Final Report Summary - D-CODE (DC/DC COnverter-based Diagnostics for PEM systems)
Executive Summary:
Note: the figures cited in the text are available in the attached pdf file that contains the same text reported below.
1.1. D-CODE Executive summary
Among the main technological challenges of PEMFC System (PEMFCS), long duration, reliability and optimal working operations of both stack and auxiliaries are issues of primary importance. Solving these problems will give a great opportunity for a decisive breakthrough towards mass diffusion of the current PEM technology. The need for advanced diagnostic tools is of prominent significance to improve the lifetime of PEMFCS. Diagnostic algorithms can detect malfunctioning of components and give alarms for faults that may hinder correct operations or induce system’s failures. The D-CODE project has developed an innovative monitoring procedure to support the diagnosis of PEM fuel cell stack for fault identification. The stack electrochemical impedance spectrum (EIS) is proposed for on-line diagnosis during on-field operations. To synthesize in a statement: the project implements the idea that a single measure (i.e. EIS) provides holistic information on the stack. Indeed, EIS is the most reliable diagnostic tool for FC and the project D-CODE moves the EIS-based diagnostic from laboratory to on-field. Therefore, the main issue was to transpose the EIS from lab-scale to on-board as well as to develop innovative FC stack monitoring to change radically the concept of on-line diagnosis. In perspective, on-board EIS may support stack degradation level analysis as well as lifetime prediction and effective control to mitigate the consequences of degradation due to chemical, thermal and humidity effects as well as load cycles or faults.
To cover most of the prospective stationary use of PEMFCSs (µCHP, backup and APU), two PEM technologies were considered, namely low temperature (LT) and high temperature (HT) for both high voltage (HV) and low voltage (LV) applications. Moreover, an additional device with a dedicated diagnostic function was developed to monitor the status of the stack when the FCS is switched-off. Towards the objective of implementing the EIS-based diagnostic functions on board of a FCS, the project addressed the development of new hardware and diagnosis algorithms that were embedded into the FCS. Moreover, a comprehensive experimental campaign was performed to identify the behaviour of the FC stacks under abnormal operating conditions to look into the influence of auxiliary devices (BoP) on stack impedance, as well as to validate the developed diagnostic algorithms. Another objective was building a diagnostic tool which may be implemented into any FCS with the minimum change of both hardware and software. Therefore, flexibility and modularity were also addressed during hardware and diagnosis algorithms design.
Two DC/DC converters for LV and HV uses were built and connected to an EIS-oriented controller board, which may be either implemented in any converter or connected via communication interface. The main feature of such a board is the control of voltage and current converter loops to inject the current stimuli to perform the EIS. This configuration guarantees the maximum flexibility to make the EIS-based diagnosis easy to install on any PEMFCS by substituting the conventional DC/DC converter or by adding the EIS control board. Three diagnostic algorithms were implemented making use of two approaches, namely model- and knowledge-based. The validation of these algorithms was successfully performed achieving the isolation of the following abnormal operations: air and fuel starvation, flooding, drying. A diagnostic tool was built and embedded into the controller of the FCS tested during the project. It was easily interfaced thanks to the use of conventional software functions and standard communication protocols. This guaranteed the development of a diagnostic tool that has industrial standards and marketable potentialities and is potentially implementable in any FCS.
Thanks to both the new DC/DC converter hardware and diagnostic algorithms, faults and potential failures associated to electrochemical processes, components faults (e.g. blower, power electronics, actuators) or having external origin (erroneous control, critical load) can be detected on-line while the system runs. Moreover, the main D-CODE project concept can be easily applied to other FC technologies (e.g. SOFC) thanks to the “universal” hardware and generic theoretical and experimental methodologies implemented.
Project Context and Objectives:
1.2.1. Introduction
The leading objective of the project D-CODE was the implementation of the electrochemical impedance spectroscopy (EIS) for on-line monitoring and diagnosis of PEM FC during on-field operations. The D-CODE project aimed at developing an innovative diagnostic procedure to detect PEM fuel cell faults and support stack degradation level analysis. To synthesize in a statement: the project implements the idea that a single measure (i.e. EIS) provides holistic information on the stack. Indeed, EIS is the most reliable diagnostic tool for FC and the project D-CODE moves the EIS-based diagnostic from laboratory to on-field. Therefore, the main issue was to transpose the EIS from lab-scale to on-board as well as to develop innovative FC stack monitoring to change radically the concept of on-line diagnosis. Moreover, an additional diagnosis function was developed to monitor the status of the stack when switched-off. The project plans were conceived in such a way to analyse and develop each component of the diagnosis tool and this was mostly achieved during the first part of the activity, then in the second period all elements of hardware and algorithms were tested and assembled for the final validation.
A diagnostic tool made of hardware and software parts and based on the electrochemical impedance spectroscopy concept was built; it gives direct and meaningful information about the FCS status during real operations. The project achieved its targets whose main results are i) the application of the EIS-based diagnosis for on-field application and ii) the development of affordable diagnostic equipment. It was demonstrated the possibility to move the EIS from controlled, clean and costly context (i.e. laboratory) towards unattended, contaminated and cheap environment (i.e. on-field). The D-CODE diagnostic concept relies on the combination of power electronics hardware and diagnostic algorithms, whose functions can be easily extended to other applications of PEM fuel cell systems and, in perspective, to all FC technologies as well. The D-CODE project’s outcomes are expected to improve management and operational capabilities of PEM FCS as well as to increase maintenance time and reduce degradation rate. These achievements are crucial and will foster the deployment of PEM fuel cells for on field use.
As shown hereinafter, the objectives were achieved with satisfactory results. Such a positive outcome was obtained thanks to a good balance of complementary skills brought by the partners that cover all relevant areas: industry, innovation, research, and academia. This guaranteed the development of a diagnostic tool that has industrial standards and marketable potentialities and a Technology Readiness Level of 4 (i.e. TRL 4 – technology validated in lab) .
1.2.2. Context
In the last years PEM fuel cells have demonstrated great potentialities for automotive and stationary applications. Research and industry communities tackled significant technological problems with remarkable achievements that make the PEM fuel cells outstanding solutions for future energy and environment challenges. Materials and production technologies have guaranteed the attainment of notable performance at costs that are getting closer to those of the conventional energy conversion systems. Moreover long duration, reliability and optimal working operations of both stack and auxiliaries are issues of primary importance and further efforts are required for their improvement. The classical approach to face these problems relies mostly on basic research (e.g. new materials) with massive recourse to expensive and long experimental campaigns, often driven by trial and error analysis. A step forward in solving these problems is the exploitation of advanced methodologies available in the broad area of applied research. Among others, advanced control and diagnosis solutions – supported by ad hoc experiments and modelling approaches – would provide suitable solutions towards the achievement of more performing and durable FC. This paradigm would benefit also other FC technologies (e.g. SOFC).
Following the trend seen in the high tech industry, a reduction in the actual empiricism together with more intelligent tools installed on-board may lead to better performance and longer lifetime. The project D-CODE faced these problems from the point of view of monitoring and diagnosis, which are the first efforts which can impact on performance and lifetime of PEMFC system. Diagnostic algorithms can detect components malfunctions and give alarms for faults that may hinder correct operations or induce system’s failures. The work has progressed its main activity to develop an advanced monitoring and diagnostic tool making use of electrochemical impedance spectroscopy (EIS). EIS can provide significant insights into the stack status and can support effective control to mitigate the consequences of degradation due to chemical, thermal and humidity effects as well as load cycles or faults.
The EIS is a steady-state technique able to observe phenomena occurring in electrochemical systems. On a FC, small-amplitude sinusoidal perturbations are superimposed over the nominal operating current at several discrete frequencies. Each injected perturbation induces a change into the output voltage that is out-of-phase with respect to the current perturbation signal. The electrochemical impedance is the frequency-dependent proportionality factor between the voltage response and the current signal; its real and imaginary components may be plotted to generate the Nyquist plot with characteristic semicircle shapes. Furthermore, the EIS is the response of the system to codified and reproducible inputs (stimuli) generated on purpose under dynamic operations. The electrochemical impedance provides the most meaningful information about the cell electrochemical processes providing unique information on the status of the FC.
It is worth noting that the impedance spectrum brings much information with only one measurement; this is in contrast with conventional techniques, which need more than one measurement to generate a single feature . Indeed, the techniques available today rely on either simple heuristic approaches or burdensome mathematical models combined to many experimental measurements. Models and maps are used to generate reference (i.e. normal) system features to be compared to the actual ones, extracted from a real working system. To make useful diagnosis, redundant features must be computed and interpreted.
1.2.3. Project objectives
The objective of the project D-CODE was the development of a diagnostic tool, made of hardware and software parts and based on the electrochemical impedance spectroscopy concept, which gives direct and meaningful information about the FC status during real operations. At methodological level, the project aimed at reducing the empiricism found in most control and diagnostic algorithms and devices. Therefore, a more general objective was the exploitation of advanced techniques introducing in the field those innovations and approaches relying on advanced concepts and theories.
The main innovation of D-CODE is the exploitation of the Electrochemical Impedance Spectroscopy during real operations of fuel cell system (FCS). The on-board implementation of EIS is a breakthrough innovation for FC area because of its potentiality for the monitoring of the FC systems while running on-field. This was the main technological objective, which has been achieved making use of conventional hardware and manufacturing technologies at affordable industrial costs.
At the beginning of the project two main targets were set:
1. The first target dealt with the development of the power stage and the control strategy of DC/DC converter to measure the stack impedance spectrum. Towards this goal two devices for high and low voltage applications were built making use of power electronic methodologies. Furthermore, high flexibility has been achieved through the development of an EIS-oriented controller board, which may be potentially implemented on conventional converter as well. Also the integration of EIS functions through signal injection was achieved together with filtering and data collection for EIS analysis. The data exchange with FCS controller was successfully tested and integrated within the main management functions.
2. The second target was the derivation of a set of indicators from the spectrum to quantitatively evaluate on-board either a drift or a difference between the actual status and nominal/expected performance of the stack/system. Two monitoring approaches, namely model- and knowledge-based were developed and tested. The diagnostic algorithms provide a quantitative framework that reduces the empiricism of the analysis performed today to derive the diagnosis of FC at both on-field and lab scale.
To achieve these goals several technological issues were addressed, among others:
• Derive FC degradation/diagnostic information while the system runs.
• Implement EIS functions to control the DC/DC converter.
• Embed the EIS-based diagnostic tool into the FCS control unit.
A relevant advantage gained by the project concerns with future on-field use that would benefit from the functional scheme where single components are linked together (see figures 1.1 1.3 to 1.7 and 1.19 to 1.21). According to that scheme, two converters (for LV and HV use) and one EIS board, which is able to interface with both converters, were designed and several prototypes built. In order to better demonstrate the portability of the developed algorithms, characterization and on-line validation have been performed on another commercial fuel cell system, the Nexa Ballard Power Module. Furthermore, the analysis on LTFC in switched-off conditions was considered worth of attention for backup system as well as for other applications that may benefit from the diagnosis in such conditions (e.g. portable, automotive). A dedicated device was designed and built for the purpose of switched-off diagnosis; it is worth mentioning that the application of such a device to other FC technologies (e.g. SOFC) is straightforward. Thanks to these enhancements the industrial implementation of the diagnosis tool is flexible and affordable.
A list of the technical objectives achieved is reported below:
• Characterize LT and HT FC systems in different operating conditions.
• Test in climate chamber the LT PEMFC when switched-off.
• Provide experimental data ad electrochemical impedance spectra for diagnositic algorithms development and validation.
• Design and manufacture the HV and LV DC/DC converters accounting to industry standards.
• Develop the EIS board to control both HV and LV converters and embed the spectrum analysis functions.
• Implement signal injection and data acquisition for the Impedance Spectroscopy analysis.
• Identify methodologies to infer on the FC status from model- and knowledge-based approaches for on-line application.
• Implement and validate model- and knowledge-based monitoring and diagnostic algorithms.
• Develop a technique to infer on the FC status when switched-off.
• Implement the interface for data exchange among converter, EIS board and FCS management system.
• Test the hardware and verify the whole assembly (FC & HW).
• Validate the monitoring schemes under normal operating conditions.
• Test the faults isolation to identify components and processes that operate under faulty conditions; infer on the stack status.
• Improve algorithms and system management.
The attainment of the technical objectives listed above has generated several outcomes with large benefits at industry and research levels; towards these goals some actions were taken to spread the knowledge gained thanks to the achievement of the following objectives:
• Disseminate the project idea via web-based technologies.
• Disseminate the results through scientific publications
• Organize two workshops to
o discuss initially with scientific and industrial communities and then to
o present the EIS-based diagnostic tool and gather suggestions from international scientific and industrial communities.
• Implement student exchange programs among the partners.
• Initiate the process for results’ exploitation towards production and commercialization.
• Participate to an international event (Hannover Fair 2014) for final dissemination activity.
• Results exploitation and project follow-up.
Project Results:
1.3. Main S&T results
The D-CODE project structure was conceived on the basis of three main activities: i) the experimental characterization of LT and HT PEMFC systems to deepen the influence of control and BoP components on the EIS spectrum; ii) the manufacturing of hardware components (i.e. power electronics and FC system) with specific features so as to proper implement diagnostic algorithms based on EIS measurements; iii) the development of diagnostic tools exploiting the information embedded into EIS spectra in order to detect and isolate system components deviation from normal operating conditions. The project was structured into four research work packages (see figure 1.2) two more WPs were devoted to project management and to the dissemination of project results, respectively. The work package 2 (WP2) focused on the experimental activity to characterise both LT and HT fuel cell systems and to test diagnostic hardware and related strategies. WP3 was dedicated to the hardware design and realization of both the DC/DC power converters and the EIS command board. The activity of WP4 led to the development of the algorithms for the diagnosis. WP5 was finalised to the development of the management system to drive the DC/DC converter for EIS purposes and to improve the FCS control strategies taking advantages of the information gathered from the EIS-based diagnosis.
1.3.1. Project general description
1.3.1.1. Fuel cell systems characterization
A large number of characterization tests were conducted on the FC systems to increase data significance and repeatability; the main activity was conducted in the frame of the project, further data set were provided by Fraunhofer (ISE) and Aalborg University . For low temperature FC a conventional and a modified Dantherm DBX2000 system were used, in the latter the BoP is electrically decoupled from the stack. The experimental campaign also investigated the effect of BoP components and grounding on the impedance measured on the stack. Furthermore, the modified system was tested in faulty conditions, i.e. some parameters in the on-board controller were changed to induce faults in the FC stack operation. The DBX2000 modified version was also tested with the external high voltage DC/DC converter manufactured by CIRTEM (Figure 1.3-b) as well as the low voltage one purchased from MICROPI and built according to the design made by UNISA and shown in Figure 1.3-a. Both converters were coupled with the EIS external board designed and built by BITRON.
Dedicated tests were designed and carried out at EIFER to record EIS spectra on the stack using only the converter. An on-line validation procedure was developed and the diagnostic tools were validated with the system running in normal as well as in faulty conditions, while the converters were used to record EIS measurements on the stack. The quality of the result is very good considering the complexity of the configuration and the difficulty of carrying out impedance spectroscopy measurement on a relatively high power system, such as the DBX2000. Figure 1.4 reports the spectra measured through the EIS board at different loads for the assembly of Figure 1.3-a (i.e. DBX2000 modified and LV converter). The EIS were measured at low, mean and high currents; in Figure 1.4 the points at low frequencies are on the right side of each spectrum. At 5A the EIS exhibits some noise at low frequency, as the amplitude of the voltage AC current is very low and then very sensitive to any perturbation. At 40A, the EIS measurement has high quality on the whole range of frequencies. This is because the higher the current, the higher the signal-to-noise ratio.
In order to better demonstrate the portability of the developed algorithms, characterization and on-line validation were performed at UFC on another commercial LT fuel cell system, the Nexa Ballard Power Module. Figure 1.5 shows the hardware assembly with the same configuration of Figure 1.3-b the HV converter is another prototype built by CIRTEM.
The high temperature PEMFC system assembled by Serenergy A/C was characterized at EIFER with a protocol similar to the one used for the low temperature Dantherm BDX2000 system. The tests were first performed with the control parameters default values and then repeated with different stack temperature, in turn, maximum and minimum values. The HT PEMFC system was also tested with the low voltage DC/DC converter coupled with the EIS external board and the stack impedance was measured (Figure 1.6).
To meet the requirements for the diagnosis of backup systems in switched-off condition, which is a critical task to guarantee system serviceability also for other FC applications, Dantherm in collaboration with the Aalborg University carried on the characterization in climate chamber of stacks in switched-off conditions. The new tests included not only humidity but also temperature cycles. To perform this activity UNISA designed an impedance analyser easily to embed into a DC/DC converter, which offers the possibility to perform EIS measurements system in switched-off conditions and detect the polymeric membrane humidification level. The Figure 1.7 shows the device developed by MICROPI connected with the FCS during the tests performed at EIFER.
The activity concerning the switched off analysis consisted in the measurements of the impedance on the DBX2000 modified system in switched off conditions at ambient temperature and humidity (see Figure 1.7). An example of the obtained results is given in Figure 1.8. The left diagram shows the impedance module versus the investigated frequency range. The right diagram shows the Nyquist plot of the same measurements. Both the results here presented are in accordance with the ones provided by Aalborg University (see Bidoggia et al.). Such results confirm the capability of the impedance analyser developed within the project and the feasibility of performing such a measure on the FC system for on field installations.
The table 1.1 shows a synopsis of the six FC arrangements (system and stack) investigated during the experimental campaign performed at EIFER and UFC; in the table the two supplementary sets made available from Frauhnofer ISE and Aalborg University are also reported for the sake of completeness. The cases shown in figures 1.3 through 1.7 are referenced into the relevant boxes; cross marks denote the other arrangements. It is worth mentioning that at the beginning of the project only two configurations were conceived for LT and HT FC, i.e. the LT DBX standard and HT Serenergy respectively.
1.3.1.2. Hardware design and manufacturing
The EIS board, developed by BITRON, is shown in Figure 1.9; this device is able to control both the HV and LV DC/DC converters, both shown in Figure 1.3. The EIS board output, where the variable frequency stimulus is produced by the firmware, is connected to either the voltage or current loops of the DC/DC converter, depending on the frequency range in which the stimulus lies. The two inputs to the EIS board are the stack current and voltage waveforms: they are processed, at each frequency of the generated stimulus, in order to produce an output file containing the real and imaginary parts of the stack impedance. During the signal processing, the firmware takes into account, among other things, the transient effect generated by the change of the frequency of the applied stimulus and the presence of noise, e.g. at switching frequency, affecting the two waveforms. The flexibility of the measurement is improved by the use of a configuration file, which is read by the EIS firmware as a first step of the measurement; it includes the list of frequencies (0.01 Hz - 2 kHz) to be analysed and the characteristics of the stimulus to be produced, frequency-by-frequency. The file is transferred to the EIS board by the supervisor (see element 3 of Figure 1.1) that runs on the FCS control unit. This allows to easily customizing the sequence of the measurement process.
The two DC/DC converters were developed in order to comply with different FC applications: stationary grid-connected, by means of the HV converter and backup, stand alone, by means of the LV converter. Both converters were designed to process the power produced by a 1.9 kW Dantherm system with a working voltage within the range 28-56V. As a consequence, the HV converter required a step-up topology, developed through a non-isolated solution in order to ensure the highest possible efficiency. The LV converter, which has to supply a 48V load, is based on a step-up-down SEPIC topology with coupled inductors. The project has also contributed to the development of a precision impedance analyser (see Figure 1.7) aimed at performing the EIS analysis of the stack in switched-off conditions. The analyser is able to cover the frequency range 1kHz-200kHz and can apply a sinusoidal stimulus from 1 mA to 1 A.
To perform the EIS measurements on a FC integrated in a complete system some guidelines were followed. Among others the main issues were those related to a proper interface and data exchange among Supervisor, Monitoring & Diagnosis modules and the FCS controller whose connections are shown in the Figure 1.1. Moreover, the impedance measure should not be influenced by any electrical component except the stack. A deep analysis of the measured spectra revealed several issues related to perform EIS measurements directly on a fully integrated fuel cell system. These issues were primarily caused by the fact that the fuel cell stack cannot be electrically isolated from the balance of plant and the system controller, since there is a circuitry hindering the EIS measurements of the isolated stack. Therefore, a modified fuel cell system had to be developed by Dantherm. The most difficult task in developing the modified system consisted in decoupling the DC/DC converter and the stack, as this modification required a major revision of the Dantherm DBX2000 embedded system software in order to make sure that the system was able to run reliably without compromising the basic safety of the system, in which the DC/DC plays a major role. This part of the work was inspired also by the need for having a system that complies with industrial standards. Other simpler solutions would have been possible for the purpose of testing only, but without achieving a complete validation at lab level. Now a solid study on that issue is available and will surely help during future actions oriented towards further implementation of the proposed methodology on field.
The schematic for the standard and the modified DBX2000 system is illustrated in Figure 1.10. From such representation it can be observed that the DC/DC converter was entirely electrically decoupled from the fuel cell stack (see the scheme on the right), and that the stack power was routed directly out through the main DC connectors of the control box. Also a separate stack current measurement had to be added.
In order to support the testing to be performed on the modified DBX2000 unit, it was chosen to develop a new monitoring and logging application for the system. The application communicates with the system via CAN-bus. The new monitoring application supported access to all system variables, and also supported logging these variables during testing on the system. On top of that, the application also provided a graphical interface, with the possibility of plotting charts for different system variables (see Figure 1.19).
For the purpose of integrating EIS enabled DC/DC converters, an architecture was developed, where the integration of the individual necessary devices was performed via a target PC, which served as the link between the fuel cell main controller, the DC/DC converter (with EIS board connected) and the diagnostic algorithms, with embedded EIS functionalities. The main advantage of this approach consisted in its flexibility in accepting codes or applications written in different programming languages. The solution would also virtually support any communication protocol chosen for communicating with the EIS-enabled DC/DC converter. Furthermore, the target PC had much more computational power compared to an embedded controller. The functional links among the components is presented on Figure 1.11. This scheme focuses more on signals communication if compared to that of Figure 1.1 which reports on the components functions.
On the DBX2000 system’s side the solution required a major modification to the software in the system’s embedded controller as well as a modification to the Labview monitoring and logging application for the system. Using this architecture it was possible to integrate the EIS enabled DC/DC converters as well as EIS measurement hardware developed in the project.
1.3.1.3. Diagnostic tools development and implementation
The model-based diagnostic tool developed by UNISA consists of two tasks: the identification of the parameters of an Equivalent Circuit Model of the stack, performed by exploiting the experimental spectra, and the detection and isolation of abnormal conditions via comparison of the identified parameters to their reference values. Such reference values are mapped as function of FC nominal operating conditions and modelled via black-box models specifically developed to estimate normal conditions values. The EIS spectra acquired in abnormal conditions were used for algorithm validation. An example of the online application of this diagnostic algorithm for DBX2000 system (see Figure 1.2) is shown in Fig. 1.12 where the values attained by one model’s parameter (e.g. the membrane resistance) are compared with the nominal ones. In this figure, the coloured areas highlight the parameters deviation (abnormal operations) at different operating currents. The isolation process is then performed making use of a signature matrix, which accounts for the information brought by four ECM model parameters (see the bottom side of the figure).
A detailed scheme of the on-line diagnostic procedure was conceived, with all the required phases, from the fault detection to the isolation. Moreover, among all the possible faults taken into account in a preliminary analysis, only drying and flooding were considered as the ones currently inducible for a validation of the proposed algorithm. Figure 1.13 illustrates the comprehensive model-based diagnostic tool proposed and developed by UNISA.
For the development of the aforementioned ECM parameters models, the knowledge of current load and stack temperature is essential. This requirement led to the development of a specific communication protocol between the UNISA algorithm and the DBX2000 system control software, so as to acquire the current and temperature measurements for each EIS spectrum. Analytical regressions have been computed for the system state variables (i.e. stack voltage, stack temperature and blower power) estimation in normal operating conditions. These regressions are essential to perform a plausibility check to understand if the system variables are at the expected operating condition.
To retrieve the essential information from the EIS board, the UNISA diagnostic algorithm reads from a text file the EIS measured variables (i.e. frequency, impedance real part, imaginary part, module and phase). These information are treated in order to identify the spectrum and compute the ECM parameters values.
Two knowledge-based diagnostic algorithms were developed by UFC: a fuzzy-logic one and an advanced neural-network based one. The former consists of a Fuzzy Clustering algorithm for automatic learning from experimental data and a Fuzzy Inference System (FIS) for decision-making. It was validated online on two different LT stacks (see Figures 1.3 and 1.5). The Fuzzy Clustering algorithm consists of four steps: feature extraction and selection, fuzzy clustering, and diagnostic rule design, showed in Figure 1.14 where the upper part is related to the off-line training and the lower part to the online testing. Since it combines both fuzzy clustering and fuzzy logic, it is thereby called “double-fuzzy” methodology.
Another diagnostic algorithm was developed, it is based on liquid state machine (LSM) which is an advanced architecture evolved from the traditional recurrent neural network. The main idea is to create an internal network randomly and keep the connections fixedly; the only part to be trained is the output weights .The main structure is shown in Figure 1.15. The stimulated stack voltage at a low frequency is used as the neural network’s input and the health status is used as its output. The main considerations of utilising stimulated stack voltage as inputs are: (1) the stack voltage is a credible reflection of FC degradation; (2) when stimulated by an AC sinusoidal current, it includes also the information about ac impedance, which is believed to be discriminant in diagnosing flooding and drying.
The experimental results demonstrate the algorithms high accuracy rate (100%), computational efficiency (less than 5s), easy implementation and good portability. Besides the LT type, the methodology is verified on the HT stack (Figure 1.6) which further shows its capability for fault diagnosis of PEMFC stacks regardless LT or HT types. Figure 1.16 shows an example of UFC algorithm for fault-diagnosis application to NEXA System. Three different operating conditions were analysed on the NEXA system (normal, air starvation and air surplus) and the stars represent the location of each operating point with respect to the three considered regions (i.e. clusters).
The diagnostic methodology proposed by EIFER is a data-based approach. First, a set of statistical reliable data in different conditions (i.e. flooding, drying and normal) needs to be taken. Second, the conditions of the fuel cell are defined as an output and associated with the referring states. Third, an Adaptive Neural Fuzzy Interference System (ANFIS) is trained with the prepared data set. The Adaptive Neural Network (ANN) teaches a Fuzzy Interference System (FIS) with a predefined structure and compares the output of the FIS with the known output of the data set. One run of the ANN algorithm is called an epoch. During this training phase, several epochs are required for proper results of the FIS. In general, a data set is randomly divided in sets for teaching the system and sets for testing the trained FIS. The error between the known output and the output of the FIS can be used as an assessment of the tool performance. Ultimately, the trained FIS represents a model of the states inside the cell. It is worth remarking that any data-based approach can only identify conditions with noticeable differences in the measured data sets. Another important point is to avoid and filter noise as much as possible. For that reason, different approaches for data pre-treatment were investigated. The algorithm scheme is presented in Figure 1.17. In the Figure 1.18 the results for the NEXA system are reported showing three cases, namely normal, drying and flooding operations for the NEXA System running at 10 A.
In Figures 1.19 1.20 and 1.21 the interfaces (see item 6 of Figure 1.1) of the three diagnostic tools developed by UNISA, UFC and EIFER are shown. It is worth remembering that the diagnostic tools developed by the D-CODE project may detect the following abnormal operations:
• Air starvation
• Fuel starvation
• Flooding
• Drying
Potential Impact:
1.4. Expected impact
1.4.1. Expected outcomes and target achieved
With the aim of analysing the future impact of the results achieved by the Project D-CODE, it is worth describing its outcomes with respect to those envisaged in both annual and multi-annual implementation plan documents (AIP and MAIP) .
With respect to the RTD priorities of the FCH-JU 2009 call, the project results will have a remarkable impact towards the key objective of “… placing Europe at the forefront of fuel cell and hydrogen technologies worldwide and enabling the market breakthrough of fuel cell and hydrogen technologies, thereby allowing commercial market forces to drive the substantial potential public benefits …”. Indeed, the proposed diagnostic tool represents a step forward to enhance the performance of FC systems, guarantees the advancement of the technology and improves the commercial competitiveness of the systems that might embed it. Although developed for PEM, the diagnostic tool might be extended to any FC technology through the methodological process outlined by D-CODE. Looking at the industrial scenario of the FC market, the proposed solution contains several technological innovations (e.g. on-board EIS, model- and knowledge-based diagnostic algorithms) that neither the EU industries nor their non-EU competitors have yet in the portfolio.
Moreover the AIP focused on applied research activities “ … directed towards developing components and sub-systems with improved performance, durability and cost for all three technologies in order to achieve system application readiness … ”. The project’s approaches, methodologies and applications belong to the broad area of applied research; this complies with the objective of advancing the actual technology through the enhancement of components (here component means either physical device or software) that are able to improve FCS. The objectives achieved have a direct impact on both performance improvement and durability; the former is linked to the capability of forecasting FC status thanks to the precise knowledge of the stack status via EIS monitoring, whereas the latter deals with the correct and timely detection of faults, whose knowledge may support the controller to activate precise countermeasures to maximize the system readiness.
The analysis of the results summarised above permits to state that the diagnostic tool designed, built and tested during the project fits within the boundaries set by the multi-annual implementation, whose one objectives was the achievement of “… new or improved materials as well as reliable control and diagnostics tools both at component and system levels … directed towards developing components and sub-systems … leading to step change improvements over existing technology in terms of performance, endurance, robustness, durability and cost for all technologies …”.
The results and the objectives just reviewed deals with technical issues whose impact spreads towards social and economic areas. Once implemented on-board of FC, the diagnostic tools will advance their performance contributing to increase the overall effectiveness of these energy conversion systems, as well as to make them more attractive. Indeed the solution proposed has a positive impact on the operational costs because of its effects on PEM stacks reliability and lifetime, which can be improved thanks to the monitoring and diagnosis tools. These benefits might be gained by all PEM applications (CHP, APU, backup) thanks to the flexibility of diagnostic tool, which can also be implemented on other FC technologies (e.g. SOFC) by following the methodological steps proposed. Moreover, the electrochemical impedance spectrum can be measured via low cost hardware, which was built following industrial standards. These points represent further advantages to reduce development costs and consequently purchasing price. Therefore, future commercialization can be easily promoted increasing the number of FC systems.
Finally all the above technical and economic advancements achievable through the implementation of the diagnostic tool will increase the diffusion of the fuel cell with an indirect impact on the major environmental issues.
1.4.2. Dissemination
The results of D-CODE have been disseminated through different ways to increase the awareness of both the scientific and industrial communities. The communication strategy was based on three pillars: workshops, website and scientific communications (i.e. publications).
According to the project schedule, two workshops were successfully organised. The former was held on June 2012 during the French Research Network on Fuel Cells meeting. More than 50 researchers attended the two special sessions on FCS diagnosis, whose 9 lectures were given by the invited speakers. The coordinator’s keynote speech was given during a plenary session with about 250 attendants (see Figure 1.4.1 left). A second workshop was held in Karlsruhe, Germany (April 2013). It was organized in the framework of the Conference FDFC (Fundamentals and Developments of Fuel Cell) and gathered 81 attendants (see Figure 1.4.1 right). Eleven presentations were delivered, followed by interesting comments and questions from the audience. The last event organized focused on the dissemination towards industrials. The D-CODE results were presented at the Hannover Fair in April 2014, during 1 week (see the pictures of Figure 1.4.2). The purpose was to show in a booth the hardware (i.e. the fuel cell system, the new converters, the control board, the diagnosis algorithm). The project’s booth was located in the group exhibit “Fuel Cell and Battery Exhibition”, the Europe's largest exhibition in this field. In the booth, the hardware developed in the project was exhibited and the results presented on posters (see Figure 1.4.3). Brochure and flyers were also distributed to the visitors. Sixty contacts were established with visitors coming from 20 different countries. A presentation of the results was done by the coordinator in the technical forum, the video is available on the Fair website as well as on “YouTube” (see the bottom right corner of the Figure 1.4.3). The participation to the Fair was a valuable opportunity for the consortium to meet several important actors in hydrogen and fuel cell activities.
The official website of the project is available at the URL https://www.d‐code‐jti.eu. On the same website all main achievements of the project are presented to disseminate the results and to increase the awareness of the industrial and research communities about the D-CODE project. The website has been regularly fed with the main achievements of the project: photos of the developed hardware, public milestones of the project, information on the events organized, references of the scientific publications co-authored by the partners and public documents available for downloading. As of July 12rd 2014, a search on the Google database using the keys sequence “on-line EIS diagnosis PEM fuel cell” provided a list of relevant links and the project D-CODE was in the first page at 5th rank.
To disseminate the scientific knowledge along the project, seven papers have been published through different ways. First, two articles have been published in a major journal which is International of Hydrogen Energy (Impact Factor: 3.55). Four papers have also been presented in major conferences publishing proceedings (3 in FDFC 2013, 1 in IEEE-IECON 2013). An article is in final revision in another high level journal, Journal of Power Sources (IF: 4.675) and should appear in autumn 2014. The list of the papers is reported below:
1. Zhixue Zheng, Raffaele Petrone, Marie-Cécile Péra , Daniel Hissel, Mohamed Becherif, Cesare Pianese, Nadia Yousfi-Steiner, Marco Sorrentino, “A review on non-model based diagnosis methodologies for PEM fuel cell stacks and systems”, International journal of hydrogen energy, vol.: 38, number: 21, pages: 8914-8926, DOI: 10.1016/j.ijhydene.2013.04.007.
2. Raffaele Petrone, Zhixue Zheng, Daniel Hissel, Marie-Cécile Péra, Cesare Pianese, Marco. Sorrentino, Mohamed Becherif, Nadia Yousfi-Steiner, “A review on model-based diagnosis methodologies for PEMFCs”, International journal of hydrogen energy, vol.: 38, number: 17, pages: 7077-7091, DOI: 10.1016/j.ijhydene.2013.03.106.
3. Angelo Esposito, Bastian Ludwig, Cesare Pianese, Soeren Boedker, "Experimental Characterization of a PEMFC System with Electrochemical Impedance Spectroscopy", 5th Fundamentals & Development of Fuel Cells Conference (FDFC), April 16-18, 2013, Karlsruhe, Germany.
4. Zhixue Zheng, Marie-Cécile Péra, Daniel Hissel, Mohamed Becherif, “Diagnosis of proton exchange membrane fuel cell （PEMFC） stack based on fuzzy clustering”, 5th Fundamentals & Development of Fuel Cells Conference (FDFC), April 16-18, 2013, Karlsruhe, Germany.
5. Raffaele Petrone, Zhixue Zheng, Daniel Hissel, Marie-Cécile Péra, Cesare Pianese, Marco. Sorrentino, Mohamed Becherif, “Implementation of EIS measurements on an embedded commercial system”, 5th Fundamentals & Development of Fuel Cells Conference (FDFC), April 16-18, 2013, Karlsruhe, Germany.
6. Zhixue Zheng, Raffaele Petrone, Marie-Cécile Péra, Daniel Hissel, Mohamed Becherif, Cesare Pianese, “Diagnosis of a Commercial PEM Fuel Cell Stack via Incomplete Spectra and Fuzzy Clustering”, 39th Annual Conference of the IEEE Industrial Electronics Society (IECON) November 10-13, 2013, Vienna, Austria.
7. Zhixue Zheng, Marie-Cécile Péra, Daniel Hissel, Mohamed Becherif, Kréhi-Serge Agbli, Yongdong LI, “A double-fuzzy diagnostic methodology dedicated to on-line fault diagnosis of PEMFC stack”, Journal of power sources, February 2014 – available on line, DOI: 10.1016/j.jpowsour.2014.07.157.
At educational level 3 PhD students were involved in the project, moreover several Bachelor and Master students worked on the D-CODE topics. Zhixue Zheng (UFC) has prepared her Doctoral thesis in the framework of D-CODE. She defends her PhD thesis in September 2014, at UFC, Belfort, France in front of an international jury (French, Spanish, Swiss and Chinese members in the jury). Raffaele Petrone (UNISA) was supervised by UNISA and UFC and defended its Doctoral thesis on his work developed in the framework of D-CODE. He’s currently performing a post-doc at UFC on a related subject. Pierpaolo Polverino (UNISA) focused its Doctoral work on SOFC, his experience on software and hardware integration allowed the development of the fault identification algorithms and its interfacing to PEMFC; he’s now with UNISA as research assistant. Some students at Bachelor and Master levels developed their final work in the frame of the D-CODE project. The teaching staffs of UFC and UNISA deliver several courses on energy conversion technologies for undergraduate, graduate students, summer schools and tutorials. The D-CODE results will feed the fuel cells sections of these courses.
The dissemination of the results will continue after the end of the project as a part of the main communication activity of each partner, either research institution or industry. Indeed further scientific publications will be issued as follow-up of the work done during the project. It is envisaged that the results achieved will be further checked and improved to make them interesting for publication in International Journals. Moreover, the work performed has set solid bases on which further research work can be built leading to more publications as well as other research projects, which in all cases will acknowledge the role and the results of D-CODE. Academic partners will include in their teaching activity the study performed within D-CODE and the experience made. This will spread out at a larger extent the concepts of diagnosis as well as of the on-board application of EIS. At industrial level the work performed is part of the technical know-how of these partners, which may also add the developed hardware in their catalogues. Therefore, several dissemination actions will be also taken as part of the normal communication strategy of each partner. Moreover, other dissemination actions could be performed if requested by the FCH-JU (e.g. annual review events).
1.4.3. Exploitation of results
The project D-CODE has demonstrated the feasibility of using a low-cost EIS-board in conjunction with a DC/DC converter to perform EIS-based PEMFC diagnosis. This enables fuel cell systems to be monitored during operation or in standby mode, contributing to an improvement in their reliability without adding excessive cost. The DC/DC converters and the EIS board are production-ready devices. The future step is to further develop the application to take it out of the laboratory and into wide scale field testing, with a view to making the solution commercially available for a wide variety of fuel cell applications.
The consortium looks forward to exploit both hardware (converters, EIS board) and software (diagnosis algorithms) outcomes. Limited but necessary improvements are required to implement the tool for on-field use. The TRL achieved by D-CODE ranks the level 4 (technology validated in lab); TRL5 (technology validated in relevant environment) can be easily achieved, even TRL6 (technology demonstrated in relevant environment) could be reached as well. Therefore a follow-up would focus on the improvement of the proof of concept towards on-field application to valorise the project results. Moreover both hardware and diagnostic algorithms could also be easily exploited at laboratory level with an effective reduction of equipment costs; this would require some changes in the implementation and in the hardware to comply with laboratory standards. Future work may support the research towards the study of other faults or failures to be isolated on-board by processing the impedance spectrum via either model- or knowledge-based approaches. Further actions are also required on the methodological side to develop an integrated experimental and computational framework for the design and manufacturing of EIS-based diagnosis tools. This would simplify the implementation on any PEMFC and may reduce the effort required for diagnosis algorithms identification and customization. An ad hoc activity might be settled for stack and system identification under mimicked faults for the development and testing of diagnosis algorithms.
The project’s outcomes will serve as a basis to provide increased knowledge on the potentialities offered by advanced diagnosis, on-field deployable diagnostic tool and related hardware. It is worth remarking the close interaction between diagnosis and state-of-health evaluations as well. The availability of advanced algorithms (i.e. EIS-based) for the condition monitoring of FC represents a first step towards the implementation of advanced control, as well as effective prognosis algorithms. Future cooperation among the partners will support significant research on PEM lifetime via advanced control (either feedback or adaptive) and the development of tools aimed at evaluating system residual lifetime as well as expectable end-of-life performance. Apart from the application to other FC technologies (i.e. SOFC), the developed methodologies and hardware can be also extended to other electrochemical devices (e.g. battery).
The D-CODE partners will exploit the know-how gained during the project along three main lines of action, namely at industrial, research and educational levels, either within cooperative or autonomous work. Cooperative activity will be exploited by some partners potentially interested in the implementation of the EIS on commercial FC. Notably DANTHERM Power may lead a group to continue the development of the actual tool for the implementation on board of their systems. In that case the work will focus on processes, methodologies and standards at industrial level to make the diagnostic tool an effective instrument to improve performance and increase the FC lifetime. This will guarantee a better competitiveness and may represent a reference for the entire community, thus fostering a widespread implementation of advanced diagnostic tools.
On the power electronics side BITRON could continue the development of the EIS board and include it in their catalogue. In such a case, the board may be purchased by customers wishing to use a low cost instrument capable of injecting sinusoidal waveforms for the purpose of analysing the dynamic response of other electrochemical devices (e.g. batteries). Electrical devices can also be considered for the dynamic response at frequencies below 1 kHz.
CIRTEM could exploit the experience achieved in the design and building of high voltage DC/DC converters for several applications other than fuel cells.
Moreover, the entire diagnostic tool (EIS board, converter and algorithms) might be installed at lab scale as low cost spectrometer. Moreover, the algorithms proposed by UNISA, UFC and EIFER are ready for on board implementation of any spectrometer and could be implemented in software packages for the EIS data analysis and FC diagnosis.
All research partners will continue the researches started in the project; the exploitation of the results will continue towards both basic and applied research leading to further cooperation opportunities. As a first step, a common paper between the D-CODE partners will be coordinated by UFC in order to synthesise the results obtained with the algorithms developed by UFC, UNISA and EIFER.
List of Websites:
1.5. The project public website
The D-CODE website was structured and implemented in the first part of the project and mostly used for the data exchange among the partners through the intranet section. During the project, as soon as new outcomes were made available and the main dissemination events were closed, the website was updated with new information. Below is reported a summary of the website structure then the description of the actions made during the second period is given.
The website was implemented by EIFER and is hosted on their server. The web address is https://www.d-code-jti.eu. It’s divided into two main sections, namely, access-free and intranet area, which is restricted to the project partners only. Overall, the web manager registered about 3000 contacts. The Figure 1.6.1 is a snapshot of the welcome page.
Public webpage
The public web page has 6 menu tabs (Home, About, Partners, Contacts, Events and Products & Documents). The sections Events and Products & Documents were updated continuously during the second period to add the information about the main dissemination events and the documents describing the results of the projects. The website has been regularly fed with the main achievements of the project: photos of the developed hardware, public milestones of the project, information on the events organized in the framework of the project, references of the scientific publications co-authored by the partners and public documents available for downloading.
In the Figure 1.6.2 the snapshot of the upper part of the section event is shown, here the main information on the participation to the Hannover Fair (April 2014) are reported. Several URLs connect to other website with further information on the hosting event. A list of the main documents describing the project and its results is available in the section documents as shown in the Figure 1.6.3.
Upon decision of the Steering Committee held on March 28, 2014 the website will be active for two years after the end of the project. This is one of the main follow up actions that may help valorising the project outcomes.
Note: the figures cited in the text are available in the attached pdf file that contains the same text reported below.
1.1. D-CODE Executive summary
Among the main technological challenges of PEMFC System (PEMFCS), long duration, reliability and optimal working operations of both stack and auxiliaries are issues of primary importance. Solving these problems will give a great opportunity for a decisive breakthrough towards mass diffusion of the current PEM technology. The need for advanced diagnostic tools is of prominent significance to improve the lifetime of PEMFCS. Diagnostic algorithms can detect malfunctioning of components and give alarms for faults that may hinder correct operations or induce system’s failures. The D-CODE project has developed an innovative monitoring procedure to support the diagnosis of PEM fuel cell stack for fault identification. The stack electrochemical impedance spectrum (EIS) is proposed for on-line diagnosis during on-field operations. To synthesize in a statement: the project implements the idea that a single measure (i.e. EIS) provides holistic information on the stack. Indeed, EIS is the most reliable diagnostic tool for FC and the project D-CODE moves the EIS-based diagnostic from laboratory to on-field. Therefore, the main issue was to transpose the EIS from lab-scale to on-board as well as to develop innovative FC stack monitoring to change radically the concept of on-line diagnosis. In perspective, on-board EIS may support stack degradation level analysis as well as lifetime prediction and effective control to mitigate the consequences of degradation due to chemical, thermal and humidity effects as well as load cycles or faults.
To cover most of the prospective stationary use of PEMFCSs (µCHP, backup and APU), two PEM technologies were considered, namely low temperature (LT) and high temperature (HT) for both high voltage (HV) and low voltage (LV) applications. Moreover, an additional device with a dedicated diagnostic function was developed to monitor the status of the stack when the FCS is switched-off. Towards the objective of implementing the EIS-based diagnostic functions on board of a FCS, the project addressed the development of new hardware and diagnosis algorithms that were embedded into the FCS. Moreover, a comprehensive experimental campaign was performed to identify the behaviour of the FC stacks under abnormal operating conditions to look into the influence of auxiliary devices (BoP) on stack impedance, as well as to validate the developed diagnostic algorithms. Another objective was building a diagnostic tool which may be implemented into any FCS with the minimum change of both hardware and software. Therefore, flexibility and modularity were also addressed during hardware and diagnosis algorithms design.
Two DC/DC converters for LV and HV uses were built and connected to an EIS-oriented controller board, which may be either implemented in any converter or connected via communication interface. The main feature of such a board is the control of voltage and current converter loops to inject the current stimuli to perform the EIS. This configuration guarantees the maximum flexibility to make the EIS-based diagnosis easy to install on any PEMFCS by substituting the conventional DC/DC converter or by adding the EIS control board. Three diagnostic algorithms were implemented making use of two approaches, namely model- and knowledge-based. The validation of these algorithms was successfully performed achieving the isolation of the following abnormal operations: air and fuel starvation, flooding, drying. A diagnostic tool was built and embedded into the controller of the FCS tested during the project. It was easily interfaced thanks to the use of conventional software functions and standard communication protocols. This guaranteed the development of a diagnostic tool that has industrial standards and marketable potentialities and is potentially implementable in any FCS.
Thanks to both the new DC/DC converter hardware and diagnostic algorithms, faults and potential failures associated to electrochemical processes, components faults (e.g. blower, power electronics, actuators) or having external origin (erroneous control, critical load) can be detected on-line while the system runs. Moreover, the main D-CODE project concept can be easily applied to other FC technologies (e.g. SOFC) thanks to the “universal” hardware and generic theoretical and experimental methodologies implemented.
Project Context and Objectives:
1.2.1. Introduction
The leading objective of the project D-CODE was the implementation of the electrochemical impedance spectroscopy (EIS) for on-line monitoring and diagnosis of PEM FC during on-field operations. The D-CODE project aimed at developing an innovative diagnostic procedure to detect PEM fuel cell faults and support stack degradation level analysis. To synthesize in a statement: the project implements the idea that a single measure (i.e. EIS) provides holistic information on the stack. Indeed, EIS is the most reliable diagnostic tool for FC and the project D-CODE moves the EIS-based diagnostic from laboratory to on-field. Therefore, the main issue was to transpose the EIS from lab-scale to on-board as well as to develop innovative FC stack monitoring to change radically the concept of on-line diagnosis. Moreover, an additional diagnosis function was developed to monitor the status of the stack when switched-off. The project plans were conceived in such a way to analyse and develop each component of the diagnosis tool and this was mostly achieved during the first part of the activity, then in the second period all elements of hardware and algorithms were tested and assembled for the final validation.
A diagnostic tool made of hardware and software parts and based on the electrochemical impedance spectroscopy concept was built; it gives direct and meaningful information about the FCS status during real operations. The project achieved its targets whose main results are i) the application of the EIS-based diagnosis for on-field application and ii) the development of affordable diagnostic equipment. It was demonstrated the possibility to move the EIS from controlled, clean and costly context (i.e. laboratory) towards unattended, contaminated and cheap environment (i.e. on-field). The D-CODE diagnostic concept relies on the combination of power electronics hardware and diagnostic algorithms, whose functions can be easily extended to other applications of PEM fuel cell systems and, in perspective, to all FC technologies as well. The D-CODE project’s outcomes are expected to improve management and operational capabilities of PEM FCS as well as to increase maintenance time and reduce degradation rate. These achievements are crucial and will foster the deployment of PEM fuel cells for on field use.
As shown hereinafter, the objectives were achieved with satisfactory results. Such a positive outcome was obtained thanks to a good balance of complementary skills brought by the partners that cover all relevant areas: industry, innovation, research, and academia. This guaranteed the development of a diagnostic tool that has industrial standards and marketable potentialities and a Technology Readiness Level of 4 (i.e. TRL 4 – technology validated in lab) .
1.2.2. Context
In the last years PEM fuel cells have demonstrated great potentialities for automotive and stationary applications. Research and industry communities tackled significant technological problems with remarkable achievements that make the PEM fuel cells outstanding solutions for future energy and environment challenges. Materials and production technologies have guaranteed the attainment of notable performance at costs that are getting closer to those of the conventional energy conversion systems. Moreover long duration, reliability and optimal working operations of both stack and auxiliaries are issues of primary importance and further efforts are required for their improvement. The classical approach to face these problems relies mostly on basic research (e.g. new materials) with massive recourse to expensive and long experimental campaigns, often driven by trial and error analysis. A step forward in solving these problems is the exploitation of advanced methodologies available in the broad area of applied research. Among others, advanced control and diagnosis solutions – supported by ad hoc experiments and modelling approaches – would provide suitable solutions towards the achievement of more performing and durable FC. This paradigm would benefit also other FC technologies (e.g. SOFC).
Following the trend seen in the high tech industry, a reduction in the actual empiricism together with more intelligent tools installed on-board may lead to better performance and longer lifetime. The project D-CODE faced these problems from the point of view of monitoring and diagnosis, which are the first efforts which can impact on performance and lifetime of PEMFC system. Diagnostic algorithms can detect components malfunctions and give alarms for faults that may hinder correct operations or induce system’s failures. The work has progressed its main activity to develop an advanced monitoring and diagnostic tool making use of electrochemical impedance spectroscopy (EIS). EIS can provide significant insights into the stack status and can support effective control to mitigate the consequences of degradation due to chemical, thermal and humidity effects as well as load cycles or faults.
The EIS is a steady-state technique able to observe phenomena occurring in electrochemical systems. On a FC, small-amplitude sinusoidal perturbations are superimposed over the nominal operating current at several discrete frequencies. Each injected perturbation induces a change into the output voltage that is out-of-phase with respect to the current perturbation signal. The electrochemical impedance is the frequency-dependent proportionality factor between the voltage response and the current signal; its real and imaginary components may be plotted to generate the Nyquist plot with characteristic semicircle shapes. Furthermore, the EIS is the response of the system to codified and reproducible inputs (stimuli) generated on purpose under dynamic operations. The electrochemical impedance provides the most meaningful information about the cell electrochemical processes providing unique information on the status of the FC.
It is worth noting that the impedance spectrum brings much information with only one measurement; this is in contrast with conventional techniques, which need more than one measurement to generate a single feature . Indeed, the techniques available today rely on either simple heuristic approaches or burdensome mathematical models combined to many experimental measurements. Models and maps are used to generate reference (i.e. normal) system features to be compared to the actual ones, extracted from a real working system. To make useful diagnosis, redundant features must be computed and interpreted.
1.2.3. Project objectives
The objective of the project D-CODE was the development of a diagnostic tool, made of hardware and software parts and based on the electrochemical impedance spectroscopy concept, which gives direct and meaningful information about the FC status during real operations. At methodological level, the project aimed at reducing the empiricism found in most control and diagnostic algorithms and devices. Therefore, a more general objective was the exploitation of advanced techniques introducing in the field those innovations and approaches relying on advanced concepts and theories.
The main innovation of D-CODE is the exploitation of the Electrochemical Impedance Spectroscopy during real operations of fuel cell system (FCS). The on-board implementation of EIS is a breakthrough innovation for FC area because of its potentiality for the monitoring of the FC systems while running on-field. This was the main technological objective, which has been achieved making use of conventional hardware and manufacturing technologies at affordable industrial costs.
At the beginning of the project two main targets were set:
1. The first target dealt with the development of the power stage and the control strategy of DC/DC converter to measure the stack impedance spectrum. Towards this goal two devices for high and low voltage applications were built making use of power electronic methodologies. Furthermore, high flexibility has been achieved through the development of an EIS-oriented controller board, which may be potentially implemented on conventional converter as well. Also the integration of EIS functions through signal injection was achieved together with filtering and data collection for EIS analysis. The data exchange with FCS controller was successfully tested and integrated within the main management functions.
2. The second target was the derivation of a set of indicators from the spectrum to quantitatively evaluate on-board either a drift or a difference between the actual status and nominal/expected performance of the stack/system. Two monitoring approaches, namely model- and knowledge-based were developed and tested. The diagnostic algorithms provide a quantitative framework that reduces the empiricism of the analysis performed today to derive the diagnosis of FC at both on-field and lab scale.
To achieve these goals several technological issues were addressed, among others:
• Derive FC degradation/diagnostic information while the system runs.
• Implement EIS functions to control the DC/DC converter.
• Embed the EIS-based diagnostic tool into the FCS control unit.
A relevant advantage gained by the project concerns with future on-field use that would benefit from the functional scheme where single components are linked together (see figures 1.1 1.3 to 1.7 and 1.19 to 1.21). According to that scheme, two converters (for LV and HV use) and one EIS board, which is able to interface with both converters, were designed and several prototypes built. In order to better demonstrate the portability of the developed algorithms, characterization and on-line validation have been performed on another commercial fuel cell system, the Nexa Ballard Power Module. Furthermore, the analysis on LTFC in switched-off conditions was considered worth of attention for backup system as well as for other applications that may benefit from the diagnosis in such conditions (e.g. portable, automotive). A dedicated device was designed and built for the purpose of switched-off diagnosis; it is worth mentioning that the application of such a device to other FC technologies (e.g. SOFC) is straightforward. Thanks to these enhancements the industrial implementation of the diagnosis tool is flexible and affordable.
A list of the technical objectives achieved is reported below:
• Characterize LT and HT FC systems in different operating conditions.
• Test in climate chamber the LT PEMFC when switched-off.
• Provide experimental data ad electrochemical impedance spectra for diagnositic algorithms development and validation.
• Design and manufacture the HV and LV DC/DC converters accounting to industry standards.
• Develop the EIS board to control both HV and LV converters and embed the spectrum analysis functions.
• Implement signal injection and data acquisition for the Impedance Spectroscopy analysis.
• Identify methodologies to infer on the FC status from model- and knowledge-based approaches for on-line application.
• Implement and validate model- and knowledge-based monitoring and diagnostic algorithms.
• Develop a technique to infer on the FC status when switched-off.
• Implement the interface for data exchange among converter, EIS board and FCS management system.
• Test the hardware and verify the whole assembly (FC & HW).
• Validate the monitoring schemes under normal operating conditions.
• Test the faults isolation to identify components and processes that operate under faulty conditions; infer on the stack status.
• Improve algorithms and system management.
The attainment of the technical objectives listed above has generated several outcomes with large benefits at industry and research levels; towards these goals some actions were taken to spread the knowledge gained thanks to the achievement of the following objectives:
• Disseminate the project idea via web-based technologies.
• Disseminate the results through scientific publications
• Organize two workshops to
o discuss initially with scientific and industrial communities and then to
o present the EIS-based diagnostic tool and gather suggestions from international scientific and industrial communities.
• Implement student exchange programs among the partners.
• Initiate the process for results’ exploitation towards production and commercialization.
• Participate to an international event (Hannover Fair 2014) for final dissemination activity.
• Results exploitation and project follow-up.
Project Results:
1.3. Main S&T results
The D-CODE project structure was conceived on the basis of three main activities: i) the experimental characterization of LT and HT PEMFC systems to deepen the influence of control and BoP components on the EIS spectrum; ii) the manufacturing of hardware components (i.e. power electronics and FC system) with specific features so as to proper implement diagnostic algorithms based on EIS measurements; iii) the development of diagnostic tools exploiting the information embedded into EIS spectra in order to detect and isolate system components deviation from normal operating conditions. The project was structured into four research work packages (see figure 1.2) two more WPs were devoted to project management and to the dissemination of project results, respectively. The work package 2 (WP2) focused on the experimental activity to characterise both LT and HT fuel cell systems and to test diagnostic hardware and related strategies. WP3 was dedicated to the hardware design and realization of both the DC/DC power converters and the EIS command board. The activity of WP4 led to the development of the algorithms for the diagnosis. WP5 was finalised to the development of the management system to drive the DC/DC converter for EIS purposes and to improve the FCS control strategies taking advantages of the information gathered from the EIS-based diagnosis.
1.3.1. Project general description
1.3.1.1. Fuel cell systems characterization
A large number of characterization tests were conducted on the FC systems to increase data significance and repeatability; the main activity was conducted in the frame of the project, further data set were provided by Fraunhofer (ISE) and Aalborg University . For low temperature FC a conventional and a modified Dantherm DBX2000 system were used, in the latter the BoP is electrically decoupled from the stack. The experimental campaign also investigated the effect of BoP components and grounding on the impedance measured on the stack. Furthermore, the modified system was tested in faulty conditions, i.e. some parameters in the on-board controller were changed to induce faults in the FC stack operation. The DBX2000 modified version was also tested with the external high voltage DC/DC converter manufactured by CIRTEM (Figure 1.3-b) as well as the low voltage one purchased from MICROPI and built according to the design made by UNISA and shown in Figure 1.3-a. Both converters were coupled with the EIS external board designed and built by BITRON.
Dedicated tests were designed and carried out at EIFER to record EIS spectra on the stack using only the converter. An on-line validation procedure was developed and the diagnostic tools were validated with the system running in normal as well as in faulty conditions, while the converters were used to record EIS measurements on the stack. The quality of the result is very good considering the complexity of the configuration and the difficulty of carrying out impedance spectroscopy measurement on a relatively high power system, such as the DBX2000. Figure 1.4 reports the spectra measured through the EIS board at different loads for the assembly of Figure 1.3-a (i.e. DBX2000 modified and LV converter). The EIS were measured at low, mean and high currents; in Figure 1.4 the points at low frequencies are on the right side of each spectrum. At 5A the EIS exhibits some noise at low frequency, as the amplitude of the voltage AC current is very low and then very sensitive to any perturbation. At 40A, the EIS measurement has high quality on the whole range of frequencies. This is because the higher the current, the higher the signal-to-noise ratio.
In order to better demonstrate the portability of the developed algorithms, characterization and on-line validation were performed at UFC on another commercial LT fuel cell system, the Nexa Ballard Power Module. Figure 1.5 shows the hardware assembly with the same configuration of Figure 1.3-b the HV converter is another prototype built by CIRTEM.
The high temperature PEMFC system assembled by Serenergy A/C was characterized at EIFER with a protocol similar to the one used for the low temperature Dantherm BDX2000 system. The tests were first performed with the control parameters default values and then repeated with different stack temperature, in turn, maximum and minimum values. The HT PEMFC system was also tested with the low voltage DC/DC converter coupled with the EIS external board and the stack impedance was measured (Figure 1.6).
To meet the requirements for the diagnosis of backup systems in switched-off condition, which is a critical task to guarantee system serviceability also for other FC applications, Dantherm in collaboration with the Aalborg University carried on the characterization in climate chamber of stacks in switched-off conditions. The new tests included not only humidity but also temperature cycles. To perform this activity UNISA designed an impedance analyser easily to embed into a DC/DC converter, which offers the possibility to perform EIS measurements system in switched-off conditions and detect the polymeric membrane humidification level. The Figure 1.7 shows the device developed by MICROPI connected with the FCS during the tests performed at EIFER.
The activity concerning the switched off analysis consisted in the measurements of the impedance on the DBX2000 modified system in switched off conditions at ambient temperature and humidity (see Figure 1.7). An example of the obtained results is given in Figure 1.8. The left diagram shows the impedance module versus the investigated frequency range. The right diagram shows the Nyquist plot of the same measurements. Both the results here presented are in accordance with the ones provided by Aalborg University (see Bidoggia et al.). Such results confirm the capability of the impedance analyser developed within the project and the feasibility of performing such a measure on the FC system for on field installations.
The table 1.1 shows a synopsis of the six FC arrangements (system and stack) investigated during the experimental campaign performed at EIFER and UFC; in the table the two supplementary sets made available from Frauhnofer ISE and Aalborg University are also reported for the sake of completeness. The cases shown in figures 1.3 through 1.7 are referenced into the relevant boxes; cross marks denote the other arrangements. It is worth mentioning that at the beginning of the project only two configurations were conceived for LT and HT FC, i.e. the LT DBX standard and HT Serenergy respectively.
1.3.1.2. Hardware design and manufacturing
The EIS board, developed by BITRON, is shown in Figure 1.9; this device is able to control both the HV and LV DC/DC converters, both shown in Figure 1.3. The EIS board output, where the variable frequency stimulus is produced by the firmware, is connected to either the voltage or current loops of the DC/DC converter, depending on the frequency range in which the stimulus lies. The two inputs to the EIS board are the stack current and voltage waveforms: they are processed, at each frequency of the generated stimulus, in order to produce an output file containing the real and imaginary parts of the stack impedance. During the signal processing, the firmware takes into account, among other things, the transient effect generated by the change of the frequency of the applied stimulus and the presence of noise, e.g. at switching frequency, affecting the two waveforms. The flexibility of the measurement is improved by the use of a configuration file, which is read by the EIS firmware as a first step of the measurement; it includes the list of frequencies (0.01 Hz - 2 kHz) to be analysed and the characteristics of the stimulus to be produced, frequency-by-frequency. The file is transferred to the EIS board by the supervisor (see element 3 of Figure 1.1) that runs on the FCS control unit. This allows to easily customizing the sequence of the measurement process.
The two DC/DC converters were developed in order to comply with different FC applications: stationary grid-connected, by means of the HV converter and backup, stand alone, by means of the LV converter. Both converters were designed to process the power produced by a 1.9 kW Dantherm system with a working voltage within the range 28-56V. As a consequence, the HV converter required a step-up topology, developed through a non-isolated solution in order to ensure the highest possible efficiency. The LV converter, which has to supply a 48V load, is based on a step-up-down SEPIC topology with coupled inductors. The project has also contributed to the development of a precision impedance analyser (see Figure 1.7) aimed at performing the EIS analysis of the stack in switched-off conditions. The analyser is able to cover the frequency range 1kHz-200kHz and can apply a sinusoidal stimulus from 1 mA to 1 A.
To perform the EIS measurements on a FC integrated in a complete system some guidelines were followed. Among others the main issues were those related to a proper interface and data exchange among Supervisor, Monitoring & Diagnosis modules and the FCS controller whose connections are shown in the Figure 1.1. Moreover, the impedance measure should not be influenced by any electrical component except the stack. A deep analysis of the measured spectra revealed several issues related to perform EIS measurements directly on a fully integrated fuel cell system. These issues were primarily caused by the fact that the fuel cell stack cannot be electrically isolated from the balance of plant and the system controller, since there is a circuitry hindering the EIS measurements of the isolated stack. Therefore, a modified fuel cell system had to be developed by Dantherm. The most difficult task in developing the modified system consisted in decoupling the DC/DC converter and the stack, as this modification required a major revision of the Dantherm DBX2000 embedded system software in order to make sure that the system was able to run reliably without compromising the basic safety of the system, in which the DC/DC plays a major role. This part of the work was inspired also by the need for having a system that complies with industrial standards. Other simpler solutions would have been possible for the purpose of testing only, but without achieving a complete validation at lab level. Now a solid study on that issue is available and will surely help during future actions oriented towards further implementation of the proposed methodology on field.
The schematic for the standard and the modified DBX2000 system is illustrated in Figure 1.10. From such representation it can be observed that the DC/DC converter was entirely electrically decoupled from the fuel cell stack (see the scheme on the right), and that the stack power was routed directly out through the main DC connectors of the control box. Also a separate stack current measurement had to be added.
In order to support the testing to be performed on the modified DBX2000 unit, it was chosen to develop a new monitoring and logging application for the system. The application communicates with the system via CAN-bus. The new monitoring application supported access to all system variables, and also supported logging these variables during testing on the system. On top of that, the application also provided a graphical interface, with the possibility of plotting charts for different system variables (see Figure 1.19).
For the purpose of integrating EIS enabled DC/DC converters, an architecture was developed, where the integration of the individual necessary devices was performed via a target PC, which served as the link between the fuel cell main controller, the DC/DC converter (with EIS board connected) and the diagnostic algorithms, with embedded EIS functionalities. The main advantage of this approach consisted in its flexibility in accepting codes or applications written in different programming languages. The solution would also virtually support any communication protocol chosen for communicating with the EIS-enabled DC/DC converter. Furthermore, the target PC had much more computational power compared to an embedded controller. The functional links among the components is presented on Figure 1.11. This scheme focuses more on signals communication if compared to that of Figure 1.1 which reports on the components functions.
On the DBX2000 system’s side the solution required a major modification to the software in the system’s embedded controller as well as a modification to the Labview monitoring and logging application for the system. Using this architecture it was possible to integrate the EIS enabled DC/DC converters as well as EIS measurement hardware developed in the project.
1.3.1.3. Diagnostic tools development and implementation
The model-based diagnostic tool developed by UNISA consists of two tasks: the identification of the parameters of an Equivalent Circuit Model of the stack, performed by exploiting the experimental spectra, and the detection and isolation of abnormal conditions via comparison of the identified parameters to their reference values. Such reference values are mapped as function of FC nominal operating conditions and modelled via black-box models specifically developed to estimate normal conditions values. The EIS spectra acquired in abnormal conditions were used for algorithm validation. An example of the online application of this diagnostic algorithm for DBX2000 system (see Figure 1.2) is shown in Fig. 1.12 where the values attained by one model’s parameter (e.g. the membrane resistance) are compared with the nominal ones. In this figure, the coloured areas highlight the parameters deviation (abnormal operations) at different operating currents. The isolation process is then performed making use of a signature matrix, which accounts for the information brought by four ECM model parameters (see the bottom side of the figure).
A detailed scheme of the on-line diagnostic procedure was conceived, with all the required phases, from the fault detection to the isolation. Moreover, among all the possible faults taken into account in a preliminary analysis, only drying and flooding were considered as the ones currently inducible for a validation of the proposed algorithm. Figure 1.13 illustrates the comprehensive model-based diagnostic tool proposed and developed by UNISA.
For the development of the aforementioned ECM parameters models, the knowledge of current load and stack temperature is essential. This requirement led to the development of a specific communication protocol between the UNISA algorithm and the DBX2000 system control software, so as to acquire the current and temperature measurements for each EIS spectrum. Analytical regressions have been computed for the system state variables (i.e. stack voltage, stack temperature and blower power) estimation in normal operating conditions. These regressions are essential to perform a plausibility check to understand if the system variables are at the expected operating condition.
To retrieve the essential information from the EIS board, the UNISA diagnostic algorithm reads from a text file the EIS measured variables (i.e. frequency, impedance real part, imaginary part, module and phase). These information are treated in order to identify the spectrum and compute the ECM parameters values.
Two knowledge-based diagnostic algorithms were developed by UFC: a fuzzy-logic one and an advanced neural-network based one. The former consists of a Fuzzy Clustering algorithm for automatic learning from experimental data and a Fuzzy Inference System (FIS) for decision-making. It was validated online on two different LT stacks (see Figures 1.3 and 1.5). The Fuzzy Clustering algorithm consists of four steps: feature extraction and selection, fuzzy clustering, and diagnostic rule design, showed in Figure 1.14 where the upper part is related to the off-line training and the lower part to the online testing. Since it combines both fuzzy clustering and fuzzy logic, it is thereby called “double-fuzzy” methodology.
Another diagnostic algorithm was developed, it is based on liquid state machine (LSM) which is an advanced architecture evolved from the traditional recurrent neural network. The main idea is to create an internal network randomly and keep the connections fixedly; the only part to be trained is the output weights .The main structure is shown in Figure 1.15. The stimulated stack voltage at a low frequency is used as the neural network’s input and the health status is used as its output. The main considerations of utilising stimulated stack voltage as inputs are: (1) the stack voltage is a credible reflection of FC degradation; (2) when stimulated by an AC sinusoidal current, it includes also the information about ac impedance, which is believed to be discriminant in diagnosing flooding and drying.
The experimental results demonstrate the algorithms high accuracy rate (100%), computational efficiency (less than 5s), easy implementation and good portability. Besides the LT type, the methodology is verified on the HT stack (Figure 1.6) which further shows its capability for fault diagnosis of PEMFC stacks regardless LT or HT types. Figure 1.16 shows an example of UFC algorithm for fault-diagnosis application to NEXA System. Three different operating conditions were analysed on the NEXA system (normal, air starvation and air surplus) and the stars represent the location of each operating point with respect to the three considered regions (i.e. clusters).
The diagnostic methodology proposed by EIFER is a data-based approach. First, a set of statistical reliable data in different conditions (i.e. flooding, drying and normal) needs to be taken. Second, the conditions of the fuel cell are defined as an output and associated with the referring states. Third, an Adaptive Neural Fuzzy Interference System (ANFIS) is trained with the prepared data set. The Adaptive Neural Network (ANN) teaches a Fuzzy Interference System (FIS) with a predefined structure and compares the output of the FIS with the known output of the data set. One run of the ANN algorithm is called an epoch. During this training phase, several epochs are required for proper results of the FIS. In general, a data set is randomly divided in sets for teaching the system and sets for testing the trained FIS. The error between the known output and the output of the FIS can be used as an assessment of the tool performance. Ultimately, the trained FIS represents a model of the states inside the cell. It is worth remarking that any data-based approach can only identify conditions with noticeable differences in the measured data sets. Another important point is to avoid and filter noise as much as possible. For that reason, different approaches for data pre-treatment were investigated. The algorithm scheme is presented in Figure 1.17. In the Figure 1.18 the results for the NEXA system are reported showing three cases, namely normal, drying and flooding operations for the NEXA System running at 10 A.
In Figures 1.19 1.20 and 1.21 the interfaces (see item 6 of Figure 1.1) of the three diagnostic tools developed by UNISA, UFC and EIFER are shown. It is worth remembering that the diagnostic tools developed by the D-CODE project may detect the following abnormal operations:
• Air starvation
• Fuel starvation
• Flooding
• Drying
Potential Impact:
1.4. Expected impact
1.4.1. Expected outcomes and target achieved
With the aim of analysing the future impact of the results achieved by the Project D-CODE, it is worth describing its outcomes with respect to those envisaged in both annual and multi-annual implementation plan documents (AIP and MAIP) .
With respect to the RTD priorities of the FCH-JU 2009 call, the project results will have a remarkable impact towards the key objective of “… placing Europe at the forefront of fuel cell and hydrogen technologies worldwide and enabling the market breakthrough of fuel cell and hydrogen technologies, thereby allowing commercial market forces to drive the substantial potential public benefits …”. Indeed, the proposed diagnostic tool represents a step forward to enhance the performance of FC systems, guarantees the advancement of the technology and improves the commercial competitiveness of the systems that might embed it. Although developed for PEM, the diagnostic tool might be extended to any FC technology through the methodological process outlined by D-CODE. Looking at the industrial scenario of the FC market, the proposed solution contains several technological innovations (e.g. on-board EIS, model- and knowledge-based diagnostic algorithms) that neither the EU industries nor their non-EU competitors have yet in the portfolio.
Moreover the AIP focused on applied research activities “ … directed towards developing components and sub-systems with improved performance, durability and cost for all three technologies in order to achieve system application readiness … ”. The project’s approaches, methodologies and applications belong to the broad area of applied research; this complies with the objective of advancing the actual technology through the enhancement of components (here component means either physical device or software) that are able to improve FCS. The objectives achieved have a direct impact on both performance improvement and durability; the former is linked to the capability of forecasting FC status thanks to the precise knowledge of the stack status via EIS monitoring, whereas the latter deals with the correct and timely detection of faults, whose knowledge may support the controller to activate precise countermeasures to maximize the system readiness.
The analysis of the results summarised above permits to state that the diagnostic tool designed, built and tested during the project fits within the boundaries set by the multi-annual implementation, whose one objectives was the achievement of “… new or improved materials as well as reliable control and diagnostics tools both at component and system levels … directed towards developing components and sub-systems … leading to step change improvements over existing technology in terms of performance, endurance, robustness, durability and cost for all technologies …”.
The results and the objectives just reviewed deals with technical issues whose impact spreads towards social and economic areas. Once implemented on-board of FC, the diagnostic tools will advance their performance contributing to increase the overall effectiveness of these energy conversion systems, as well as to make them more attractive. Indeed the solution proposed has a positive impact on the operational costs because of its effects on PEM stacks reliability and lifetime, which can be improved thanks to the monitoring and diagnosis tools. These benefits might be gained by all PEM applications (CHP, APU, backup) thanks to the flexibility of diagnostic tool, which can also be implemented on other FC technologies (e.g. SOFC) by following the methodological steps proposed. Moreover, the electrochemical impedance spectrum can be measured via low cost hardware, which was built following industrial standards. These points represent further advantages to reduce development costs and consequently purchasing price. Therefore, future commercialization can be easily promoted increasing the number of FC systems.
Finally all the above technical and economic advancements achievable through the implementation of the diagnostic tool will increase the diffusion of the fuel cell with an indirect impact on the major environmental issues.
1.4.2. Dissemination
The results of D-CODE have been disseminated through different ways to increase the awareness of both the scientific and industrial communities. The communication strategy was based on three pillars: workshops, website and scientific communications (i.e. publications).
According to the project schedule, two workshops were successfully organised. The former was held on June 2012 during the French Research Network on Fuel Cells meeting. More than 50 researchers attended the two special sessions on FCS diagnosis, whose 9 lectures were given by the invited speakers. The coordinator’s keynote speech was given during a plenary session with about 250 attendants (see Figure 1.4.1 left). A second workshop was held in Karlsruhe, Germany (April 2013). It was organized in the framework of the Conference FDFC (Fundamentals and Developments of Fuel Cell) and gathered 81 attendants (see Figure 1.4.1 right). Eleven presentations were delivered, followed by interesting comments and questions from the audience. The last event organized focused on the dissemination towards industrials. The D-CODE results were presented at the Hannover Fair in April 2014, during 1 week (see the pictures of Figure 1.4.2). The purpose was to show in a booth the hardware (i.e. the fuel cell system, the new converters, the control board, the diagnosis algorithm). The project’s booth was located in the group exhibit “Fuel Cell and Battery Exhibition”, the Europe's largest exhibition in this field. In the booth, the hardware developed in the project was exhibited and the results presented on posters (see Figure 1.4.3). Brochure and flyers were also distributed to the visitors. Sixty contacts were established with visitors coming from 20 different countries. A presentation of the results was done by the coordinator in the technical forum, the video is available on the Fair website as well as on “YouTube” (see the bottom right corner of the Figure 1.4.3). The participation to the Fair was a valuable opportunity for the consortium to meet several important actors in hydrogen and fuel cell activities.
The official website of the project is available at the URL https://www.d‐code‐jti.eu. On the same website all main achievements of the project are presented to disseminate the results and to increase the awareness of the industrial and research communities about the D-CODE project. The website has been regularly fed with the main achievements of the project: photos of the developed hardware, public milestones of the project, information on the events organized, references of the scientific publications co-authored by the partners and public documents available for downloading. As of July 12rd 2014, a search on the Google database using the keys sequence “on-line EIS diagnosis PEM fuel cell” provided a list of relevant links and the project D-CODE was in the first page at 5th rank.
To disseminate the scientific knowledge along the project, seven papers have been published through different ways. First, two articles have been published in a major journal which is International of Hydrogen Energy (Impact Factor: 3.55). Four papers have also been presented in major conferences publishing proceedings (3 in FDFC 2013, 1 in IEEE-IECON 2013). An article is in final revision in another high level journal, Journal of Power Sources (IF: 4.675) and should appear in autumn 2014. The list of the papers is reported below:
1. Zhixue Zheng, Raffaele Petrone, Marie-Cécile Péra , Daniel Hissel, Mohamed Becherif, Cesare Pianese, Nadia Yousfi-Steiner, Marco Sorrentino, “A review on non-model based diagnosis methodologies for PEM fuel cell stacks and systems”, International journal of hydrogen energy, vol.: 38, number: 21, pages: 8914-8926, DOI: 10.1016/j.ijhydene.2013.04.007.
2. Raffaele Petrone, Zhixue Zheng, Daniel Hissel, Marie-Cécile Péra, Cesare Pianese, Marco. Sorrentino, Mohamed Becherif, Nadia Yousfi-Steiner, “A review on model-based diagnosis methodologies for PEMFCs”, International journal of hydrogen energy, vol.: 38, number: 17, pages: 7077-7091, DOI: 10.1016/j.ijhydene.2013.03.106.
3. Angelo Esposito, Bastian Ludwig, Cesare Pianese, Soeren Boedker, "Experimental Characterization of a PEMFC System with Electrochemical Impedance Spectroscopy", 5th Fundamentals & Development of Fuel Cells Conference (FDFC), April 16-18, 2013, Karlsruhe, Germany.
4. Zhixue Zheng, Marie-Cécile Péra, Daniel Hissel, Mohamed Becherif, “Diagnosis of proton exchange membrane fuel cell （PEMFC） stack based on fuzzy clustering”, 5th Fundamentals & Development of Fuel Cells Conference (FDFC), April 16-18, 2013, Karlsruhe, Germany.
5. Raffaele Petrone, Zhixue Zheng, Daniel Hissel, Marie-Cécile Péra, Cesare Pianese, Marco. Sorrentino, Mohamed Becherif, “Implementation of EIS measurements on an embedded commercial system”, 5th Fundamentals & Development of Fuel Cells Conference (FDFC), April 16-18, 2013, Karlsruhe, Germany.
6. Zhixue Zheng, Raffaele Petrone, Marie-Cécile Péra, Daniel Hissel, Mohamed Becherif, Cesare Pianese, “Diagnosis of a Commercial PEM Fuel Cell Stack via Incomplete Spectra and Fuzzy Clustering”, 39th Annual Conference of the IEEE Industrial Electronics Society (IECON) November 10-13, 2013, Vienna, Austria.
7. Zhixue Zheng, Marie-Cécile Péra, Daniel Hissel, Mohamed Becherif, Kréhi-Serge Agbli, Yongdong LI, “A double-fuzzy diagnostic methodology dedicated to on-line fault diagnosis of PEMFC stack”, Journal of power sources, February 2014 – available on line, DOI: 10.1016/j.jpowsour.2014.07.157.
At educational level 3 PhD students were involved in the project, moreover several Bachelor and Master students worked on the D-CODE topics. Zhixue Zheng (UFC) has prepared her Doctoral thesis in the framework of D-CODE. She defends her PhD thesis in September 2014, at UFC, Belfort, France in front of an international jury (French, Spanish, Swiss and Chinese members in the jury). Raffaele Petrone (UNISA) was supervised by UNISA and UFC and defended its Doctoral thesis on his work developed in the framework of D-CODE. He’s currently performing a post-doc at UFC on a related subject. Pierpaolo Polverino (UNISA) focused its Doctoral work on SOFC, his experience on software and hardware integration allowed the development of the fault identification algorithms and its interfacing to PEMFC; he’s now with UNISA as research assistant. Some students at Bachelor and Master levels developed their final work in the frame of the D-CODE project. The teaching staffs of UFC and UNISA deliver several courses on energy conversion technologies for undergraduate, graduate students, summer schools and tutorials. The D-CODE results will feed the fuel cells sections of these courses.
The dissemination of the results will continue after the end of the project as a part of the main communication activity of each partner, either research institution or industry. Indeed further scientific publications will be issued as follow-up of the work done during the project. It is envisaged that the results achieved will be further checked and improved to make them interesting for publication in International Journals. Moreover, the work performed has set solid bases on which further research work can be built leading to more publications as well as other research projects, which in all cases will acknowledge the role and the results of D-CODE. Academic partners will include in their teaching activity the study performed within D-CODE and the experience made. This will spread out at a larger extent the concepts of diagnosis as well as of the on-board application of EIS. At industrial level the work performed is part of the technical know-how of these partners, which may also add the developed hardware in their catalogues. Therefore, several dissemination actions will be also taken as part of the normal communication strategy of each partner. Moreover, other dissemination actions could be performed if requested by the FCH-JU (e.g. annual review events).
1.4.3. Exploitation of results
The project D-CODE has demonstrated the feasibility of using a low-cost EIS-board in conjunction with a DC/DC converter to perform EIS-based PEMFC diagnosis. This enables fuel cell systems to be monitored during operation or in standby mode, contributing to an improvement in their reliability without adding excessive cost. The DC/DC converters and the EIS board are production-ready devices. The future step is to further develop the application to take it out of the laboratory and into wide scale field testing, with a view to making the solution commercially available for a wide variety of fuel cell applications.
The consortium looks forward to exploit both hardware (converters, EIS board) and software (diagnosis algorithms) outcomes. Limited but necessary improvements are required to implement the tool for on-field use. The TRL achieved by D-CODE ranks the level 4 (technology validated in lab); TRL5 (technology validated in relevant environment) can be easily achieved, even TRL6 (technology demonstrated in relevant environment) could be reached as well. Therefore a follow-up would focus on the improvement of the proof of concept towards on-field application to valorise the project results. Moreover both hardware and diagnostic algorithms could also be easily exploited at laboratory level with an effective reduction of equipment costs; this would require some changes in the implementation and in the hardware to comply with laboratory standards. Future work may support the research towards the study of other faults or failures to be isolated on-board by processing the impedance spectrum via either model- or knowledge-based approaches. Further actions are also required on the methodological side to develop an integrated experimental and computational framework for the design and manufacturing of EIS-based diagnosis tools. This would simplify the implementation on any PEMFC and may reduce the effort required for diagnosis algorithms identification and customization. An ad hoc activity might be settled for stack and system identification under mimicked faults for the development and testing of diagnosis algorithms.
The project’s outcomes will serve as a basis to provide increased knowledge on the potentialities offered by advanced diagnosis, on-field deployable diagnostic tool and related hardware. It is worth remarking the close interaction between diagnosis and state-of-health evaluations as well. The availability of advanced algorithms (i.e. EIS-based) for the condition monitoring of FC represents a first step towards the implementation of advanced control, as well as effective prognosis algorithms. Future cooperation among the partners will support significant research on PEM lifetime via advanced control (either feedback or adaptive) and the development of tools aimed at evaluating system residual lifetime as well as expectable end-of-life performance. Apart from the application to other FC technologies (i.e. SOFC), the developed methodologies and hardware can be also extended to other electrochemical devices (e.g. battery).
The D-CODE partners will exploit the know-how gained during the project along three main lines of action, namely at industrial, research and educational levels, either within cooperative or autonomous work. Cooperative activity will be exploited by some partners potentially interested in the implementation of the EIS on commercial FC. Notably DANTHERM Power may lead a group to continue the development of the actual tool for the implementation on board of their systems. In that case the work will focus on processes, methodologies and standards at industrial level to make the diagnostic tool an effective instrument to improve performance and increase the FC lifetime. This will guarantee a better competitiveness and may represent a reference for the entire community, thus fostering a widespread implementation of advanced diagnostic tools.
On the power electronics side BITRON could continue the development of the EIS board and include it in their catalogue. In such a case, the board may be purchased by customers wishing to use a low cost instrument capable of injecting sinusoidal waveforms for the purpose of analysing the dynamic response of other electrochemical devices (e.g. batteries). Electrical devices can also be considered for the dynamic response at frequencies below 1 kHz.
CIRTEM could exploit the experience achieved in the design and building of high voltage DC/DC converters for several applications other than fuel cells.
Moreover, the entire diagnostic tool (EIS board, converter and algorithms) might be installed at lab scale as low cost spectrometer. Moreover, the algorithms proposed by UNISA, UFC and EIFER are ready for on board implementation of any spectrometer and could be implemented in software packages for the EIS data analysis and FC diagnosis.
All research partners will continue the researches started in the project; the exploitation of the results will continue towards both basic and applied research leading to further cooperation opportunities. As a first step, a common paper between the D-CODE partners will be coordinated by UFC in order to synthesise the results obtained with the algorithms developed by UFC, UNISA and EIFER.
List of Websites:
1.5. The project public website
The D-CODE website was structured and implemented in the first part of the project and mostly used for the data exchange among the partners through the intranet section. During the project, as soon as new outcomes were made available and the main dissemination events were closed, the website was updated with new information. Below is reported a summary of the website structure then the description of the actions made during the second period is given.
The website was implemented by EIFER and is hosted on their server. The web address is https://www.d-code-jti.eu. It’s divided into two main sections, namely, access-free and intranet area, which is restricted to the project partners only. Overall, the web manager registered about 3000 contacts. The Figure 1.6.1 is a snapshot of the welcome page.
Public webpage
The public web page has 6 menu tabs (Home, About, Partners, Contacts, Events and Products & Documents). The sections Events and Products & Documents were updated continuously during the second period to add the information about the main dissemination events and the documents describing the results of the projects. The website has been regularly fed with the main achievements of the project: photos of the developed hardware, public milestones of the project, information on the events organized in the framework of the project, references of the scientific publications co-authored by the partners and public documents available for downloading.
In the Figure 1.6.2 the snapshot of the upper part of the section event is shown, here the main information on the participation to the Hannover Fair (April 2014) are reported. Several URLs connect to other website with further information on the hosting event. A list of the main documents describing the project and its results is available in the section documents as shown in the Figure 1.6.3.
Upon decision of the Steering Committee held on March 28, 2014 the website will be active for two years after the end of the project. This is one of the main follow up actions that may help valorising the project outcomes.
