CORDIS - EU research results

Green Retrofitting of Existing Ships

Final Report Summary - REFRESH (Green Retrofitting of Existing Ships)

Executive Summary:
The underlying idea of the REFRESH was to identify retrofitting solutions considering mutual interaction of integrated shipboard systems and functions or services.
The work carried out within the first half of the project resulted in developing a versatile computational platform. The software brings together the rapid modelling techniques with a powerful sensitivity framework. This allows a thermo-hydraulic network (fully integrated with an electrical grid, a control layer and top-level functions, such as propulsion) to virtually reproduce any level of complexity to be assessed for energy performance as well as for system integrity. The framework identifies important interactions between components and physical parameters (e.g. heat-exchange or flow characteristics) playing dominant role in the systems’ performance, hence providing a benchmark for future tasks, such as optimisation.
In REFRESH, the optimisation of energy performance was carried out for a number of systems (such as cooling system for a VLCC or HVAC system of a passenger ferry), but the most complete studies were undertaken for a cargo handling system of an oil tanker. These studies involved the optimisation of the system’s topology as well as its sizing and operational efficiency.
One of the major difficulties in assessing energy performance of ship systems is the amount of data required for running simulations and validating results. Accuracy of numerical predictions is highly dependent on both the quality of the data and the resolution of the models. Furthermore, high-resolution models often produce large amount of data that is used as final results or just for quality control. Operational data is also necessary for validation and verification of the predictions, and all this results in a need for efficient technologies for data storage and information management. This complex task was addressed within the project by a dedicated team aiming to develop a database solution, universal enough to meet the needs of various partners and different applications. An additional requirement imposed to the development of the database was to make it compatible with advanced data-mining techniques and probabilistic modelling (Bayesian Network). For this, in REFRESH the suitability of data-intensive models in retrofitting problems was investigated in parallel to parametric modelling.
All the tools and methodologies developed or refined in the project were brought together, unveiling the required strategy for handling the whole energy modelling application process, which allows the installation on-board of a DST for efficient operation based on DEM.
The tool installed on-board a RoPax ship aims to use parametric models of the ship propulsion and auxiliary plant as well as live feeds from the navigational system and ship’s automation data to identify operational deficiencies in. Once the efficiency issues are properly evaluated the tool produces a list of actions and recommendations for improving energy efficiency in operation. Furthermore, the system stores the data for use in finding retrofitting solutions for the efficiency failings originating in non-optimal performance of the systems’ components.
The above description would not be complete without mentioning major difficulties encountered during the project’s execution. Most of these stemmed from the project’s inherent over-reliance on operational data, which proved to be far too scarce and too unreliable to provide the basis for accurate assessment and generalisation of the results. Other information-related challenges were encountered while calibrating numerical models as well as during the installation of the DST (backward compatibility of data exchange interfaces within on-board automation systems). The latter issue may require paying particular attention for the regulatory framework on monitoring, reporting and verification. On the other hand, the troublesome reliance highlights the importance of numerical prediction (combined with targeted audits) and use of sensitivity analysis for (educated) decision making in retrofitting, both applied during the course of the project.

Project Context and Objectives:
REFRESH addressed the existing merchant fleet, with the aim to treat energy efficiency as an integral part of the life-cycle a ship. Such ambitious task was taken through modelling and simulation of energy flows on-board for various operational conditions and for extended periods of time. REFRESH shared the same methodological approach with the project TARGETS ( which covers in detail the hydrodynamic performance of a ship (both hull and appendages) as well as the systems performance on-board; the above being the reason for the joint meeting held in March 2013 between REFRESH and TARGETS partners, to share ideas and early results. Considering that a limited number of choices for hydrodynamic improvement of a ship can take place after its delivery, REFRESH focused exclusively on the energy assessment of the onboard systems.
Stemming from the fact that any retrofitting investment taken under consideration (i.e. listed among WP1’s Energy Saving Potentials) will be scrutinised by technical feasibility and cost-benefit analysis, it is necessary to provide the operator with sufficient insight into the energy performance of a ship in order to make an informed decision. The above is implemented by
• Identification of underperforming systems and components, and
• Virtual replacement of them with alternatives and assessment of their effects on key performance indicators (KPI).

Moreover, the intended benefit following the upgrading of the ship systems needs to be verified in practice, for example with the savings in fuel consumption. This calls upon the development of a monitoring system that will collect data from various components on-board, compare it to the optimal or predicted consumption (for a given set of environmental and loading conditions) and advice the crew accordingly.

In particular, the project focused on the following four development areas:
• Energy audits and retrofitting specification: The process starts by assessing the current status of the ship by on-board energy audits. The output of this step is the recommendation of technical solutions to improve the fuel consumption of the ship followed by shipyard specification for implementing these changes in a retrofitting project.
• Dynamic Energy Modelling (DEM): Although the energy audits introduce a very practical approach to the project, it is necessary to couple them with mathematical models of energy flows on-board. The reason is that energy audits provide only a facet of the performance of the ship lasting only a few days (i.e. as long as it takes to complete the audit), whereas a theoretical model can simulate several scenarios and combinations of operating conditions.
• Optimisation for energy efficiency: On the basis of theoretical modelling and past experience, the optimal KPI values for any combination of the environmental conditions and ship state (e.g. full load condition and nominal speed) can be identified.
• Life-cycle energy management: The final step in the process is to establish an energy management system, which will be comprised by a database fed with real data collected on-board, application of a data mining methodology, and, most important, a decision-support tool to provide energy efficiency oriented advisory to the crew on-board by combining actual performance values with simulations of alternative scenarios.

Project Results:
Work-package 1 - Retrofitting specification and budgeting
First step identified while approaching existing ships energy efficiency and application of retrofitting solutions have been the investigation of the problematic areas as well as the identification of the available retrofit solutions to take into consideration.
Main objective of the first work-package was to deploy the energy audit methodology in view to identify the energy deficiencies on-board and, thus, to pinpoint various configurations and modifications of existing and new systems that would improve the overall performance of the vessel.

REFRESH project Consortium, in particular through the participating operators (NAVIOS, ANEK & CSME – the latter representing Maran Tankers Management Inc.), offered vessels representing the major existing ship types.
During the project’s Kick-Off Meeting, a selection of 6 vessels was nominated for energy auditing, including

• Two (2) RoPax ships
• Two (2) Bulk Carriers (Handymax or Panamax, Capesize)
• Two (2) Tankers (Aframax, VLCC)

Such decision was based on a number of technical reasons addressing ships’ age, main equipment, operators and consultants experience based feedbacks.
Energy audits were conducted by AMS (i.e. bulk carriers and tankers) and TV (inspecting the Ropax vessels).

All major systems’ health and current performance was analysed and reported.
Drawback identified and discussed was the need to base the investigation upon available information from the on-board systems, with partial help from the tools that can be carried on-board by auditors; on the other hand, the level of information collected allowed the experts to prepare a list of Energy Saving Potentials (ESPs) deemed suitable for each vessel, taken from a wider ESPs list discussed and agreed with the Consortium.

Such ESPs were analysed for both technical and financial feasibility.
Benefit was examined for each ESP and for each vessel under consideration; results were, then, expressed through allocation of expected yearly benefits from each single solution within specific fuel bill reduction ranges.

Starting from this early stage of the project, both operational and technical (i.e. requiring actual refit works) solutions were taken into consideration; this having in mind efficiency improvements through a combination of actions, where the impact of each solution applied takes advantage from the presence of the others (e.g. at least by changing its boundary conditions).

Having collected necessary input on-board and run benefit estimation from the available solutions for improving the vessels’ efficiency, the last task of the work-package focused on cost estimations on technologically feasible ESP configurations for the different ship types.
Shipyards have provided economic data related to retrofitting systems on board, replacement of energy deficient systems, etc.
Based on this exercise, resulting in cost estimations summarized and compared through the return of investment, most cost effective configurations were defined and reported.

Work-package 2: Dynamic Energy Modelling

While clearly the first work-package provided information on where the room for improvements are and also on what actions would be recommended, based on available technology, it also highlighted the need for a more analytic approach to estimate the impact such solutions on the operation of the main systems and their interactions.

The second work-package, leaded by SSRC, then, aimed at applying the energy models concept to ship’s systems.

In order to facilitate the assessment of the energy performance of the ships under consideration, it is necessary to develop a series of energy models for the onboard systems. The underlying principle is that the ship is treated as an autonomous energy system, which is comprised by a series of energy systems (propulsion system, fuel oil system, etc.). Every system is decomposed further into energy components (pumps, heat exchangers, etc.).

The building blocks of the Dynamic Energy Modelling (DEM), main subject of the WP2 and central concept of the whole project, were the mathematical models of generic components, integrated into a platform and then assembled in different combinations and operational parameters for modelling the shipboard systems.
The modelled systems are divided into three major groups:
- Hydraulic systems are associated to transferring of liquid masses such as coolant, fuel and lube oils, ballast water or liquid cargo.
- Thermo-hydraulic systems are associated to thermal energy conversion (e.g. steam for FO heating) and transfer (e.g. maintaining temperature of an engine block)
- The electrical systems are responsible for the electrical power generation and distribution of electrical energy to individual consumers.
The developed energy routines are integrated with use of in-house software that allows the necessary functionality (i.e. systems configurations) and computational efficiency for any given ship itinerary and time interval.

Due to the bottom-up modelling followed in DEM, the individual energy performance of every considered component and system on-board could be obtained for a given set of environmental conditions (part of the operational profile of the ship) and time interval. Detailed knowledge at this level allows the following possible applications of DEM, investigated throughout REFRESH project:
- Design: DEM can be used for the assessment of the energy efficiency of a large number of design alternatives in terms of system / component selection, power plant configuration, and operational profile.
- Operation: Considering that the energy systems configuration of an existing ship is fixed for most of its operational life, application of DEM has the potential to offer the framework for energy management, thus providing a decision-support tool in real time that will indicate the best energy savings strategy when operational conditions vary.
- Retrofitting: DEM can be used for the identification of systems with poor performance in terms of energy efficiency and rational selection (cost-effectiveness and return-on-investment) of the upgrade strategy. Due to its integrative nature, DEM can provide fertile ground for assessing the potentials of new technology and innovation (renewable energy, waste heat recovery, etc.).

The development of energy models, leaded by University partners and heading towards the simulation of energy flows in a sufficient resolution to analyse the energy performance and evaluate solutions, was strictly linked with the input coming from audits (i.e. systems performance data) and operators, through systems arrangements and operational data (e.g. ships’ noon reports).
Such fundamental exchange of information from actual vessel operation to academia, started with the first task of the work-package (i.e. ‘data collection’) and continued for a whole year while the development of models continued, since initial expectations from developers proven to be far from developers’ expectation; this was not just about not-optimal number of parameters tracked by operators, which is predictable, but also resolution and database consistency.
Lesson learnt from such process was that, without frequent direct use from ship managers, operators, ship personnel, of the recorded data, any human errors and calibration issues remains unnoticed until it’s too late to verify and amend; this highlighting the need of a cultural change in the way the ships are operated, still relying too much on ship’s personnel experience and not sufficiently looking at daily systems operational data, which could provide critical information about machinery’s health and performance.

Based on available ships’ data and identified ESPs, the developers succeeded to finalize and share executable files based on energy models implementation and set up in order to simulate the energy flows of three different systems (Simulation of energy performance, task 2.3) i.e.:
1. Heating ventilation and air conditioning system
2. Cargo unloading system, and
3. Cooling water system

A number of simulations were conducted in order assess the performance of these systems. For each simulation i.e. case scenario, a specific variation of the weather (external forcing functions) and/or the operational profile (set points and/or number of components) was considered.

On the developed models, in view of their, optimization, their application for generating Bayesian Networks and their use for producing a prototype innovative Decision Support Tool, sensitivity analysis and uncertainty quantification was performed.
Aim of such activity was to identify those components or model parameters on the ship power plant propitious of inducing large output or uncertainties. In both cases, (i.e. sensitivity analysis and uncertainty quantification), the assessment is perform first by calculating/estimating the system sensitivity coefficients, and then subjecting them to either to operational variations from the nominal input values (sensitivity analysis) or to input uncertainties (uncertainty quantification).

It was then noted that, when addressing the ship power plant in terms of sensitivity, the inherent thermo hydraulic network acts a source of vast number of parameters/variables difficult to be screened out.
Due to the closed-loop nature of this sub-system, it is impossible to estimate the sensitivity coefficient without solving the system (direct method).
Following discussions on the matter between the involved partners, an “adjoint system” is constructed, which allows to orderly an effective calculate all the first order local sensitivity coefficients for the whole thermo-hydraulic framework.
In the case of the Auxiliary Power plant sensitivity analysis, the adjoint sensitivity analysis procedure (ASAP) becomes a practical tool to compute the whole network first order local sensitivity coefficient.
Thus the ship auxiliary power plant can be reduced to a handful number of system parameters.
For this particular case it is shown on the relevant project deliverable that, the thermo-hydraulic framework is reduced from 95 links and 21 tanks, to 20 links and 5 tanks, given a margin of 1% parameter non-dimensional sensitivity. Of course once the number of parameters in the mentioned network has been filtered, their sensitivities can be incorporated to the auxiliary power plant electric distribution open loop chain.

The last task of the ‘Dynamic Energy Modelling’ work-package focused creating parametric energy models, to be used as a computational engine for the Decision Support Tool prototype.

Main objective of the task was to investigate feasibility of developing regression models for complex systems (such as ship auxiliary plant) in order to reduce computational effort in predicting energy performance without compromising accuracy of the prediction.
The calculations were performed with basic settings for the data generated within T2.3 (Simulation of energy performance) and reported in the relevant deliverable (D2.4) with the exception the third case study (i.e. cooling system) the input data was not pre-processed.
The results obtained and presented in the last WP2’s report (D2.6) demonstrated successful application of response surface design to all three different energy models, although non-optimal data quality available from ships was highlighted, and can be resumed as follows:
- overall satisfactory accuracy;
- possible incompleteness/too low order of the model and/or use of averaged data might be behind patterned residuals in cargo handling example;
- presence of outliers (or some other data contamination) is believed to have negative impact on HVAC example;
- in case of the cooling system regression model can produce very accurate results without any significant deviations from the simulated data.

Cargo handling and HVAC systems raises the attention on the importance of formal identification of dominant parameters.
It also needs to be noted that dynamics was “implicit” in the investigated cases and that the models were applied either to
- averaged/aggregated results of time domain simulations, or to
- Steady state results; i.e. the results were obtained with use of dynamic energy models but time histories e.g. FOC = f(t,…,…,) were not analysed.

Overall, the results presented in the foregoing were promising and demonstrated that response surface methodology can be employed for rapid and accurate utilisation of results of time domain simulations of energy systems.
Obviously, the response surfaces designed in the foregoing are valid for specific systems developed within WP2 with specific operational ranges.
Nevertheless, the results presented above demonstrate that the response surfaces to be used for DST are relatively easy to construct and are accurate enough, hence rule out need of use DEM calculations in on-board applications.

Work-package 3: Optimisation and Benchmarking

The aim of the third work-package was to make use of the available input from the previous activities and elaborate on the energy efficiency optimisation of a ship in both the retrofitting and the operational phase.

WPL NTUA leaded also the first task, towards the creation of a virtual model for the systems arrangement, this including the process of selection of an appropriate geometry modelling software (in this case: Rhinoceros®) among numerous available software packages on the basis of set criteria and examples of parametric modelling and visualization of energy models.

Among 11 candidate software and based on the following criteria Rhinoceros® was chosen and applied:
- User learning time
- Level of detail
- User-friendly interface
- Available tools for parametric design
- Connectivity with other software
- Hardware performance requirements

The main objective of the developed software is to create a simplified model of any possible arrangement with focus on the space requirements rather than the detail of each component. The aim is to create a model that can be coupled with any energy model. This is accomplished by developing a script in Rhinoceros which has the same input with the energy model but only extracts the necessary spatial information from the input file and ignores the rest.
The developed software is capable of communicating with various different file formats that are commonly used in energy models such as CSV format, SIMULINK© (.mdl) and also XML format.

The virtual model (*.3dm file) and the folders with the inputs for the presented examples (SIMULINK simple hydraulic example, Aframax pump room, iSySE demo model and the simplified engine room arrangement) were finalized and shared in the relevant deliverable document (D3.2) with user instruction in order to properly run the scripts.

The second task of the work-package dealt with the ship optimization for retrofitting.

Different design retrofit solutions were generated based on WP1 results, each of them having a spatial and energy footprint.

Each solution present different performance characteristics assessed against the decision makers’ criteria, thus creating a Multi Criteria Decision Making (MCDM) problem.

The design space was, then, investigated along with the optimum solutions, using standard Multi-Objective optimization software.

To resume, the ESP solutions suitable for retrofitting were optimised with respect to the following objective functions:
1) Exhaust gas minimization/Minimization of fuel consumption
2) Minimization of the retrofitting cost
3) Maximization of the Net Present Value (NPV)
The only constraint of the optimisation study was that the NPV is positive.
Such optimization problem required to be expressed in mathematical form, hence ‘Normed Weighted Sum Method’ is selected and is described in the relevant deliverable document (D3.3) discussed with partners and approved.

The identification of optimal solution is then made by evaluating all possible combinations of the proposed ESPs, for all ship types addressed in the REFRESH project and by taking into consideration the preference of the Decision Maker (DM).

The third task of WP3, on the basis of the developed energy models in WP2 and thereafter refinements, as well as the previous work-package 3 work on different design/retrofit solutions, focused on the optimization of the operational life of the ship and the required re-adjustments for the so many different operational conditions that may be encountered.
As agreed among partners, the objectives of the optimization exercise for each developed model refer to the minimization of fuel consumption, while considering exhaust gas emissions for regulation compliance.
Such operational optimization process was implemented in Matlab’s environment and it was presented for the following systems
1. Cargo handling system of an existing Aframax tanker;
2. HVAC system of a ROPAX vessel;
3. Ballasting procedure optimization (with use of variable speed drives);

The tanker case presented the following results:
Although in general slower pump speed requires less energy from the pumps which results in less steam consumption and therefore less fuel is burned from the boiler, it’s flagged up that shore orders regarding the required pressure or flow rate at the manifolds during the discharging procedure force the operator to maintain the RPM at certain speed. This means that the only parameter that can be selected by the operator is the number of pumps that will be used in order to maintain the shore requirements at the manifolds. Furthermore, if it’s possible to agree at the beginning of the procedure a lower pressure than the maximum allowable (without overlooking other constraints, such as the time spent in port) it is possible to have significant savings, especially in terminals with long distances. Indicatively, a reduction in the manifold pressure by 10% can provide 8% less fuel consumption. Quantification of steam/fuel consumption reduction due to lower RPM is clearly quantified.

For the ROPAX HVAC optimization case, the following remarks were presented:
- The return air ratio is constant, equal to 70% as expected, since it is easier to maintain the same temperature when the air is recirculated.
- It is worth noting that the heating, during the winter and autumn months is much more energy intensive than the cooling during summer. (It might be counterintuitive considering that the vessel is in the Mediterranean Sea; of course results discussed with operator, ANEK)
- As expected, there is a strong influence of the set-points of the heating and cooling system into the amount of energy the system uses (to be as aligned as feasible with external temperature, within commercial related constraints); quantification of this helps the operators to give proper priority to their decisions.

Aframax tanker, ballasting procedure case presented the following result:
From the analysis of operational data it is possible to note that the water ballast pump is not operating in rated conditions, characterized by a rotational speed of 1180 rpm. Feeding the induction motor rated voltage and frequency (440 V / 60 Hz), operating points are correctly characterized by a rotational speed very close to the induction motor rated speed (1191 rpm). In order to utilize the water ballast pump at its rated speed it is necessary to use a speed variable drive.
Simulation results quantified the savings, in way that, of course, could be repeated easily on other vessels starting from relevant energy models.

From the obtained optimization results, it can be observed that the accurate simulation/prediction of the behaviour of any sub-system of a vessel is of outmost importance and a proper operation/selection of the crucial parameters for each modelled system can provide significant savings in terms of energy/fuel consumption. On the basis of the conducted optimizations, valuable guidelines for the efficiency of ship operation were developed, which may be disposed to and exploited by ship operators.

The last task of the third work-package, ‘benchmarking studies’, dealt with further review of WP3 results from experience, involving Class, shipyards, operators and consultants.

Such review was particularly critical towards all ship specific commercial and technical reasons causing the boundary conditions assumed in the optimisation to be different and thus modifying the outcome of proposed energy efficient solutions. Thus it is imperative that no solution should be applied (in particular, nothing should be installed) any vessel on the basis that “it works for a similar vessel”. Even the operational profile of a particular ship can have an adverse effect on the expected benefit of an installed ESP. Hence another lessons learnt for developers is that a multi-objective task means that the systems will have to be exhaustive, since small variations of the model can have a massive impact on the outcome (e.g. a proposed set of operational and refit solutions should be verified not only against current actual loading conditions, trading route etc, but also against all other reasonable set of boundary conditions that may occur, thus raising the reliability of R.O.I.s and allowing easier application of the same or similar studies to other vessels).
More specifically, solutions examined included maintenance procedures such as overhaul of Main Engine and Diesel Generators, while in terms of pure retrofitting the ESPs reviewed included installation of high efficiency electric motors, homogenisers, variable speed drive pumps and replacement of incandescent lams with CFLs. All examined solutions demonstrate revenue but depending on the application it might not be economically viable, not even within a 5-year perspective.
Only the most basic solutions, including for example the replacement of incandescent lamps by higher efficiency ones, resulted in an unmistakable benefit for every application, representing a series of subtle solutions with minor cost that can be added up to improve a vessel’s overall energy efficiency.
The case of the replacement of the conventional cargo handling system of an AFRAMAX tanker with deep well pumps, which demonstrates exceptional results in terms of savings, accompanied on the other hand by a very high capital investment required, was penalised by experience based feedback highlighting that there is also high risk involved, due to the difficulty in repair and maintenance of such systems.
Cargo handling optimisation for a tanker vessel calculating both figures of discharging time and consumption, discussed for possible application on a larger number of vessels though REFRESH project operators, demonstrated that the parameters affecting the Cargo Handling operations are not easily manageable. (e.g. pump speed which is one of the major parameters affecting fuel consumption upon discharging operations is directly associated with discharging time. Hence it cannot always be altered, as it depends on parameters imposed by port facilities).
HVAC optimisation for a RoPax vessels resulting from on-board awareness on the cost of non-optimal temperature set points and their dependency on the operational profile of the subject vessel, still look promising after partners’ review, although absolute figures of savings were not presented, as it would have required a specific long term data collection on historical set points and temperatures within cabins and public areas.
Work-package 4 - Monitoring & management in operation

Developed in parallel with the optimisation activities (WP3), and leaded by SSRC, the work-package focused on the data-intensive processes required so to prepare the ground for the DST oriented activities of work-package 5.

The first task, ‘data collection’, allowed to identify and describe the data sources to provide the necessary information for the task at hand.

It aimed, on one hand to identify the origin of data both from prior developments in REFRESH and from on-board measurements (operational data), and, on the other hand, to link the theoretical and operational data under a framework, notwithstanding the complexity (and interdependency) of on-board systems.

Taking into account the complexity of the task at hand, the Consortium of REFRESH decided to combine practical and theoretical data and hence elaborate on data that can be grouped in three categories as follows:

- Items (machinery components and their configurations) that would deem a retrofitting project successful in terms of overall improvement of the energy efficiency of a ship; input coming from the first work-package;
- Identification of dominant parameters from the simulation and sensitivity analysis and uncertainty quantification for a given operational profile of a ship, which is WP2 input;
- Contribution of practical experience and data concerning fuel consumption of ships, new contribution requested to partners at this stage.

Architecture and requirements of the database were then set.

Main requirements set for the DB to serve the purpose of the energy management of the ship, were set as follows:
- identify the data that needs to be stored in the database;
- define who are the main users of the database and how they need to access it;

Taking into account that the main planned purpose of the DB would be to serve the WP5’s Decision Support System (DSS), then the Data mining tool (Bayesian Networks) of Task 4.3 and potentially the owner and other users.
Planned uses of the DB were mainly the Decision Support System (DSS), the Data mining tool (Bayesian Networks) and potentially the owner and other users.

Besides, the database was to be structured generic enough to accommodate different types of ships, but with a particular attention to ROPAX ships as they possess more complex systems configuration and are more susceptible to optimisation than the alternatives available in this project, i.e. cargo vessels. Similar consideration leaded to the decision to prepare the prototype DST for one of the ANEK operated ROPAX ships.

Looking in particular at the DB required use, it was decided that a relational database model would have been the most suitable solution for the requirements of the REFRESH project.

Of course “Comprehensive Energy Efficiency database”, deals with the development of a database that has to hold information related to the energy management of a specific vessel. Indeed, essential information (data) for the exercise is to store the input and output operational variables subjected to a set of fixed characteristics/parameters. So, in a ship energy context this is translated into speed, distance and cargo transported as input variables, and fuel consumption as output quantity. In addition, the collection of additional ship/voyage characteristics, (i.e. sea state, water consumption etc.) will improve the calculations and validation of simulation results.

Two different types of data were, then, considered to populate the database: Operator data and simulated DEM data. Operator data includes among others, environmental conditions, RPM, speed, fuel consumption, etc., while simulated data constituted the output of DEM as per WP2 developments.

For what regards the structure of the DB, it has to be noted that organisation of the information is based, up to a certain extent, on the STEP-ISO 10303 and the SFI code, which is a system for classification of technical and economic ship information developed by the Ship Research Institute of Norway in 1972.

The database has been developed using Ms Sql Server and a user guide has been developed by the task leader. To simplify the use of the database, a user interface has been developed to demonstrate the data manipulation. The database software can be used to create new ship data, edit existing ships, edit and create machinery components, electrical components and other categories. The user interface also gives the option to import and export data form and to the database storage.

The last two tasks of the work-package focused on the consolidation and the verification of the database information and the knowledge models in place, in view to properly extract and use the information for decision-support purposes.

Heading towards a mathematical model to look for meaningful patterns in data, from the different data mining techniques that exist, Bayesian Networks were chosen to analyse the data related to energy usage on-board vessels.
Two different types of data were analysed: simulation data and operational data from noon reports.

Bayesian networks were learnt from the available data and were presented within projects deliverales D4.4 and D4.5.

Bayesian Network structures learnt from data need to be assessed in terms of their ability to represent the involved processes and in terms of their structure, parameters connectivity, visualisation and usefulness.
Two steps of assessments (basis and detailed) were performed and the elaboration of the received feedback and comments facilitated the improvement of BN setting up procedure.

Most notable findings reported were the following:
- BN identify correctly the correlated variables and clusters corresponding to either physical parameters (i.e. multiple nodes linked to a single parameter, e.g. ME load) or to functionality shared by number of components (e.g. aux engine working hours can be grouped into single function)
- In cases where detailed and extensive information is available the BN networks are effective and powerful tool. In case of incomplete, sparse or infrequent data however, (typical case in marine industry and specifically in the case of older vessels for which noon reports are often the only permanent source of data), the missing information need to be complemented by experts judgement to form the constraints for the parameters connectivity.
- Similarly BN are a useful tool to extract information, which needs to be trained and used with special care. The simpler models can be more easily understood but they are less useful, whereas the complex model (i.e. providing substantial piece of information which can be used to assist the ship operators and marine experts to get more insight of the ships systems operation) require special attention during their set up and training process.
- BN is constructed based on a data derived from a well-structured physical system, the reduction process will bring, with every iteration, a structure of the resulting BN closer to parametric model. As a consequence, if the physical network is largely known (i.e. the dominant parameters can be readily identified) and the inferences are not of prime objective the parametric models are faster and easier alternative to BN in setting up the empirical models. The same may apply in case of small datasets.

Following the above, and in parallel with the work on DST specification and the discussions and following decision about the most suitable vessel for DST prototype installation, the Consortium take the decision to use parametric models (presented for the first time in WP2) instead of Bayesian Networks for integration into the Decision Support system.

Work-package 5 – Decision support tool
Central activity within the work-package was the development of a system able to assist the crew to make decisions for optimal energy management on-board the ship.
The concept is the tool to utilise data from the present and past in order to propose means of reducing the energy consumption, not only in real time but for the scope of future planning as well.

As a basis for the identification of the required functionalities of the DST, the principles of operation of the decision support tool were defined and summarised in just few bullet points corresponding to main tasks performed by the system:
1. Identify current status of the ship’s plants;
2. Identify deficiencies in operation of the ship’s propulsion system and auxiliary power plants;
3. Identify potential corrective actions for improving efficiency of operation;
4. Verify feasibility of the corrective actions and prepare list of recommended actions;
5. Report points 1.-4. to the user and accommodate for the user’s input.

During this phase, MV Olympic Champion,a ROPAX vessel operated by ANEK lines, was selected for installation on-board of the DST system.

System architecture is built based on the underlying idea behind to derive operational recommendations based on calculations rather than direct measurements (e.g. calculated vs. measured torque). Such architecture reduces significantly the installation costs and lessens the need for large quantities of data needed for deriving the regression models. Furthermore, use of first principle tools allows for generating wide spectrum of data, including responses to “off-the-average” parameters and such knowledge should reduce overall uncertainty in the decision support process.

Consequently the outlined architecture of the DST is based on 4 interconnected modules serving specific tasks related to the operation of the tool, i.e.:
A. Hardware and/or software interfaces to the ship’s control, monitoring and automation systems;
B. Database tools for storing, managing and accessing of operational data, as required by the tool (and the User);
C. Decision support engine responsible for delivering to the User recommendations on optimised (from point of view if energy efficiency) configurations;
D. GUI for data visualisations and information exchange.

During the installation process, which started in autumn 2014, but took more than 3 months due to many different issues, including vessel availability (unplanned repairs in Perama dock-yard for more than 1 month) and technical difficulties in connecting the new system with the existing automation on-board without prompt support from the maker, such architecture was modified and the amendments aimed mainly at increasing reliability of the prototype.
The major change was introduced to the interaction between hardware interfaces, database tool and the GUI. In the original configuration the DB tools (B) played a role of the central module of the DST platform whereas the actual installation is centred at the DST application and GUI with DB tools playing a peripheral (client) role. In the revised configuration the GUI manages all the underlying data processes (acquisition, parsing, processing and reporting) and therefore there is no need for an external interfaces or plugin to the database. This also makes the database tool and the DST largely independent, a clear advantage particularly during the testing and debugging.

The main process carried out by the module is split into two sub-processes, denoted as “Evaluation” and “Verification and Compliance”, respectively. In terms of the basic functionality of the DST, as outlined earlier, the former process corresponds to points 1-3 (status-deficiencies-corrective actions), whereas the latter process is summarised by the point 4 (feasibility of corrective actions). The decision support process is initiated either automatically (timer event) on manually, following request from the user. Details of the process are presented in the within project deliverable D5.2.

To resume, based on the above architecture, the DST functionality consists of selecting the most appropriate configuration and advice on speed adjustment given the vessel current state. This is can be done due to the diversified nature of the ship power plant (multiple operating engines for propulsion).
The objective function is reduction of overall fuel oil consumption; hence it must be estimated for each engine based on calculation of the load of each operating engine.
The load is calculated based on response surfaces and intermediate measured parameters from the KD-30 automation system and navigation system via ENIRAM’s Data integration cabinet.
The DST graphical user interface shows the status of the consumers (main engines). In addition, advisory actions are displayed. These consist of changing configuration (engaged or disengage engines) and/ or changed of operational speed.

The software REFRESH Decision Support Tool (DST) has been distributed to partners and uploaded as technical deliverable (D5.4 - Development of hosting software and integration of supporting models) in a form of a single executable file Refresh_DST.exe accompanied by several libraries and ancillary files.
Installation and use of the software was also detailed into a specific cover document.

In particular, the DST software was developed as a standalone 32-bit MS Windows application.
The particular choice of development environment (Embarcadero RAD Studio XE3) was primarily dictated by ease of GUI development.
Furthermore, the above mentioned RAD studio supports the application development and runtime platform FireMonkey allowing for development of applications for multi-devices (including portable devices with non-Windows operating systems) with highly customizable controls. The latter feature was particularly important for development of user-friendly, transparent and intuitive GUI. Of course developers (leaded by SAS, i.e. Brookes Bell) specific expertise was very important, in such decisions, proposed mainly by WP leader and discussed at Steering Committee level taking into account best solutions generally available to achieve WP5 targets and also developers skills and experience.

As reported on deliverable D5.4 it has to be noted that the DST in its current form is an early prototype. Furthermore, the development needs to be perceived in the context of its intended use i.e. on-board support for operation. These aspects created the basic constraints for the development.
In the course of development, due to (mostly technical) challenges, the plan of development had to be reviewed in order to reflect the knowledge gained in the process.
The top priorities were given to:
- Simplicity of the GUI and clarity of visual information – the crew is “flooded” with information hence the DST aimed at presenting only the most important information and in a self-apparent way. The purpose of DST is to assist not to distract;
- Prioritising goals – the advisory actions need to be clearly defined and of measurable benefits as any dubious or questionable recommendations would not be followed;
- Recommendations must play informative role and not to take the form of requests;
- The tool needs to be easy in use and reliable.

While implementing the software the developers followed the rule that “less is more”, i.e. less overall number of functions but in the areas where there are biggest benefits expected; underlying idea behind this philosophy is that the need for additional DST functionality would grow should its value were proved. This is a key element in order to understand possible future application of such DST concept based on energy models, because a market oriented approach favours a step by step approach rather than relatively big investments to get full level of detail on all systems.

As reported on technical deliverable D5.5 the prototype DST installed on-board MV Olympic Champion was tested on-board and in laboratory and satisfied the set evaluation criteria, as follows:

1. The On-board Decision Support Tool system is installable by the end-user.
2. The On-board Decision Support Tool GUI presents the data correctly and according to the data that it reads from the vessel’s systems and the manual inputs. The GUI is operable and responds to the commands of the user.
3. The On-board Decision Support Tool presents the operational recommendations via the GUI.
4. The Onboard Decision Support Tool was not fully operable in the field testing on-board Olympic Champion. However, the functionality was complemented after the field testing, and in the later laboratory testing round, the DST was found to be operable. Thus this criterion is judged as being satisfied, with remark of the later done amendments.

The task leader, i.e. ENIRAM, in addition to the above, provided us with further observations and recommendations, including the following:
- The GUI operates fluidly and the values are clearly presented.
- The GUI doesn’t cope well with different screen resolutions and resizing of the application window. This in understandable at this phase of the application implementation, but note is made to record the fact.
- It is recommended that the input method of the user-input data is re-evaluated. The rolled dial is not quite self-evident for the user, and the dial limits the range of input values.

Work-package 6 – Concept demonstration

All the main findings and results from the previous project activities were verified and demonstrated through specific demonstration cases.
As planned, such activity addressed Bulk Carriers, Tankers, Containerships and ROPAX vessels.

The ROPAX case addressed mainly the DST functionalities, flagging up the vessel’s data availability issues and the prototype status of the software that, of course, proposing an innovative concept making use of energy models to run simulations rather than usual simple data trending analysis, will require to be improved progressively after REFRESH project completion. Among the conclusions taken by ANEK with Tecnoveritas and S.C. supervision it is noted that the DST and its application can be directly/reasonably quickly transferred to a ship of similar propulsion configuration (diesel mechanical, twin-screw CPP) but cannot be directly ported to other ship types. This is mostly due to the fact that in case of simpler ships (e.g. single ME, directly driven fixed-pitch propeller) the energy saving potential will be related to factors such as trim or draught, i.e. these that did not played significant role in the case reported in the foregoing.
Nevertheless, the methodology presented here and advanced during the REFRESH project is universal and can be transferred to any ship type or, indeed, to any power plant.

The tanker demonstration case, although no energy saving opportunity endowed with undoubtedly good ROI was found, successfully proved how energy models can be deployed to evaluate impact of actions such as the benefits of a “wisely time-managed” ballast handling and the influence of different loading/discharging sequences.

Through the Bulk Carrier demonstration case, the operation of the integrated sea water/ fresh water cooling system of a Handymax bulk carrier was investigated through energy models simulation.

Having run several different cases of operating the system pumps, including constant and variable speed (i.e. ref. to Variable Speed Drives retrofitting solution) were examined, with results and conclusions confirming successful application of the energy models based simulations.

Benefit from VSD installation was then quantified through simulation.
If on one hand the simulation tool proved to be working properly and could be used as tool for evaluating the energy saving potential for the ship auxiliary systems and the techno-economic study of the installation of variable frequency drives for the ship main pumps and fans, on the other hand the specific retrofit was not found viable for the specific existing vessel. It would make sense; form a return of investment point of view, for a new building rather than a retrofit for this type of vessel.

The containership demonstration case addressed mainly the use of Bayesian Networks, demonstrating their application and also highlighting the link between their effectiveness and input data quality.
The analysis presented showed how the Fuel Oil consumption of Main Engine, Boiler and Diesel Generators seem to be influenced by some of the variables recorded in the noon reports according to the findings of the Bayesian Networks and the regression analysis.
These results were obtained from the available data which represent a small dataset, which need to be considered while noting that the relationships derived for the Boiler FO consumption were weaker than the ones derived for the Main Engine and the DG consumptions.
The uncertainty about data quality is the reason why TV concludes that, before being able to use the models presented here to predict the Fuel Oil Consumption on board the container vessel, these relationships need to be scrutinised and their validity checked.

Potential Impact:
The REFRESH project aims to make the following contributions towards the greening of shipping operations:
- Provide a tool for the quantification and rationalization of energy management on-board of commercial ships;
- Offer an optimization capabilities to be integrated in the decision support system for energy efficiency, allowing the crew to identify the optimal energy efficiency oriented operation of the on-board systems;
- Elaborate on the development of a monitoring methodology building on the collection of operational data, to interact with decision support and thus support the crew in daily operation and the ashore personnel in fleet management in the medium and long run;
- Elaborate on the compilation of a set of operational guidelines along the lines of green retrofitting and operation of the ships under consideration.

REFRESH end products include the dynamic energy modelling, optimization methodologies for retrofitting and operation, and monitoring methodology (all these tools and solutions were incorporated into the demonstration platform of work packages WP5 and WP6).
During the 36 months of the project several tasks produced viable results that allowed the consortium to move forward in the pursuit of the main objective but also generated some “ready to use” results that can be exploit by the partners, which were taken in consideration while preparing the project’s exploitation plan.
During the project, several energy deficiencies have been identified for different types of ships. In addition, energy saving potential methods have been considered and evaluated according to their technical and financial feasibility.
The outcome of the optimization studies was very important in that it demonstrated there is a significant need for critical review before any ESP solution is actually implemented. Impact of ESPs will depend on ship type, its size and configuration as well as the operational profile of a particular vessel. Thus, there is a need for the instruments with the ability to simulate the various functions of a vessel in the short and long‐term so that potential benefits can be quantified. Such computational platform is necessary for de-risking the decision making process. Moreover, in order to carry out a multi‐objective optimisation the systems has to be exhaustive in order to capture nature and scale of the investigated processes. Furthermore, the outcome of simulations, monitoring and optimisation needs to be gauged against financial instruments in order to properly assess (crucial in retrofitting) period of investments return.
The project developed tools for simulating performance of on board systems and allowing for optimizing their operation in given working conditions. All the tools developed or advanced during the project executions were integrated within a dynamic decision support tool, capable to assist crew at ship’s management.
On the other hand it should be borne in mind that due to the restricted of available data the estimated ESPs and their contribution to the energy plan of ship is limited only for the cases which were tested. Application of these ESPs on a greater sample of ships will evaluate better their contribution to the energy efficiency.
With regard to the quantification of the expected impact, it is estimated that vessel and fleet management has a great deal to gain from the optimization of whole vessel and individual systems’ operation and energy efficiency. Individually, there are claims of 1-2% reduction of fuel consumption (and subsequently CO2 emission) from the optimization of some systems which can add up to a substantial 10-15% for a single vessel.
In particular, as highlighted by NTUA, the utilization of the provided guidelines from the ROPAX and TANKER optimization studies offers promising solutions to improve the overall energy performance of each vessel type.
For what concerns the future of the application of the DEM approach, interesting conclusions were taken during the last REFRESH project workshop, held in Monaco, V.Delta premises, on Friday, February 27th.
It was discussed and agreed that such in deep analysis as dynamic energy models built component by component proved to be feasible for existing vessels only when looking into specific systems where low energy performance was already identified (i.e. either by direct ship-owner experience based feedback or energy audit finding or even a first less detailed energy modelling based analysis). Thus, based on the project experience, the DEM full application to ships looks more promising on new-builds. This applies particularly to new ships built in Europe, thanks to the more complex average vessel type and also to the presence of shipyards already familiar with monitoring and advisory systems (e.g. Meyer Turku, within REFRESH Consortium), which are mainly developed in Europe (e.g. one of the leaders in the business sector is ENIRAM, again part of REFRESH Consortium).

Back to existing vessels application, it was agreed that, after REFRESH project experience as well as general improvements on the subject driven by the largest players in the shipping sector, the challenge is not anymore to be able to model, simulate and advise properly on room for savings based on proper energy performance related data; now the real challenge is to take such a work to low budget, poorly equipped, very simple ships, which are still today the majority of all the vessels sailing.
Thus, in such background, in view of improving operational efficiency and also properly identifying and also financially assessing retrofit solutions, it is suggested to use REFRESH project tools and results to develop and offer to the shipping market a step by step advisory product. But this product will require proper data input, hence investment on on-board measuring tools. Such cost could be afforded only based on room for savings previously identified through a simplified analysis, adapted to the available level of information and targeting to identify where the deficiencies are and only roughly calculate possible savings.

List of Websites:
The REFRESH website is the key point of access to all relevant information of the project, including general information, the public results as well as the private section. The public section contains an introduction to the project, its concept and objectives, the project partners, news regarding the progress of work and newsletters, and all publicly available deliverables. The private section of the website is dedicated to activities concerning the execution of the project and it is accessible only by project partners.
The project web page is accessible through the following web address:
The general structure of the web page is the following:
- Home page
- Background and Objectives
- News
- Participants
- Work Packages
- Deliverables

Project's coordinator contact
V.Delta ltd