Final Report Summary - SAFEWIND (Multi-scale data assimilation, advanced wind modeling and forecasting with emphasis to extreme weather situations for a secure large-scale wind power integration) Executive summary:The European project SAFEWIND has developed leading-edge research in short-term forecasting of wind power by:-Delivering state-of-the-art solutions to facilitate large-scale integration of wind energy into electricity networks. -Bringing its solutions close to the business processes of the European power systems industry. -Developing academic excellence and European leadership in the field with more than 125 scientific publications (36+ journal papers).-Creating worldwide business opportunities for high-end European technology. SMEs in the project already use the new knowledge acquired to provide forecasting services. -A successful public workshop presented the project results to a broad public from academia and industry. Project Context and Objectives:I. Project context:Due to the variable nature of the wind resource, the large-scale integration of wind power causes several difficulties in the operation and management of a power system. Short-term forecasts of wind generation from a few hours up to a few days ahead are necessary for carrying out various management tasks related to the integration of wind generation into power systems (i.e. quantification of reserves, optimal scheduling, congestion management, optimal trading strategies a.o). In the past 20 years, considerable research has been carried out in the field of wind power forecasting leading to several operational tools. However, the resulting wind forecasting technology presents several bottlenecks. Depending on the sensitivity of end-users, the quality of the forecasts is not adequate for about 10-20% of times. In situations such as weather fronts or wind speeds near cut-off speed, the forecast error may reach high levels. Large forecast errors significantly reduce the value and acceptability of wind power due to their impact on the grid, the penalties they involve when trading and other consequences. II. Scientific and technical objectives:The project addresses the following scientific and technical objectives :-Definition and identification of extreme events Forecasters, energy producers and grid operators have different views on what extremes related to wind generation are. One of the objectives of the project is to produce a catalogue which identifies and classifies extreme situations as a function of their origin, nature or impact.-Large-scale vision of wind power forecasting by the development of an adequate information management infrastructure The focus is to consistently collect, store and distribute the measurement data that are needed to assess the prevailing meteorological conditions over Europe with regard to wind energy. The related vision consists in having for the first time a coherent dataset that is based on standard meteorological data as well as wind farm data, which in combination represents the European 'wind energy weather'. The knowledge of the prevailing weather situation at a certain time is a pre-requisite to detect relevant large-scale extreme events. The work in this part will be directed towards the design and implementation of a slim but robust data management system that will host the data. -Alerting and Data Assimilation Techniques for Improved Short-term Wind Power Predictability The aim here is to develop methods to adequately monitor and assess the weather situation over Europe in order to detect severe deviations in the wind power forecast due to extreme events. Then, react on such a deviation by issuing suitable alerts to users that a forecast error is occurring, and by producing improved updates of the prediction in the short-term (0-6 h).In order to compare the latest forecast field with the prevailing meteorological situation over Europe different data assimilation techniques are developed ranging from the use of rather basic pattern recognition methods towards more advanced meteorological analysis techniques based on surface measurements and wind farm data. The aim is to obtain a fairly precise picture of the current 'wind energy weather' including not only synoptic weather data but also wind farm measurements. This snapshot of the current weather situation is used to detection extreme errors in wind field and wind power prediction and is the starting point to calculate short-term updates. A major innovation of this work, lies in the direct inclusion of wind farm data into advanced data assimilation techniques focussing on the variables that mainly determine the wind power output. -Optimized Ensemble Forecast Systems Applied to Wind Power Prediction The project aims at delivering the meteorological component for skilful, probabilistic wind power predictions based on ECMWF's Ensemble Prediction System (EPS). The work on the EPS aims at improving ensemble forecasts (wind and wind power) at all forecast ranges (0-15 days). Special attention is paid to forecast extreme events in the medium-range with higher accuracy in order to facilitate the integration of wind power in the power system in any weather situation. A set of EPS configurations will be evaluated using newly developed probabilistic skill score measures that are based on observations of different sources (e.g. certified WMO measurement sites, installed LIDAR measurements, etc.).Emphasis will be given to the combination of high resolution deterministic forecast products with ensemble predictions by minimizing probabilistic skill scores. The optimally combined meteorological forecast system will be linked to wind power forecasting tools and improvements in wind power forecasts will be evaluated on all relevant time scales that are interesting to end-users. -Novel Methods for Wind Power Forecasting and Extremes Specific attention is given to predicting extremes situations in the short to medium term (up to 2-3 days ahead) with purely statistical and probabilistic methods accounting for the evolution of meteorological variables, possibly indicating different weather regimes. Due to this time scale, this work relies more on Numerical Weather Predictions (NWPs) than on the use of measurements. A large component of this work relates to the question of estimating and communicating prediction uncertainty, as this is a crucial aspect of wind power forecasting, and also to warning forecast users about potential forthcoming extremes. In parallel, other objectives are related to the aim of decreasing the average prediction error, the risk exposure due to the large errors in wind power forecasts for the case of point forecasts, and to a more robust prediction of extreme events. For that purpose, novel methods will be developed for regime-switching modelling (and forecasting), for conditional forecast combination, as well as for better accounting for cut-off events when modelling the curve for the conversion of wind to power.-Wind Resource Assessment vs Predictability One of the objectives of the project is to explore synergies between the forecasting and resources assessment areas. An innovative concept will be examined which consists in taking into account the predictability of a potential site for a wind farm as a design parameter in the resources assessment study. This is expected to be beneficial for the case of projected wind farms which are going to participate directly in an electricity market rather than a feed-in tariff system. In that case, the pay-back of the investment depends among others on the revenue from the participation to the electricity market. However, such revenue is reduced because of penalties induced from forecast errors. Thus, i.e. among two potential sites, one in complex terrain with high resource but low predictability and another in flat terrain with low resource but high predictability a compromise between resource and predictability might be beneficial to consider. -Assessment of benefits from new measuring technologies for better estimation of external conditions, resource assessment and forecasting. Wind speed measurements are the basis for power production calculations and forecasting. As wind turbines have grown larger and higher, with their rotors covering even larger portions of the boundary layer, larger discrepancies are being reported. There are strong indications that these discrepancies are mainly due to the measurement of the wind speed at hub height, which however cannot any longer be considered as adequate. New remote sensing techniques such as the Lidar device, being investigated within the UpWind project, offer the possibility to measure the wind profile (wind speed, wind direction and turbulence) at more heights both below and above hub height in a competitive way without the need of huge met towers. The knowledge of the wind profile over the rotor will remove much of the uncertainty associated to the power measurement and will make power predictions more accurate. A more detailed input at more levels will also enhance the performance of the forecasting software. -Demonstration of operational benefits A validation phase, in synergy with the ANEMOS.plus project, permits to evaluate in realistic, operational conditions the merits of the most promising models developed here.Project Results:To implement the project objectives, the various tasks of the project were organised into 8 work packages and supported by a dissemination and a management work package. The foreground was reported in the form of contractual deliverables as well as project deliverables. A great number of them is public and available through the web site of the project. Both of them make a total of around 50 reports. Beyond contractual reporting, the partners of the project have been very active in disseminating the project results. In total, by the date of this reporting, a number of 36 peer-reviewed journal papers have been accepted and published. 3 books have been supported by the project. Moreover, a considerable number of results were presented at international conferences in the form of posters (some of them with outstanding poster awards) or oral presentations. In total, around 95 publication of this type were made. It is thus far from trivial to enumerate or rank all the results. Through the project a specific process was followed to identify highlight results. In this Section, the overview of the results per work package is presented. Reference to the corresponding Deliverables are given for further details. Then, a selection of highlight results is presented. The highlight results include : -The generation of a catalogue for extreme events (Dc-1.2 ). This contains points of views of the different communities including end-users, modellers, meteorologists a.o. -A 100-page report on the State of the Art (Deliverable Dp-1.5) mapping the with detail the wind forecasting technology (details in Annex of Highlights).-Establishment of channels for strategic collaboration between the wind power community and research in meteorology. This is essential for the project to have an impact on the research policies in meteorology that permits to account for the particularities of the renewable energies sector (Dp-1.1).Furthermore, this WP permitted to establish the working basis for the rest of the project through the definition of a set of test cases for the validation of the various research results (Dp-1.5) the data collection in collaboration with the industrial partners of the project (Dc-1.6) and finally the definition of end-user requirements including an assessment of the needs in different climates like in Asia (Dp-1.3). The objective of WP-3 (Design and Development of Advanced Data Management Infrastructure for Large-Scale Wind Power Forecasting) is to consistently collect, store and distribute the measurements that are needed in the project to assess the prevailing regional meteorological conditions over Europe with regard to wind energy and wind power forecasting. The aim is to design and implement a slim but robust data management system that gives project partners a fast and efficient access to quality controlled measurements from different original sources in a standardized format. Initially the data requirements for the project were defined (Dc-3.1). Then, the data management system was designed and implemented and a first version was released and supported by a web interface with which the partners can download the data (Dp-3.3). In collaboration with WP-1, the data collected were checked and migrated to the data management system (Dp-3.2). This WP handled the confidentiality issues with the data providers. In the 2nd Period the data management system was finalised and delivered after data migration. The quality control procedures were developed as a function of the characteristics of the data. All a data sets were checked and assigned quality flags. Additionally, a possibility for a direct command-line type access is given. The WP-3 supported the maintenance of the tool until the end of the project and completed successfully its objectives.The main results of this WP can be summarised as:-A dedicated data management system to wind power forecasting applications was fully implemented and populated with data (details in Annex of Highlights). -The data management system provides a common and reliable basis to develop the research work in the project. It includes a quality control functionality which ensures that all data is checked consistently. Access to the system is provided through a user-friendly web interface that permits to visualize and download the data contained in the system.-The data management system is fully operational. First data sets in standard DEPRI format without quality checks were made available to partners in June 2009, the fully error-checked data was available from October 2010 onwards. The last data updates were made available in July 2012. The operational Meteo France data have been updated until the end of the project.The objective of WP-2 (Benefits from New Measurement Devices for Improved System Design) has been to examine if and how lidars and sodars can be used for forecasting at a local scale and how they can improve the accuracy of resource assessment. These objectives have been achieved by a combination of experimental and theoretical studies. An initial state-of-the-art report (Dp 2.1) set the technological scene regarding remote sensing and forecasting at the outset of the project.At the core of the work package we have performed two major field campaigns of lidar and sodar measurements, one in flat terrain where the instruments are expected to perform well, and a second in highly complex terrain, where the flow in-homogeneity is known to give rise to measurement errors. These campaigns were carried out over the first 3 years of the project, starting with the flat terrain campaign. The size of the database was augmented by using in addition previously measured data from the EU UpWind project giving a total of well over 12 months of flat terrain lidar data. Before the complex terrain campaign it was decided to perform an inter-comparison of the instruments first at the flat terrain site. This work is reported in the two deliverables comprising Dc 2.2 (Overview and Assessment).Measurements at the complex terrain site in Spain occurred somewhat later than originally planned due to technical problems at the site and extreme weather conditions. Even so, the measurements provide 14 months of complex terrain remote sensing data including 8 months with simultaneous measurements from 2 lidars and a sodar (a continuation of the inter-comparison, now in complex terrain). The complex terrain measurement campaign is reported in the deliverables comprising Dc 2.3 (Overview and Assessment.)Having gathered the data, the analysis could begin. In work task 2.4 we studied both the flat complex terrain data with particular emphasis on classifications based on stability (flat terrain) and Froude number (complex terrain). An outcome from this work is a proposal for lidar testing to be carried out both in flat and complex terrain in order to give a more complete picture of the abilities of particular instruments. Details of the classification studies and ideas concerning the proposed new combined testing paradigm can be found in deliverable Dp 2.4.Work task 2.5 was concerned with how well lidars are able to measure short term extreme events such as gusts and wind shear. First an experimental analysis showed that the pulsed lidar was surprisingly good at such measurements although with some attenuation in comparison with cup anemometer measurements as expected. The continuous wave lidar has only a duty cycle of around 20% at any given height (assuming 5 measurement heights) and this seriously impedes its ability to register extreme events. A theoretical model for predicting this measurement performance would first need to know how each instrument measures spectra of wind speed fluctuations. We therefore examined how the pulsed lidar measures spectra, a non-trivial task as it turned out but with surprising and interesting results. It turns out that because of the multiple beams at certain frequencies the same wind will be sampled twice - giving rise to large peaks in the spectrum. This result was confirmed by experimental data. This work has been published and the entire results for the work task can be studied in Dp 2.5.After the mid-term assessment it was decided to perform an extra task in which lidar data measured at distances of 1-2km would be studied to see to what extent it is possible to predict wind shear over this range. Extra resources were allocated and a suitable data set was identified. The results are promising and indicate a clear benefit of using lidars for very short term forecasting. These findings will be submitted for journal publication shortly.Work task 2.6 was concerned with examining how the concept of the equivalent wind speed developed as part of the UpWind project, could be used to improve resource assessment. Firstly we looked at using the equivalent wind speed in the site calibration procedure prior to a power curve measurement in complex terrain. It was found that a slight decrease in the uncertainty could be achieved. More significantly we found that the equivalent wind speed introduced to the estimation of the annual energy production at a proposed new wind turbine site could reduce the uncertainty by several percent. This is a major new finding and potentially will change the way resource assessment is carried out in the future. Both these studies are to be found in the two deliverables comprising Dp 2.6. A journal paper will shortly be submitted explaining our findings concerning using the equivalent wind speed for annual energy production estimates.All the measured remote sensing data have been made available to the project participants in a database. This provided a central and logistically efficient means of storing the data and making it readily accessible. A MySQL database server was used for this. A brief summary of the database and the contents is given in Dp 2.7.In conclusion Work Package 2 has made some significant findings concerning the use of lidars for short term forecasting and has made important advances in understanding how lidars register extreme events.The highlight include:-A database of lidar data supported by 2 measurement campaigns at flat (Denmark) and complex (Spain) terrains in SAFEWIND and synergy with UpWind. -Innovative methodology for the evaluation of Remote Sensing instruments (details in Annex of Highlights). -Understanding of how lidars measure extreme winds (details in Annex of Highlights)-Improved energy production estimates by accounting for the wind shear. Identified an important new use for the equivalent wind speed concept in resource assessment (details in Annex of Highlights).Work package WP-4 of the SAFEWIND project is called 'Alerting and Data Assimilation Techniques for Improved Short-term Wind Power Predictability' and focus on the shortest-term forecast range (0-6h). New methods and/or new and existing data sources shall be investigated to adequately monitor and assess the (energy) weather situation of Europe for detection of severe deviations in the wind power forecast. Severe deviations shall trigger suitable alerts that are issued to users that a forecast error is occurring. The alert will be combined with an improved update of the prediction in the shortest-range (0-6 h). Updates of wind power forecasts are of high importance to grid operators to activate reserves in case high deviations to previously forecasted wind power are likely to occur. Furthermore wind park operators that sold their wind energy at the power marked and are likely to fall short, have a high interest to buy balancing energy at the spot market rather than to pay high penalties in case of non-fulfillment of their delivery contract. A major innovation of this work package lies in the direct inclusion of wind farm data into advanced data assimilation techniques focusing on the variables that mainly determine the wind power output. Hence, in addition to the standard meteorological observations, like e.g. 10m wind speed, also wind speeds at hub height from actual wind farm sites are considered. Moreover, the work package aims at providing new algorithms that automatically detect relevant differences between forecasted and observed weather situations. Below the main results are given as a function of the initial objectives of the Work Package:-Characterization and modeling of the spatio-temporal evolution of wind or power prediction errors for simulation of forecast correction purposes The work has been carried out by the partners ARMINES and DTU.IMM (now DTU Wind Energy) in Task 4.1. The excellent wind power data provided by EnergieNet.DK was used to study the propagation of errors in wind power forecasts in dependence of the meteorological situations (e.g. wind speed and direction). A statistical model based on Conditional Parametric Vector Auto-Regressive (CP-VAR) has been developed to reduce the one hour ahead RMSE forecast error up to 18% by including the information of the forecast error of surrounding groups of wind farms. Details of this work are given in Deliverables Dc-4.9 and Dp-4.1. The following two objectives :-Data assimilation to produce updated wind fields over entire Europe using observations and latest high resolution NWP -Nudging for wind energy forecasts using real-time observations from different spatially diverse sources such as meteorological or wind farm data to produce a wind analysis for numerical weather prediction models refer to Tasks 4.2 and 4.4 and have been executed by the partners UNIOL and DTU.RISOE (now DTU Wind Energy).Both partners successfully used the mesoscale WRF (Weather and Research Forecasting) model with its data assimilation capabilities Four-Dimensional Data Assimilation (FDDA) and 3-dimensional variational data assimilation (3dVar) to assimilate meteorological observations and nacelle wind speeds. Due to compute limiting resources the model domain had to be restricted to the North Sea and surrounding countries. The work was not performed under real-time conditions, i.e. time delays for input data and observations have not been considered, yet. However, it has been successfully shown that conventional and new observations types (e.g. nacelle wind speeds) are able to positively alter the WRF forecast despite the known fact that usually in data assimilation the effect of some hundred observations is rather limited compared to the very high number of grid points of a mesoscale model. Details of the work are given Deliverables Dc-4.9; Dp-4.2 Dp-4.4. The following three objectives : -Development of methods for updating the wind power forecast based on a correction of the wind fields with the derived wind map -Detection of extreme wind power prediction errors with the help of the meteorological fields of wind vector, surface pressure and temperature to identify phase-shifts, strong intensification of low pressure systems, or potential cut-offs of wind turbines -Development of an alerting framework that issues alerts to users if a large prediction error is detected including different levels of severity have been worked out by the partners UNIOL and ENERGYMETEO in Task 4.2 and Task 4.3. It was found that continuous monitoring of the current weather situation performs best utilizing observations of the mean sea level pressure. Those are used to obtain an AdHoc-Analysis that is compared with forecasted pressure maps. It was initially planned to compare detected and forecasted wind power to each other, but local effects due to surface roughness and orography prevent wind speed observations to be as representative as pressure observations. Furthermore, it is of great value that from the AdHoc-Analysis of mean sea level pressure the propagation of fronts can be detected. Thus, ENERGYMETEO developed a set of feature tracking modules and assess the deviation between detected weather situation and forecasted weather situation over the difference in pressure gradient that is correlated with wind power deviations. The developed Alarming Module to issue warnings of deviations from day-ahead forecasts offers different sensitivity thresholds and was successfully applied to the Danish (EnergieNet.DK) and 50Herzt (Vattenfall) control zone. Details are given in Deliverables Dc-4.9 Dp-4.3 and Dc-4.8. -Classification and reconstruction of extreme situations based on principal component analysis. The develop methodology will use multivariate methods for decomposing weather situations into modes in the time-frequency space, and permit to detect weather situations that promote extreme wind situations.The work on this objective has been carried out by the partners UCM and CENER in Task 4.5 and 4.6. The dimensionality of the large-scale atmospheric database has been reduced by Principle Component Analysis. The methodology of the ANalog PAttern Finder (ANPAF) has been developed, tested and applied to various test cases. In Spain the daily mean wind speed and maximal wind speed gust is estimated for more than 23 sites and in Denmark, Germany and Ireland the daily mean wind power production is estimated successfully with analogs in a deterministic and probabilistic mode. The results outperform climatology and require the instantaneous geopotential field at 1000 hPa as input. Details of this work are given Deliverables Dc-4.9 Dp-4.5 and Dp-4.6. -Development of a quantitative measure that indicates severe variability in the wind power production The work on this objective has been carried out by DTU.RISOE in Task 4.7 of the project. Two different methodologies have been proposed and applied to the problems of characterizing the predictability and climatology of severely variable wind speeds, which are implicitly linked to severely variable power production. With the help of classification techniques the relation to synoptic situations was successfully demonstrated. The finding that open cellular convection leads to severe wind fluctuations was even successfully modeled with the mesoscale model WRF and can be regarded as a surplus of DTU.RISOE in this task. Details of this work are given in Deliverables Dc-4.9 and Dp-4.7.As detailed above all set objectives in work package 4 have been achieved successfully. The Alarming Module by ENERGYMETEO to alert if the estimated power error (deviation) exceeds certain thresholds has been integrated in Anemos.eXtreme and runs for the 50Herz (Vattenfall) control zone. Furthermore the statistical forecast error propagation module by DTU.IMM has been integrated in Anemos.eXtreme by ENFOR to provide shortest-term updates on Danish wind power to EnergieNet.DK.Its objectives can be sorted into two categories, which are(i) the further development of appropriate verification and diagnostic tools, and(ii) the improvement of the quality of deterministic and probabilistic forecasts of wind over Europe.Regarding the former category (i), the suitable (already existing) verification approaches to wind ensemble forecasts have been reviewed. Subsequently new concepts have been proposed and developed giving more emphasis to the verification of ensemble forecasts (and derived products) for extreme events. This has consisted for instance of further considering observations (from synoptic stations) in the verification task, while accounting for their uncertainty. It has also involved the evaluation of ability of Weather Types and Extreme Forecast Indices (EFI) that inform about extremes and can be used for early detection of extreme winds, to come in the early to medium-range. On the more research side, it has been aimed at developing new ideas for the verification of ensemble forecasts (of wind and wind power) by seeing them as trajectories in the future, and by looking at their functional and marginal probabilistic properties.The main results can be summarised by the following highlights:-Verification of ensemble forecasts of wind speed against observations, accounting for observational uncertainty, for the whole European area.-New ECMWF deterministic and probabilistic products : High-resolution (deterministic) and ensemble prediction system (probabilistic) 100-meter wind fields (details in Annex of Highlights).-Special emphasis on the ability of Weather Types and Extreme Forecasts Indices to inform about coming extreme events in the early to medium-range (details in Annex of Highlights). -Bivariate adaptive recalibration of ensemble forecasts of (u,v)-winds, with improvements of probabilistic skill scores up to 15% in the early range over Europe-Proposal of new approaches to the evaluation of ensemble forecasts as trajectories.-UNIOL demonstrated successfully the added impact of 100m winds and calibrated 10m winds (CPS) for regional and point probabilistic forecasts. Improvements in CRPSS up to 25 % and in RMSE up to 50% have been obtained (details in Annex of Highlights).The objective of WP-6 (Novel Methods for Wind Power Forecasting and Extremes) was to develop and assess new methodologies for forecasting wind power generation using statistical concepts, which would answer the new varied needs of forecast users with focus on extreme events. Extreme events were indeed identified as comprising one of the most pressing challenges when it comes to an optimal integration of wind energy into existing power systems and electricity markets. They translate to severe threats to systems adequacy and operations safety, while yielding extreme costs for the actors of the power systems and electricity markets: power producers may have to deal with additional costs linked to additional maintenance and repair or market penalties, while system operators have to guarantee a proper system operation at all times. These so-called extreme events are perceived in a fairly different manner depending on the type of forecasts user. In parallel the meaning of 'extreme' in the case of the wind power application is seen differently by meteorologists, forecasters, power systems engineers, etc.The original objectives of this work package related to lead times in the early-medium range (up to 2- 3 days) while works in WP-4 and WP-5 were somewhat more focused on the short-term (a few hours ahead) and medium range (more than 2 days ahead), respectively. Shorter lead times (in the order of 10 minutes to 6 hours) were considered in WP-6 for some of the investigations related to new predictive densities and regime-switching modelling for instance since it is where the highest benefits from these new approaches could be shown. Also when looking at the use of meteorological ensemble forecasts as provided by ECMWF as input, longer lead times (up to 5-7 days ahead) were studied since an acceptable level of predictability could still be reached. Whatever the forecast length and temporal resolution, a main objective was to propose, develop and evaluate forecasting methodologies that could allow issuing: -more accurate point forecasts, since comprising the forecasting product that this the most employed by forecast users in practice,-more reliable and skilled probabilistic forecasts (in various forms), since such types of forecasts may bring the highest value as input to decision-making,-more meaningful information about forecasting uncertainty, since too complex forecast products are difficult to appraise by most forecast users, and may then provide then with misleading signals.As a combination of these three points, particular attention was given to the case of extreme events, which may be of meteorological nature e.g. the winter storm Xynthia, or of particular importance to market participation and power systems operations e.g. very large forecasting errors. Some of the modelling proposals and investigations permitted to gain new insight on the dynamics and predictability of wind power generation, for instance through the regime-switching and forecast combination tasks telling us how offsite observations may inform of regime switches and how local measurements may already provide valuable information on power generation regimes. These investigations came as a natural extension of previous works on wind power forecasting. In contrast, other work highly benefited from a collaboration between statisticians, energy forecasters, meteorologists and forecast users, for instance for the case of the optimal communication of forecast uncertainty and for the definition of event-based prediction problems. Finally, new ways to approach the wind power forecasting problem were though off through the work in this work package. As an illustrative example, the concepts of scenarios and event-based prediction open the door to the development of new forecasting methodologies and of forecasting products that may be of utmost importance to a wide range of forecast users.A selection of highlight results of this work package include (details in Annex of Highlights) :-A panel of approaches to probabilistic forecasting-A novel method for ramp forecasting with temporal uncertainty using ECMWF ensembles-New probabilistic forecasting product: Scenarios-Quantile forecasting using variability indices-An artificial intelligence based regime switching model for extreme events-A protocol for the evaluation of wind power probabilistic forecastsThe aim of WP-7 (Wind Resource Assessment vs Predictability) is to explore synergies of the forecasting and resources assessment areas. Its objective is double fold: The first one is to investigate an innovative concept; the value of Wind Power Predictability in Spatial Planning or Investment Planning (resource assessment) phase of a wind farm (Task 7.1). For this purpose, predictability is considered as a decision factor in the planning phase of wind energy by:-anticipating the economical impact of (lack of) predictability at different scales and therefore-defining strategies for optimum penetration of wind energy at European, regional and wind farm levelsThe second objective is to study the value of Extreme Wind Predictability in Site Assessment (Task 7.2). For this purpose new techniques for 50-year extreme wind (Vref) assessment based on numerical weather prediction outputs are proposed to:-avoid using historical observations which are scarce and of low quality (lack of homogeneity and representativeness)-produce a unified methodology that allows for trans-national mapping of Vref Regarding the first objective forecasting models were run on hindcast mode to produce predictability-related information that can be used to anticipate operational costs and produce a better assessment of the cost-benefit of wind energy deployment. Several case studies have been investigated in the project to illustrate this approach considering different end-user perspectives. Meteorological mesoscale models gave us the opportunity to evaluate wind power predictability and other forecast skills. In addition to a large capacity factor, small wind variability is also desirable for wind power integration. Forecast skill maps were generated. They can provide useful information for spatial planning and for analysing the sources of forecasting errors, either by topographical effects or meteorological phenomena. For the case of wind farm developpers, in countries where wind farms participate in electricity markets rather than in feed-in tariff schemes, the level of predictability might an impact on the financial performance of the wind farm. It was investigated if this factor should be taken into account as a design in the investment phase of the wind farm. It was found that including predictability as a decision factor in the planning phase of a wind farm has a very low weight compared to the capacity factor, which is evidently the main feasibility driver. For the case of Denmark and the Nordpool market, only when wind farm aggregation is considered, predictability can become more important but only explains 0.15% of the total revenues variance. Nevertheless, predictability can be already assessed during the planning phase to provide an indicator of the quality of wind for grid integration purposes.A pan-European electricity market would offers greater advantage for wind energy traders that can benefit from portfolio effects in larger geographical domains. In a case study investigated, moving from Ireland to a Western European domain that includes a few wind farms in France and Spain, implies a decrease of around 10% in the aggregated prediction error and in the associated market penalties. Regarding the 2nd objective, Vref is one of the key parameters to be determined in the process of wind trubine siting. Defined as the maximum 10 minutes average of wind speed with a recurrence period of 50 years, Vref is directly related to extreme events. The estimated value of Vref has a direct influence on the class of wind turbine that has to be defined for the site, this decision implies a risk if Vref is not properly estimated. IEC 61400-1 standard defines Vref but does not determine the method to calculate it. In this situation, depending on the method, one can obtain different values of Vref. Even more, the statistical methods usually applied to estimate Vref are not properly used if their requirements are not fulfilled; this is the case when applying the Gumbel method with 1 or 2 years of measurements, which is a very typical situation in wind resource measurement campaigns, with measurement periods usually shorter than 2 years.Long-term predictability of extreme winds with meteorological databases and downscaling models is also a research area of great potential considering the large improvements observed in global reanalyses. Still large uncertainties are to be expected due to the difficulties of NWPs and microscale models in modelling stormy weather turbulence, especially in complex terrain. Nevertheless, the results obtained in the Denmark and Spain case studies suggest that Vref mapping can be of engineering value during the planning phases of wind energy. In view of the elaboration of modern wind atlases for spatial planning purposes, it is clearly demonstrated the added value of including predictability of wind power and extreme winds together with the typical wind resource outputs. The result is an enriched wind atlas that can provide a more complete vision of the lifecycle value and cost of wind energy. The highlight results include:-New methodologies and test cases have been presented to demonstrate the feasibility of using predictability during the planning phase of wind energy-Guidelines on how to use predictability during the planning phase have been produced-These new tools allow life-cycle approaches that enable a more integrated promotion of wind energy development -Towards an enriched European wind atlas: predictability and Vref mapping are two new layers that can be integrated in spatial planning activities (details in Annex of Highlights)The aim of WP-8 (Demonstration of Operational Benefits: Anemos.eXtreme) is to implement the most promising forecasting approaches developed in the project into software prototype modules in order to be validated in an operational environment. The targeted approaches are for extreme wind detection, alarming, warning and, if possible, prediction correction will be tested in an operational environment. For this purpose, the respective software modules will be run operationally for a number of end-users including TSOs, utilities and wind farm operators.As basis for the demonstrations was used the existing wind power prediction platform Anemos (developed in the previous Anemos project), which provides wind power predictions on a professional basis. The aim in SAFEWIND is to extend this platform by integrating the new functionalities and modules developed in the project. The ensemble of the new functionalities, including data management, was implemented as a suit of modules plugging into this software framework: Anemos.eXtreme.For the demonstration of the Anemos.eXtreme modules six test cases have been set up: SONI (TSO of Northern Ireland/UK), EirGrid (TSO of Ireland), PPC (System Operator of the island of Crete), RTE (French TSO), EDF (Utility in France) and 50 Hz (TSO in Germany). For each of the test cases, a selection of the seven Anemos.eXtreme modules have been set-up according to their needs and the corresponding end-user requirements. Details on the demonstrations and the modules tested at each one are given at the highlight below.An additional objective of this work package is the set-up and evaluation of wind power predictions under different climate conditions in India. For the evaluation of the ANEMOS platform under different climatic conditions in India, the platform has been set-up by ARMINES with the support of the partner TERI. The prediction results showed high sensitivity of the models to the climatic conditions dominated by monsoons. Although overall performance was at the level of state of the art, the performance was found to be highly dependent to the type of season, opening perspectives for more season-tailored models. Seven software prototype modules have been developed in the SAFEWIND project and integrated within the framework of the Anemos prediction platform composing an ensemble called Anemos.eXtreme. The purpose of the modules is to make short term corrections of wind power forecasts, provide alerting of upcoming extreme events, predict special events like cut-off events, and provide estimations on the expected level of the uncertainty of the wind power predictions. These modules have been demonstrated in an operational environment and have been evaluated using real data. The results have been presented and discussed for each of the modules in this report.Potential Impact:I. Potential ImpactI.1. Scientific and Technological ImpactThe project had a considerable number of scientific production (36 peer reviewed published papers at journals, more than 90 poster and oral presentations at international conferences, 11 PhDs etc.). The project this way shaped the state of the art in continuity of the two previous projects ANEMOS (R&D FP5) and ANEMOS.plus (RD&D, FP7) with a total of more than 250 publications. ANEMOS was the project that marked the passage from 'deterministic' to 'probabilistic' forecasting, ANEMOS.plus in turn demonstrated how probabilistic forecasts can be optimally used for decision making in power system management for optimising wind integration. SAFEWIND pushed the knowledge and provided solutions for the challenging and extreme situations with high impact to end-users. The project developed European excellence in the field at international level.The project delivered a variety of advanced models for wind power forecasting. These models provide either forecasts and uncertainly estimations of improved quality, compared to the state of the art, or complementary forecast products (i.e. prediction risk indices, ramps forecasting, variability forecasting, extreme events forecasting like cut-offs etc), that can be of high added value for end-users in the various applications for managing wind power. These new models are designed to respond to the increasing needs of end-users. The project goes one-step further, to the operational validation of the most promising tools. This permits to evaluate the benefits of their application, get feedback from end-users on how these tools can be used to increase wind penetration, and finally get better knowledge of the operational constraints and consider them in the modelling process itself. The impact of this process is a continuous improvement of the wind power forecasting technology. Beyond short-term forecasting, SAFEWIND studies how wind predictability can contribute to spatial planning and the investment phase of wind farms. It carries pioneer research on this topic identified also as relevant in the frame of the FP7 Eranet+Topic on the New European Wind Atlas. The results and publications of SAFEWIND will most probably permit to optimise and accelerate the research work in this future project. The geographical dimension of wind forecasting is considered in the project; the project aims to show how improved performance can be achieved for regional or national forecasting using geographically distributed information. The geographical dimension extends to the European scale. Given the deployment of wind energy in Europe, the project aims at developing research towards a European vision for wind power forecasting. This is expected to be necessary in the coming years in the frame of an enhanced coordination among Transmission System Operators (TSOs) to manage large wind capacities. I.2. Impact on competitiveness of European Industry and EmploymentIn the longer term, the impact of the project results is expected to be high since they will permit a smoother integration of the large amounts of wind power projected in the next years/decades. Facilitating penetration, increases acceptability by operators. The risk that problems in the grid become a bottleneck for further development of wind installations is reduced. The development of good quality forecasting solutions facilitates the development of the forecasting service business that European companies can provide. The excellence in research in relation to the well established methodology within the project to bring its results rapidly close to the daily practice of end-users have a positive impact on the business opportunities of the industrial partners participating in the project (i.e. 3 SMEs) some of them have developed leading positions in the field. The SAFEWIND project includes research institutes and companies that provide today software products, services and consultancy related to wind power forecasting and power system management. The project will permit them to enhance their position at European and international level through the development and validation of innovative tools destined to facilitate large-scale wind energy integration.A strategic result of the project has been the successful synergy with the meteorological community. The project developed research on meteorological models oriented for first time to wind energy. This is a major contribution towards improved predictability of the renewables. In general, research in meteorology in the past was general and for the benefit of multiple applications. The collaboration with ECMWF provided awareness of the needs of the Energy Community for dedicated products. The choice of ECMWF is expected to maximise this impact thanks to the specific role it has has with respect to the national meteorological organisations in Europe. I.3. Environmental and Societal ImpactThe advanced tools proposed in this project are expected to contribute critically to the wide integration of wind power in power systems. As a consequence, higher wind integration enables a substantial reduction of GHG emissions and the mitigation of climate change. Accurate and reliable prediction tools alleviate the consequences from the fluctuating nature of wind resource. By this way, higher wind integration is achieved, as explained before, with measurable benefits such as: -The development of employment through the expansion of wind power related industry.-The sustainable development and protection of the environment in the European countries and Islands.-The contribution to the security of energy supply and to economic and social cohesion.All these benefits result in an improvement of the quality of life, and thus contribute to further improvement of the acceptability of renewables.I.4. Impact to Policies, International co-operation, co-ordination with Member State research programmes, and EU wide research networksThe use of forecasting tools enables a higher penetration of wind power in the grids, thereby allowing a larger part of the European electricity consumption to be covered by wind power. Currently, wind power installations reach a natural limit where the maximum wind power generation is commensurate with the minimum load. This typically occurs at 20% regional demand penetration. Due to the envisaged better management of wind power in the grid, this limit can be pushed further out, allowing for more of this indigenous resource to contribute to the European energy demand. This facilitates the deployment of the European priority for Security of Supply. Regarding future Research and Development (R&D) priorities in Europe in the wind sector, the partners of the project have provided input to two Working Groups (Wind Resource and Wind Integration) of the European Technology Platform (TPWind) for the update of the Strategic Research Agenda for the years 2012-2030/2050 and other policy documents. The SAFEWIND objectives are in line with the R&D priorities of ENTSOE, the European Association of TSOs. The project is considered as a smartgrid project with high impact for TSOs. It is classified within the cluster C3 of projects under the T6 Functional Project of high priority 'Tools for pan-European network observability' (pp. 76, 91, 103, ENTSO R&D Plan 2011, 'Towards 2020 Challenges and Beyond', available online at http://www.entsoe.eu). The project results were followed by the Cost Action 1002 WIRE 'Weather Intelligence for Energy', which is a network of organisations from 25 countries and focuses on renewable energy forecasting. Through this Action the results of the project have been widely spread. Finally TERI undertook throughout the project several dissemination activities to INDIA towards the ministry of Energy, Energy Agencies, Wind farm developers and owners, the Indian Wind Turbine Manufactures association, the Indian Wind Energy Association, TSO and others. It has contributed to the discussions in the last 2 years about the new grid code in India and the requirements for forecasting the output of wind farms. TERI brought knowledge from the project to these discussions (i.e . evaluation protocols, performances in the state of the art, accuracy evaluations for Indian wind farms in SAFEWIND etc). II. Main Dissemination ActivitiesII.1. Scientific (Peer-Reviewed) journal papers (36) Up to date a number of 36 journal papers have been accepted and published on the basis of the results produced in the project. At least 5 additional publications are under preparation or submitted and pending for acceptance. II.2. Scientific books (3)Three scientific books or chapters in books were supported by the results of the project. II.3. Conferences (92)A number of 92 oral and poster presentations were made at national and international Conferences with the scope to present the project or specific technical results. Some of them were invited presentations. The exhaustive list is available at http://www.safewind.eu. The following two poster presentations won 'outstanding poster awards' at the corresponding conferences :1.A. Bossavy, R. Girard, G. Kariniotakis. ' A probabilistic approach to forecast ramps of wind power production using ensembles', European Meteorological Society Annual Meeting 2011, Berlin, Germany, 1st September 2011. 2.L.v. Bremen, J. Tambke, N.Stoffels 'Spatio-temporal Smoothing of Wind Power Variability and Forecast Errors in Europe', Poster presentation, In Proceedings of the Annual EWEA 2012 Conference, Copenhagen, Denmark, 16-19 April 2012. In addition the partners participated in the organization of forecasting sessions at various conferences where oral presentations related to SAFEWIND were made. II.4. Invited presentation with political impact (1)The coordinator was invited to present the advances in the field of wind power forecasting at the French SENAT in the frame of a National Committee on the Nuclear Security and the future of the Renewables Sector. 1.G. Kariniotakis, 'La prévision de la production d'électricité d'origine éolienne et photovoltaïque', French SENAT, Oral présentation. Sécurité Nucléaire et Avenir de la Filière Nucléaire -La Maturité des Énergies Renouvelables, Table Ronde: L'Intégration des Energies Renouvelables dans les Réseaux et le Stockage de l'Énergie, 24 November 2011, Paris, France. II.5. Workshops (4)A number of 3 public workshops were organised in the frame of the project starting in September 2011 on the general topic of wind power forecasting in parallel to the ISAP 2011 Conference in Crete, Greece. In March 2012 a dedicated workshop on End-Users of Wind Power forecasting was organised by ENERGINET at Fredericia in Denmark. Finally on the 31st of August 2012 a final workshop was organised at Palais Brogniart in Paris to present the results of a project. The programme included a speech by the Scientific Officer of the project and a Round Table with recognised experts on the perspectives of Forecasting . A number of 100 participants attended.Finally, in February 2012, SAFEWIND participated the InnoGrid2020+ Conference in Brussels on smartgrids as invited project. II.6. Thesis (21)A number of 11 PhDs were launched in parallel to the project supporting it with their results. 6 of them have defended successfully by end of 2012 while the remaining are expected for 2013. A major contribution of SAFEWIND for the candidates was the possibility it gave to the young researchers to obtain valuable data through the project to carry out their work, to present their results at the various technical internal workshop of the project and discuss them with the confirmed experts and the industrial partners. The high number of industrial partners permitted to keep in mind the needs of the industry. In addition the project supported the thesis of 10 M.Sc. candidates in a similar context as the PhDs. All of them have graduated successfully. Finally, TERI (India) hosted 2 undergraduate students for intership related to the wind power forecasting activities I the project. II.7. Website/Application (2)The coordinator put in place and maintains the public web site of the project at the address http://wwww.safewind.eu A private website was put also in place to support intra-project communication (documents repository). At the address https://www. safewind.eu/workspace II.8. Press Releases (4)SONI published a press release at the occasion of the Contractual meeting organised at Belfast. Overspeed published a press release at the Technical workshop organised at Oldenburg in June 2011.ARMINES published two press releases in French and in English following the final project Workshop of 31/8/2012.II.9. Flyer (1)A flyer/factsheet on the project (objectives, consortium, expected results etc) was prepared at the beginning of the project ad distributed to various events. II.10. Articles in popular press (5)Article by G. Kariniotakis at the Special issue of the Magazine 'Tribunes Parliamentaires Européens' entitled as 'Quel avenir pour l'éolien - Which future for the Wind Energy Sector'. The title of the article is 'Quelle recherche et développement pour un leadership européen dans l'éolien? ', 'What Research and Developpment for a European Leadershiop in Wind energy ? '. Appeared in April 2012.The Times (19th August 2010) - 'Weather Eye' column: ECMWF releases new products (100m winds) for the wind energy application.Guardian newspaper (UK) article on probabilistic forecasting, 'Revealaing the real forecast', P.McSharry Oxford University, January 2010. http://www.guardian.co.uk/commentisfree/2010/jan/19/met-office-weather-forecastsA 2 page publication on Wind power forecasting and on Anemos and SAFEWIND at the Public Service Review: Science and Technology - Issue 5, December 2009. It follows editorial by Commissioner Andris Piebalgs focusing on the SET plan for Energy.Article presenting Anemos and the objectives of SAFEWIND in 'MINES AVENIR 2008', September 2008. II.11. Interviews (2)G. Kariniotakis of ARMINES gave an interview for the EWEA Web site on the theme : 'Forecasting Key to Wind energy Future' in January 2013 (see http://www.ewea.org/blog/2013/01/forecasting-the-future-at-ewea-2013/ online) P. McSharry from Oxford University gave an interview to BBC Newsnight on the theme 'Should we switch to a new way of predicting weather? ' in January 2010. http://news.bbc.co.uk/1/hi/programmes/newsnight/8448617.stm II.12. Dissemination with industries and government organisations in INDIA.TERI presented and promoted the SAFEWIND project, specially about the TERI activity and the demonstration of ANEMOS for Indian wind farms to the following organisations:1.Ministry of New and Renewable Energy, Govt. of India2.Gujarat Energy Development Agency, Govt. of Gujarat, India3.The Executive Director, Centre for Wind Energy Technology, India4.Wind turbine manufacturer and project developers including M/s Suzlon Energy Limited, M/s Regen Powertech Pvt. Ltd., M/s Enercon India Ltd., M/s Vestas Wind Technology India Pvt Ltd.5.Wind Farm Owners including M/s CLP Power India pvt. ltd., M/s Indian Oil Corporation Limited, ONGC Limited, M/s Dalmia Cements Pvt Ltd,6.The Indian Wind Turbine Manufacturers Association7.The Indian Wind Energy Association8.The Gujarat Energy Transmission Company LimitedApart from the above, the demonstration activity had been discussed one to one with various other agencies, manufacturers, owners and operators of the wind farm during Delhi Sustainable Development Summit and Renewable Energy India Expo-2011, III. Exploitation of ResultsOverview of Foreground generated in the project:The SAFEWIND has produced the following types of Foreground :-General advancement of knowledge (12 items)-Exploitation of Research and Development (R&D) results via Standards (2 items)-Commercial exploitation of Research and Development (R&D) results (11 items)The Foreground related to the commercial exploitation is mainly in the form of software modules prototypes. The models and algorithms behind these modules were developed in the various Work Packages of the project, evaluated using historical data, and as a function of their performance, they have been implemented into software prototype modules. These prototypes have been integrated into the ANEMOS wind power prediction platform and demonstrated at the demo cases of the project. This last step aimed to validate their operation under realistic operational conditions. The ensemble of these modules is named Anemos.eXtreme. Outline of exploitation plans: The software modules developed in the project are able to function operationally that is, to receive on-line input and to produce on-line output. Each module uses the standardized data formats and/or interfaces defined for the ANEMOS platform. Due to this fact, these modules can run with or without the ANEMOS wind power prediction platform and consequently can be exploited in two ways:-by the owner of the Foreground (module) and fully independently of the ANEMOS platform. No license restrictions exist which would limit the independent exploitation;-in the framework of the commercial activities related to the ANEMOS prediction platform as a whole. The exploitation plan of the ANEMOS platform is outlined below. In this case the modules are sub-licensed to the partner that takes over the exploitation of the whole ANEMOS platform. Exploitation through the ANEMOS wind power forecasting system. All modules fully integrated into the ANEMOS platform will be part in the general ANEMOS marketing activities to be conducted under the responsibility of Overspeed (SME) for the commercial part and ARMINES for the more scientific point of view. Overspeed currently covers also the maintenance of the ANEMOS Platform in the commercial applications. Other partners that are active in the commercial exploitation are ENFOR, Energy and Meteo Systems, DTU Wind Energy and CENER. A commercial web page (see http://www.windpowerpredictions.com online) was created by Overspeed to support these activities. For dissemination purposes an 8-page leaflet was also prepared. If there is interest from a potential customer, like an inquiry or a tender, Overspeed will contact the respective IP owner and discuss the availability and fitness of the module for this specific application, and will negotiate the respective license fee. It is noted that virtually all interfaces for sending data to a module, controlling it during run-time and retrieving data from the module are based on files as much as possible. All data formats used are made available to ANEMOS members (partners involved in the projects ANEMOS, ANEMOS.plus and SAFEWIND), so no restrictions apply in relation to IP issues.Exploitation of the project results individually by each paretner. Most of the partners mentioned above have a solid experience in the commercial exploitation of wind power forecasting and power system management tools and services and also on consulting services. Intellectual Property Rights protection measures.Given that the knowledge produced in the project is mainly software, protection is made through copyright. Most scientific results are published in papers (journals or conferences) and are also protected by copyright, while providing an open access of the results to the public. No patents, design rights, database rights or other actions are currently foreseen taking into account the nature of the IP rights attached to the results. Dissemination and networkingAfter the end of the project the partners decided to continue their collaboration through several activities including:-Networking: keep contact and exchange through social networks like LinkedIn. -Organisation of workshops: to keep the momentum created by the project final workshop in August 2012 the partners decided to organise an annual workshop on wind power forecasting in the next years The first one is organised in 2013 in collaboration with EWEA (Euroepan Wind Energy Association) on the 3-4 December 2013 in Rotterdam (see http://www.ewea.org/events/workshops/wind-power-forecasting online).List of Websites:http://www.safewind.eu/ Related documents Final Report - SAFEWIND (Multi-scale data assimilation, advanced wind modeling and forecasting with emphasis to extreme weather situations for a secure large-scale wind power integration)
