The following 27 posters were presented during extended refreshment breaks, lunches and the drinks reception. (Posters listed by author surname.) All workshop attendees have been given instructions on how to access the poster and oral presentations given at the workshop. EWEA members may access the proceedings free of charge via the Members’ Area or for login information contact Maura.DiRuscio@ewea.org. If you did not attend and are not from an EWEA member organisation please complete the presentations order form.
|Poster title & link to abstract||Poster presenter|
|High-resolution wind forecasts: On-demand and operational forecasting, and observational nudging||Hálfdán Ágústsson, Head of Research and Development Institute for Meteorological Research – Belgingur|
|Probabilistic forecast information for the renewable energy sector – can we bridge the gap between theory and praxis?||Isabel Alberts, Research Scientist Deutscher Wetterdienst (DWD) Offenbach Germany|
|Forecasting Wind Energy Production in complex terrain: impact of mesoscale modelling in a forecasting system||Gianluca Artipoli, Wind Energy Analyst, DEWI GmbH, Italy|
|One-day ahead Power Forecast in China: the Grid curtailment problem||Julien Berthaut – Gerentes, Scientific engineer Innovation and Research Meteodyn France|
|Wind power forecasts in the business process of a Dutch trading house||Arno Brand, Research Scientist Wind Energy ECN Petten The Netherlands|
|Wind Power Forecasting For Trading Purposes: The Importance of Managing Wind Volatility||Marta Soroa, Renewable Energy Consultant en Gnarum tecnología Spain|
|An application of PCA based approach to large area wind power forecast||Simone Sperati, Researcher Sustainable, Development and Energy Sources RSE, Italy|
|Impact of SRTM and Corine Land Cover data on calculated wind, temperature and precipitation values using WRF||Alexander de Meij, Scientific Engineer, Sustainable Development Noveltis Labege France|
|Operational wind power forecasting systems based on physical and statistical models||George Galanis, As. Professor Section of Mathematics, Hellenic Naval Academy & Atmospheric Modeling and Weather Forecasting Group, University of Athens University of Athens,Greece|
|Short-Term Forecasting of Categorical Changes in Wind Power with Markov Chain Models||Amanda Hering, Assistant Professor Applied Mathematics and Statistics Colorado School of Mines Golden, USA|
|The Value of Improved Wind Power Forecasting in the Western Interconnection||Bri-Mathias Hodge, Research Engineer Transmission and Grid Integration Group National Renewable Energy Laboratory Golden, USA|
|Development of dynamical-statistical short-range probabilistic wind prediction model for wind regimes in coastal and complex terrain||Kristian Horvath, Head of Applied Modelling Section Department for Research Meteorological and Hydrological Service, Croatia|
|Statistical Identification of Local and Regional Wind Regimes.||Karen Kazor, Graduate Student Applied Mathematics and Statistics Colorado School of Mines Golden, USA|
|From Bankable Opportunity to Operations – A New Approach||Erik Koppen Senior consultant Environment ARCADIS Nederland BV Arnhem The Netherlands|
|From a Matlab Based Wind Power Forecast to an Integrated EMS Solution||Stéphanie Lakkis, Project Engineer Energy Management Applications, Siemens AG, Austria|
|Short-term Forecasting of energy production using CFD simulations||Matteo Mana, Software Developer, Software Development WindSIm, Norway|
|RapidWind – Rapidly updated forecasts for wind energy production||Malte Müller, Researcher Research and Development Norwegian Meteorological Institute Norway|
|On the wind energy production data deficiencies: simulation from statistical models||Emil Pelikan, Institute of Computer Sciences, Prague, Czech Republic|
|Using Bayesian Model Averaging for wind farm group forecasts||Pål Preede Revheim, Research Fellow Department of Engineering Sciences University of Agder, Norway|
|Evaluation of NWP Resolution Effect on Wind Speed Forecast Quality for a Wind Farm in Central Sweden||Martin Haubjerg Rosgaard, ENFOR A/S and the Technical University of Denmark Wind Energy Department|
|FROM ENSEMBLES TO PROBABILISTIC WIND POWER FORECASTS – HOW CRUCIAL IS THE ENSEMBLE SIZE?||Jeremy Sack, Meteorologist R&D, EWC Weather Consult GmbH, Germany|
|High-resolution surface modelling for extended-range downscaling of near-surface atmospheric fields over Canada||Leo Separovic, Physical Scientist, Meteorological Research Division Environment Canada, Canada|
|EWeLiNE: Development of innovative weather and power forecast models for the grid integration of weather dependent energy sources||Scott Otterson, postdoctoral research fellow, Energy Economy and Grid Operation Fraunhofer IWES, Germany|
|The performance evaluation of WRF Model for Short Term Wind Energy Prediction System for Turkey||Elcin Tan, Instructor Dr. Meteorological Engineering Istanbul Technical University Istanbul Turkey|
|A Short Term Wind Energy Prediction System for Turkey||Elcin Tan, Instructor Dr. Meteorological Engineering Istanbul Technical University Istanbul Turkey|
|A Quantitative Assessment of the Variability of Wind||Darion Turner, Sales Trader, Trading SmartestEnergy Ltd, UK|
|Applying Model Output Statistics (MOS) in the new German Project EWeLiNE for enhanced wind forecasting for renewable power generation||Gernot Vogt, Scientific Researcher Development of Meteorological Applications, Germany|
Poster title: “High-resolution wind forecasts: On-demand and operational forecasting, and observational nudging
Poster Presenter: Hálfdán Ágústsson Head of Research and Development Institute for Meteorological Research – Belgingur
Abstract: Wind turbine and power plant operators rely on accurate forecasts and mapping of the wind resource for safe and successful operation of the utilities.
We present a state-of-the-art operational forecasting system that utilises the AR-WRF model to prepare high-resolution forecasts of wind and weather for any location in the world. The system can both be run in an operational setting (repeated forecasts) as well as in an on-demand mode where individual forecasts can be ordered in an intuative way at a short notice in a web-based interface. The system requires no prior knowledge on behalf of the user regarding atmospheric modeling and/or high performance computing. Specific post-processing modules tailored to end-user needs can be developed, and output from the system can easily be ingested into other decision support software, such as ArcGIS. The system has been actively and successfully used by Search-And-Resuce personel in operations worldwide, but is currently being pushed towards the wind energy sector and other weather dependent operations.
The current research on the system includes development of post-processing modules specific to the needs of the wind energy sector, including filtered and corrected forecasts of wind speed and power and forecasts of atmospheric icing. Forecasts have been successfully improved based on nudging of in-situ observations from an unmanned aerial vehicle and similar work is being done for other observations platform, opening the way for using in-situ, as well as remote sensing, observations to improve operational forecasts of wind inside wind parks.
This work is supported by the Icelandic Tehcnological Development Fund.
Poster title: “Probabilistic forecast information for the renewable energy sector – can we bridge the gap between theory and praxis?
Poster Presenter: Isabel Alberts Research Scientist Deutscher Wetterdienst (DWD) Offenbach Germany
Abstract: The role of renewable energies becomes more and more important in the German energy portfolio. The reasons for this are the political decision to phase out the use of nuclear power by 2022 on the one hand, and on the other hand to reduce the portion of energy production from fossil fuels in order to reach global climate protection goals. By introducing the renewable Energy Sources Act (EEG), the German government is actively promoting the role of renewable energies in the overall energy supply.
With the increasing share of the variable wind and solar energy production, the transmission system operators (TSO) face new challenges concerning the network security and stability. Therefore, high quality weather and power forecasts are indispensable to support their planning, trading and operation activities.
The overall objective of the research project EWeLiNE (Erstellung innovativer Wetter- und Leistungsprognosemodelle für die Netzintegration wetterabhängiger Energieträger) is to improve the NWP and power forecast quality particularly for the power production by wind farms and photovoltaic power plants. Besides improving the deterministic and probabilistic weather and power forecasts themselves, a primary priority of the project is to closely collaborate with TSOs in order to assemble the requirements for the development of customer-oriented weather and power forecast products. To date, only deterministic power forecasts are used in their processes and operation systems. The advantage of probabilistic forecasts is the quantification of the uncertainties associated with the weather and power forecasts. The aim of the presented part of the project is to develop appropriate means and products in order to profit from this additional information.
The EWeLiNE project will be presented with the special emphasis on the collaboration between the model developers and the end users and first ideas on how to integrate probabilistic forecast information into the decision making processes will be shown.
Poster title: Forecasting Wind Energy Production in complex terrain: impact of mesoscale modelling in a forecasting system
Poster presenter: Gianluca Artipoli, Wind Energy Analyst, DEWI GmbH, Italy
Abstract: DEWI has developed a short term power forecasting system based on a combination of physical and statistical methods. The physical model is based on global circulation forecasts, downscaled to mesoscale resolution with the atmospheric research model WRF. A set of artificial neural network (ANN) is then applied to describe statistically the relationship between the wind forecasts and the observed response of a wind farm.
Whistle the importance of both general circulation models and statistical methods have been extensively discussed in details by previous research activities, the effects of mesoscale models as intermediate step between global forecasts and statistical model is still under debate and questioned.
This paper outlines the results of a comparison of two forecasts systems, the first with the application of the mesoscale model WRF and the other with a direct linkage between the global forecasts to the ANN models. The results of the two forecasting systems have been compared with real production data. The observational dataset used for the comparison is derived by several months of operational data collected at four wind farms in Italy, in different climates. The wind farms are sited in complex terrain and affected by strong local atmospheric effects, such as stability and land-sea breezes. The comparison has been focused to assess the ability of the mesoscale model to predict phenomena at a temporal resolution higher than the typical ones of the global forecasts.
Poster title: Wind power forecasts in the business process of a Dutch trading house
Poster presentation: Arno Brand, Research Scientist Wind Energy ECN Petten The Netherlands
Abstract: WindUnie Trading has been using ECN wind power forecasts in their business process for six years. ECN developed and maintains this service on the basis of specific user requirements. In this presentation delegates will learn about the user requirements, product specifications, the match between the two, as well as experiences.
WindUnie Trading is the trading house of Coöperatie WindUnie comprising 250 members with 390 wind turbines distributed over the Netherlands. They provide select cooperation members access to various electricity markets by selling their power to the power exchange APX-ENDEX. Most members operate a single wind turbine with nominal power between 80 kW and 2 MW, and hub heights between 30 and 80 meter.
Each day, before 12:00 h LT on the day before delivery, WindUnie Trading issues a production schedule to the system operator TenneT and the power exchange APX-ENDEX. They create these production schedules by combining information on the availability of the turbines from the members with forecasts from the wind power forecasting method AVDE. Subsequently, they aggregate the forecasts, being made at the level of a single wind turbine or wind farm, to the portfolio level.
Being fed with data from a KNMI-version of the HiRLAM, ECN’s wind power forecasting method AVDE gives the expected value plus a confidence interval of the energy production with time steps of 15 minutes. AVDE uses power curves that have been optimised on the basis of the correlation between the forecasted and the measured production at the level of the grid connections. It has been found to be robust with few errors, and easy to optimise on the basis of measured productions. In addition it has been found easy to extend the portfolio with new sites. For these reasons the forecasting method AVDE satisfies the needs of WindUnie Trading.
Poster title: Wind Power Forecasting For Trading Purposes: The Importance of Managing Wind Volatility
Poster presentation: Marta Soroa Renewable Energy Consultant en Gnarum tecnología Madrid Spain
Abstract: “Energy Trading companies demand different kind of forecasts in order to achieve their business activity. In fact, their benefit highly depends on forecasts accuracy, especially on accurate wind energy forecasting, since wind energy volatility has a big impact on pool price and deviations from actual generations yield to penalizations.
Day-ahead and short-term forecasting are the most widely known services required by traders. These kinds of forecasts are mainly used to operate in different sessions at spot markets. Depending on market rules, the forecast horizon, time resolution and number of possible updates can be completely different. For instance, in Spain, forecasts can be updated in seven different sessions that require different horizons in an hourly time-basis, whereas the German market, it is closer to a continuous market and 15-min time resolution forecasts are required.
In addition, long-term wind forecasts are also highly interesting for traders working on future markets, or any other financial markets, in order to evaluate the risks associated to these long-term investments, since market prices depend on available wind energy production, among other variables.
Under these scenarios, probabilistic forecasting is the most valuable source of information that traders can use in order to evaluate different scenarios and to estimate their probability to occur. Hence, information systems are needed in order to manage all this volume of data in a comprehensive way, which helps to analyze them, taking decisions in almost real-time to maximize the benefits of trading.
Finally, the analysis would be incomplete without a performance evaluation of the forecasts, which also helps to take into account their behavior under given meteorological scenarios where uncertainties might be higher. ”
Poster title: An application of PCA based approach to large area wind power forecast.
Poster presentation: Simone Sperati, Researcher Sustainable, Development and Energy Sources RSE, Italy
Abstract: In recent years, the development of wind energy combined with its large scale integration is creating a growing interest for the predictions of the producible power related to large areas. In particular, Transmission System Operators can use such predictions for different purposes (e.g. managing overloads or reserve estimation). Furthermore the energy market operators need such predictions to forecast unbalancing cost.
The purpose of this work is to predict the power produced by all the wind farms located over the entire area of Sicily island (one of the Italian market regions), whose total installed power is 1746 MW. The study has been conducted for two years long period, considering hourly data of the aggregated wind power output of the island. For each day of the study and for the time horizons from 0 to 72 hours ahead, wind fields have been forecasted using a mesoscale meteorological model (RAMS) and ECMWF determinist forecast fields as boundary conditions. A Principal Component Analysis (PCA) has been applied on the wind speed and wind direction data extracted at 50 m above the ground on the model grid points inside the Sicily territory.
A Neural Network (NN) is then used as post processing technique of the PCA output to obtain the final wind power forecast. The input of the NN, for every forecast lead time, are the final forecast data of wind speed and direction after the PCA application. The NN has been trained with the PCA output and the power measurements on the first year of the analyzed period. For sake of comparison an alternative approach, applying NN directly on RAMS output (without PCA), has been adopted too. The study shows that the PCA introduction leads to better results in terms of RMSE, MAE and BIAS and to a lower computational time.
Poster title:Impact of SRTM and Corine Land Cover data on calculated wind, temperature and precipitation values using WRF
Poster Presentation: Alexander de Meij, Scientific Engineer Sustainable Development Noveltis Labege France
Abstract: The impact of using high resolution SRTM topography and Corine Land Cover data on the calculated meteorological variables wind speed at 10m height, temperature at 2m and precipitation in WRF is evaluated for the Lombardy region (north Italy). We compare the results with the simulation using the standard 30-arc seconds USGS Land Cover and with observations of the ARPA network for the period January – February and July – August 2008.
Our analysis shows that in general calculated average wind speeds are in general lower by WRF with the Corine Land Cover than with 30-arc second USGS, due to the larger fraction of the urban built-up category in the Corine Land Cover and agree better with the observations. Clear differences are found in calculated temperature at 2m height between the two simulations outside the city of Milan for the winter and summer period. R2 values are on average a factor 1.03 and 1.14 higher for the winter and summer period, respectively. Comparing the hit rate statistics of the precipitation events reveals that probability of detection of the precipitation event and the Hansen-Kuipers score are somewhat higher (on average 1%) by the simulation with Corine Land Cover.
Poster title: Operational wind power forecasting systems based on physical and statistical models
Poster presenter: George Galanis, As. Professor Section of Mathematics, Hellenic Naval Academy & Atmospheric Modeling and Weather Forecasting Group, University of Athens University of Athens,Greece
Abstract: “George Kallos, George Galanis, Christos Stathopoulos, Christina Kalogeri Atmospheric Modeling and Weather Forecasting Group,Department of Physics, University of Athens, Greece
The widespread growth of wind power installations and the further integration of the energy yield into the grid increase and reveals the necessity of accurate wind power forecasting. Numerical Weather Prediction systems efficiency in simulating meteorological conditions can be exploited in wind farms for the estimation of power output and the integration to the general grid.
Towards this direction, high resolution modeling is applied, covering onshore and offshore areas in order to compensate the lack of meteorological observations. For applications during the operational stage of a farm, the NWP output is post processed with statistical techniques in order to adapt the local characteristics, remove systematic biases and address major problems associated with ramping effects and extreme events. These techniques include very high resolution modeling with atmospheric systems that take into account the physical and dynamical processes, Kolmogorov-Zurbenko and Kalman filtering and non-linear stochastic wind power prediction models. Implementation of these prediction methodologies in real cases reveals the dependence of forecasting uncertainty on wind variation and previous power production.
Poster title: One-day ahead Power Forecast in China: the Grid curtailment problem
Poster Presentation: Julien Berthaut – Gerentes, Scientific engineer Innovation and Research Meteodyn France
Abstract: In this presentation, a one-day ahead Power Forecasting System is evaluated on several parks in China. The first part describes the system used: a Numerical Weather Prediction (GFS+WRF meso-downscaling) is used in combination with a Computational Fluid Dynamics micro-down scaling. This “physical” forecast is combined with an Artificial Neural Network in order to reduce the final normalized Root Mean Square Error of the system. The presentation describes deeply the ANN, with a special focus on the input variables and the architecture selection.
The second part focuses on a specific aspect of the Chinese market. In this country, the Grid Operator is using the curtailment concept as a unit commitment solution. Thus, the actual production is artificially limited by the grid itself, with no direct links with the meteorological situation. This very specific curtailment has hopefully some signature discussed in this section: some rules can be inferred from the analysis of the statistics of the production signal itself.
The third part demonstrates the ability of the ANN to efficiently predict the global output power, even with a high curtailment level. The opportunity to separate the data set between “free output power” periods and “grid curtailment” periods is explored, as well as a correction tool based on the previous inferred rules. The result is that none of these two refinements brings large benefit: the ANN correction is able to learn the behavior of the curtailed production.
Poster title: Short-Term Forecasting of Categorical Changes in Wind Power with Markov Chain Models
Poster Presentation: Amanda Hering, Assistant Professor Applied Mathematics and Statistics Colorado School of Mines Golden, USA
Abstract: As storage of wind energy is not yet feasible on a large scale, the utility must decide at each balancing time-step whether a change in conventional
energy output is required. With high penetrations of wind energy, utilities must also plan for operating reserves to maintain stability of the electricity system when forecasts for renewable energy are inaccurate. Thus, a simple forecast of whether the wind power will increase, decrease or not change in the next time-step will give utility operators an easy tool
for assessing whether changes need to be made to the current generation mix. In this work, Markov chain models based on the change in power output at up to three locations or lags in time are presented that not only produce such an hourly forecast but also include a measure of the uncertainty of the forecast. Forecasts are greatly improved when knowledge of whether the maximum or minimum wind power is currently being produced and the intrahour trend in wind power are incorporated. These models are trained, tested and evaluated with a uniquely long set of 2 years of 10 min measurements at four meteorological stations in the Pacific Northwest and perform better than a benchmark state-of-the-art wind speed forecasting model. The effect that noise in wind power has on the forecasts is also evaluated.
Poster title: The Value of Improved Wind Power Forecasting in the Western Interconnection
Poster Presentation: Bri-Mathias Hodge Research Engineer Transmission and Grid Integration Group National Renewable Energy Laboratory Golden, CO USA
Wind power forecasting is a necessary and important technology for incorporating wind power into the unit commitment and dispatch process, and is expected to increase in importance as progress is made toward a smarter grid that has higher renewable energy penetration rates. However, while there is a consensus that wind power forecasting can help utility operations with increasing wind power penetration, there is far from a consensus on the economic value of improved forecasts.
In this work, the value of improved wind power forecasting is explored in the Western Interconnection of the United States. Expanding on the Western Wind and Solar Integration Phase 2 Study at NREL, focus is placed on the sensitivities to uncertain wind power forecasts in power systems operation for medium and high-penetration wind scenarios. In particular, uniform and ramp-specific improvement scenarios are considered. A state of the art production cost environment is used to perform the generation and transmission (market) simulation, allowing for the realistic representation of operational dynamics.
The outcome of the research will facilitate a better functional understanding between wind forecasting accuracy and power systems operation. Of particular interest are: 1. correlated behavior among variables (e.g. changes in dispatch stacks, production costs, or generation by type as a function of forecasting accuracy), 2. the relative reduction in wind curtailment with improved forecasting accuracy, and 3. the value of information (e.g. which subset of forecasts provide the most value from a system-wide perspective). Economic savings due to improved unit commitment, re-dispatch, and reserve levels will also be explored.
Poster title: Development of dynamical-statistical short-range probabilistic wind prediction model for wind regimes in coastal and complex terrain
Poster Presentation: Kristian Horvath Head of Applied Modelling Section Department for Research Meteorological and Hydrological Service, Croatia
Abstract: Current state-of-the-art global and regional (mesoscale) models are still limited in representing the challenging wind conditions, such as in coastal and complex terrains. This is due to prevailing wind systems which are – due to their smaller scales – not adequately represented in typical operational weather prediction models. A new dynamical-statistical short-range (up to 3 days) wind forecasting model is being developed to support efficient wind energy integration and wind power plant management.
We present first results of the new short-range wind forecasting system prototype designed for the wind regimes prevailing in coastal and complex terrain. The wind prediction system comprises of the three refinements 1) Regional refinement of global weather model predictions through a regional (mesoscale) weather prediction model; 2) Sub-regional refinement through a sub-regional (sub-mesoscale) simplified CFD-like weather prediction model; 3) Site-specific probabilistic refinement: A self-learning, site-specific probabilistic statistical model, which utilizes both wind predictions and wind tower or nacelle observations.
Study is supported by the EU-IPA (Instrument for Pre-Accession) fund for Croatia, within the framework of the Action “Weather InteLLigence for WIND energy – WILL4WIND”.
Poster title: Statistical Identification of Local and Regional Wind Regimes
Poster presenter: Karen Kazor, Graduate Student Applied Mathematics and Statistics Colorado School of Mines Golden, USA
Abstract: Distinct wind conditions driven by prevailing weather patterns exist in every region around the globe. Knowledge of these conditions can be used to select and place turbines within a wind project, design controls, and build space-time models for wind forecasting. Identifying regimes quantitatively and comparing the performance of different regime identification methods are the goals of this research. The ability of statistical clustering techniques to correctly assign hourly observations to a particular regime and to select the correct number of regimes is studied through simulation. Pressure and the horizontal and vertical wind components are simulated under two different regimes with a first-order Markov-switching vector autoregressive model, and the following five clustering algorithms are applied: (1) classification based on wind direction, (2) k-means, (3) a nonparametric mixture model, and (4,5) a Gaussian mixture model (GMM) with one of two covariance structures. The GMM with an unconstrained covariance matrix has the lowest misclassification rate and the highest proportion of instances in which two regimes are selected. This method is applied to one year of averaged hourly wind data observed at twenty meteorological stations. The lagged wind speed correlations between neighboring sites under upwind and downwind regimes are shown to differ substantially.
Poster title: From Bankable Opportunity to Operations – A New Approach
Poster presenter: Erik Koppen Senior consultant Environment ARCADIS Nederland BV The Netherlands
Abstract: The challenges with wind energy projects are first to provide evidence that the wind resource in a given location is worth an investment (is “bankable”) and second to obtain the most energy possible out of day-to-day operations in a way that is predictable and therefore “saleable”.
”Bankability” has historically been judged by using on-site measurements over at least a year. These measurements can be as simple as a 10-metre high anemometer at the project location converted to the height of the proposed turbine hub using equations representing the atmospheric change in wind speed with height. They can also be as complex as measuring wind speed, direction and density at the proposed hub height using one or a series of tall towers. These measurements can be very expensive and time consuming.
“Saleability” can be improved through weather forecasting for the area of a wind farm but most weather models run on a 15×15 kilometre grid spacing and cannot hope to reproduce local details which have to be corrected for statistically using the historical measurement records.
Since each location (even within a wind farm site) is unique, the above approaches will have errors and uncertainties which can degrade both “bankability” and “saleability” making lenders more reluctant and power purchasers pay less for uncertain power.
A new approach which uses a state-of-the-science weather model running on a very fine scale (1×1 kilometres or less) can overcome both of these issues by providing a location specific data set of 10-minute average values over a period of 5-10 years to improve “bankability” as well as a turbine by turbine power forecast for the next day or two that has proven to be much more accurate than other approaches.
The paper will present the approach as well as validation data to prove the claims.
Poster title: From a Matlab Based Wind Power Forecast to an Integrated EMS Solution
Poster presenter: Stéphanie Lakkis, Project Engineer Energy Management Applications, Siemens AG, Austria
Abstract: “The wind penetration in the total electricity generation has been increasing over years. The European Wind Energy Association reported that by end of 2020, wind energy is targeted to reach 15.7% to 16.5% of Europe’s total electricity demand. Due to this fast growth of volatile energy provision, uncertainties are emerging in the transmission network operation. Facing this problem, TSOs need a reliable tool to monitor and forecast wind power generation within their control zone.
This abstract presents a short-term wind power forecast that was developed for integration in a control center Energy Management System (EMS). It incorporates a state of the art algorithm based on statistical and combined forecasting methods which is already in use at a major European TSO.
The forecast algorithm and the module allowing the combination of external forecasts have been adapted to an embedded system. Several aspects of operational practice are considered by the standardized solution:
- a configurable time grid
- a Graphical User Interface including contour maps
- a SCADA connection
- connections to meteorological services but also to external forecasts
- an after the fact error analysis
Moreover, the integration of a forecast tool within an EMS implies the impact on other standard applications. In order to keep the network frequency on a stable level, Load Frequency Control needs generation profiles based on reliable forecasts of load and wind power. In addition, the forecast results will be taken into account in the Security Analysis for maintaining the energy balance.”
Poster title: Short-term Forecasting of energy production using CFD simulations
Poster presenter: Matteo Mana, Software Developer, Software Development WindSIm, Norway
Abstract: More and more governments require wind farm owners to deliver production forecasts for their wind parks. In Italy, the government requires regular forecasts since beginning of January 2013. To meet these requirements, a forecasting tool has been developed combining wind speed forecasts from mesoscale weather prediction models, Neural Network (NN) corrections, and high resolution CFD simulations. Downscaling mesoscale weather predictions by CFD simulations have proven added value, but there are some bottlenecks. These include the quality of the mesoscale weather forecasts, and the accuracy of the power curve used for the energy calculation. To overcome these problems, a NN correction is used to adjust the forecasted wind speeds before they are used by the CFD. And, another NN correction is used to improve the calculated power output of the CFD simulation. In addition empirical power curves can be used.
Actual production data from sites in Italy are used to train the NN correction and to validate the outcome of the CFD simulations. To correct production data by any kind of algorithm is much more demanding than correcting wind forecasts. The production data needs to be cleaned carefully before use. And, one correction for each individual turbine is necessary to make sure that the down-time of turbines will not lead to wrong production estimates.
Depending on the data quality of the wind farm the presented forecasting system can be run with different complexity. If only mesoscale data is available the system will be based purely on the downscaling with CFD. If measurements onsite are available NN can be trained and if SCADA data of good quality can be delivered the NN strategies can be improved. The added values of the single steps will be discussed comparing the results of the forecasting tool to wind farm production data.
Poster title: RapidWind – Rapidly updated forecasts for wind energy production
Poster presenter: Malte Müller Researcher Research and Development Norwegian Meteorological Institute Norway
Abstract: The main objective of the RapidWind project is to improve wind forecasting along the Norwegian coast by preparing a high-resolution numerical weather prediction (NWP) model with rapid update cycling (3 hours). We use the non-hydrostatic Harmonie-AROME forecasting system which runs operationally at a 2.5km resolution.The rapid update cycling relies on the availability of observations with good spatial coverage at high observation frequencies. Thus, we will include observation types, such as doppler radar winds, automatic weather stations and satellite scatterometer observations to the NWP system. Along with these new data types, optimizations of the data assimilation system are considered. We will present first results of the impact of the model improvements on the wind prediction along the Norwegian coast.
Poster title: On the wind energy production data deficiencies: simulation from statistical models
Poster presenter: Emil Pelikan, Institute of Computer Sciences, Prague, Czech Republic
Abstract: It is a common practice in wind energy production modeling and forecasting to build various statistical models for electrical power produced from wind as if the data were perfect. That is, as if they contained only the variability connected with wind and power production processes.
On the other hand, the real data can contain variability related to various nuisance processes. For instance, the relation between wind speed and power produced can be easily modulated by superimposed wind farm management and maintenance processes.
It is clear that if the available data contain precise information e.g. about the periods in which maintenance of individual turbines was carried out, it is easy to reflect this knowledge when building a statistical model and hence to filter the problem out.
On the other hand, if such information is missing or if it is imprecise (e.g. in terms of timing of intervals when individual turbines are on and off for maintenance), the data are not providing unbiased account of the output curve and other features typically used for prediction purposes. In this study, we use simulations from formalized statistical models to assess how important the problem is and to show the practical consequences.
Poster title: Using Bayesian Model Averaging for wind farm group forecasts
Poster presenter: Pål Preede Revheim PhD, Research Fellow Department of Engineering Sciences University of Agder, Norway
Abstract: For operational purposes predictions of the forecasts of the lumped output of groups of wind farms spread over larger geographic areas will often be of interest. A naive approach is to make forecasts for each individual site and sum them up to get the group forecast. It is however well documented that a better choice is to use a model that also takes the cross-correlation of the forecast errors into consideration, i.e. which is able to take advantage of spatial smoothing effects. It might however be the case that some sites tends to more accurately reflect the total output of the region, either in general or for certain wind directions. It will then be of interest giving these a greater influence over the group forecast.
Bayesian model averaging (BMA) is a statistical post-processing method for producing probabilistic forecasts from ensembles, in this case the single site forecasts. Raftery et al. (2005) show how BMA can be used for statistical post processing of forecast ensembles, producing PDFs of future weather quantities. The BMA predictive PDF of the future wind power production of a group of wind farms is a weighted average of single-farm PDFs, where the weights can be interpreted as posterior probabilities and reflect the single site forecasts’ contribution to overall group forecasting skill over a training period. Single site PDFs of wind power production is modeled as gamma distributions. The parameter estimation is based on forecast-observation pairs from a training period of N days. Different values for N are tested.
The model is tested on 12 and 24 hour forecasts for groups of sites in South-Western Norway. BMA-results are compared to group persistence as well as a model considering spatial smoothing.
Raftery, A.E., et al. (2005). Using Bayesian Model Averaging to Calibrate Forecast Ensembles. Monthly Weather Review, 133, 1155-1174.”
Poster title:Evaluation of NWP Resolution Effect on Wind Speed Forecast Quality for a Wind Farm in Central Sweden
Poster presenter Martin Haubjerg Rosgaard, ENFOR A/S and the Technical University of Denmark Wind Energy Department
Abstract: For any energy system relying on wind power, accurate forecasts of wind fluctuations are essential for efficient operation. Increased forecast precision allows end-users to plan ahead within narrowing uncertainty tolerances which in turn strengthens the feasibility of wind energy. This study aims to quantify and explore various aspects of value added to wind power forecasts by downscaling global numerical weather prediction (NWP) data using a mesoscale NWP model. Farm-averaged nacelle wind speed time series forms the basis of comparison for four daily 48-hour WRF forecasts at 30km, 10km, 3.3km and 1.1km spatial resolutions for the yearlong period April 2012 to April 2013.
The wind farm is situated in central Sweden and consists of 40 Vestas V 90 turbines with a tower height of 95 meters. The installed electricity capacity is 78MWe and annually ~240 GWh are produced.
Preliminary results presented include investigation of NWP variability representation at different numerical scales, correlation between nacelle wind speed measurements and corresponding forecasted values conditioned on three 12-hour lead time horizons, and the effect of smoothing on forecast quality for the four spatial resolutions.
Poster title: FROM ENSEMBLES TO PROBABILISTIC WIND POWER FORECASTS – HOW CRUCIAL IS THE ENSEMBLE SIZE?
Poster Presentation: Jeremy Sack, Meteorologist R&D, EWC Weather Consult GmbH, Germany
Abstract: Probabilistic forecasting has become state-of-the-art in wind energy prediction. Several studies have shown that probabilistic forecasts not only provide more extensive information about future weather development, but also allow players on the wind energy market such as grid operators, risk managers or energy traders to save valuable resources.
The use of ensemble Numerical Weather Predictions (NWP) has turned out to be an elegant way of representing the forecast uncertainties depending on prevailing weather situations. It is now common practice to derive predictive densities from the ensemble’s set of trajectories and to maximize the skill through a calibration with measurements. Although expensive in terms of computing resources and money, ensembles including a large set of members have been preferred in recent studies.
The present study introduces a way of generating probabilistic forecasts obtained from a small multi-model-ensemble offering similar skills as a large single-model ensemble (ECMWF_EPS). This is achieved by an adaptive calibration of (u,v)-wind ensemble forecasts (Pinson, 2012) followed by a conditional weighted combination of wind power forecasts (Thordason, et al. 2008). The latter makes use of each member of the multi-model ensemble having distinct characteristics which is not the case for ECMWF_EPS.
Depending on lead time and location, this low-cost method yields comparable or even better results than those originating from ECMWF_EPS with respect to the continuous rank probability skill score. The question of how crucial expensive large single-model ensembles are in an operational wind power forecasting environment is discussed and a method for choosing the best option is proposed.
Poster title:High-resolution surface modelling for extended-range downscaling of near-surface atmospheric fields over Canada
Poster presenter: Leo Separovic, Physical Scientist, Meteorological Research Division Environment Canada, Canada
Abstract: In order to facilitate a larger-scale integration of wind energy within the power grids, the Government of Canada has commissioned a project carried out by Environment Canada (EC) to generate high-resolution multi-year database of near-surface wind over the entire country. The database will be generated with the limited-area Global Environmental Multi-scale (GEM-LAM) model. The final results are required to have a horizontal grid spacing of 2 km and frequency of 10 min.
The approach consists of performing a single continuous multi-year GEM-LAM simulation over a continental-scale mesh, as blending of results obtained from multiple simulations conducted over smaller domains and time frames may lead to significant spatiotemporal discontinuities. However, due to the accumulation of errors, simulations conducted over large domains and extended time frames are prone to large drifts of atmospheric and surface variables. The spectral nudging of upper-level atmospheric fields is applied to the model vertical levels in order to prevent deviations of the large-scale components from the driving analysis fields. In addition, to prevent drifts in land-surface fields, such as soil moisture and snow conditions, grid nudging of relevant surface variables towards a high-resolution surface dataset is devised. The surface dataset is generated with a high-resolution external Surface Prediction System (SPS) that simulates only land and near-surface physical processes. The atmospheric forcing for SPS simulations is derived by blending information obtained from EC’s operational regional weather forecasts, Canadian Precipitation Analysis (CAPA), observed near-surface temperatures and humidity, as well as from regional analysis of soil moisture.
This presentation will discuss the surface modelling strategies devised for this project in detail and their implications on the quality of the near-surface time series of wind and other meteorological fields, based on the validation of preliminary results of SPS and GEM-LAM simulations.
Poster title: EWeLiNE: Development of innovative weather and power forecast models for the grid integration of weather dependent energy sources
Poster presenter: Scott Otterson, postdoctoral research fellow, Energy Economy and Grid Operation Fraunhofer IWES, Germany
Abstract:Wind power forecasts have already been improved by intelligent combination of numerical weather prediction (NWP) ensembles and by assimilating online measurements at forecasted and distant locations.
The next step in wind but also in solar (PV) power forecasting research is seen in the establishment of close links between meteorology and energy economy. In addition to the combined and parallel enhancement of existing but also of new weather and power forecast tools, the integration of information related to energy management strategies into the weather model calculations shows a big room for improvement. This step will be done by the Fraunhofer IWES and the German Weather Service (DWD) within the German research project “EWeLiNE” lasting from 2012 to 2016. The project is funded by the Federal Ministry for the Environment, Nature Conservation and Nuclear Safety (BMU) and supported by the German TSOs Amprion GmbH, TenneT TSO GmbH and 50Hertz Transmission GmbH.
The project pursues the following aspects: (i) development and documentation of forecast users special requirements regarding the future grid and market integration of wind and solar energy; (ii) combined optimization of deterministic and probabilistic weather, wind power and PV forecasts; (iii) assimilation of wind power and PV measurements into the NWP models of DWD; (iv) weather regime classification; (v) application-specific NWP ensemble generation using the COSMO-DE-EPS model chain and several calibration methods; (vi) Multi-Model ensemble performance and characteristics; (vii) development of an R&D forecast platform and standardization; (viii) demonstration phase.
The paper presents the research project “EWeLiNE” with focus on the optimization potential by connecting meteorology with energy economy, the planned working steps and first results. The latter are mainly based on a detailed investigation of the current but also of the future forecast user’s requirements which is based on the evaluation of a purpose-built questionnaire answered by the German TSOs.
Poster title: The performance evaluation of WRF Model for Short Term Wind Energy Prediction System for Turkey
Poster presenter: Elcin Tan Instructor Dr. Meteorological Engineering Istanbul Technical University Turkey
Abstract: Therefore, the main purpose of this study is to evaluate the performance of WRF model in wind speed and direction prediction for 4 different wind farms located in the west part of Turkey and to compare the results of WRF model with both observations and the results of CFD based models. The WRF domains for these wind farms are constructed as 3 nested domains with the horizontal resolutions starting from 9 km by ratio 3. Each simulation has 72hr time horizon and they are performed for one-year period for each domain area. Although, according to the control run studies, WRF model has better performance than other high-resolution CFD based or physical models for average wind speeds, this study indicate that the performance of WRF model in estimating extreme winds is still evaluated as poor. So our main conclusion is that the WRF Model performance depends on the variation of wind speed accordance with domain choice. Our results also indicate that for inner WRF domain (1 km), the errors are increasing with time horizon of the model. That is, 24hr simulations have fewer errors than 48hr simulations and also 72hr simulations. The second domain solutions, with a coarser resolution (3 km), have better performance than the finest resolution domain for the extreme wind cases.
Poster title: “A Short Term Wind Energy Prediction System for Turkey
Poster presenter: Elcin Tan Instructor Dr. Meteorological Engineering Istanbul Technical University Turkey
Abstract: World energy demand is growing rapidly with developing industries and technology. Therefore, many countries today, have started to obtain benefits from renewable energy sources. For power generation, despite the discontinuous nature of wind energy, it is widely used in the world especially in the European Union countries. Moreover, wind power production forecasts are needed due to the fact that wind energy has an increasing share in total energy production and wind power generation systems has lack of energy storage.
Within the scope of this project, a short-term (0-72 h) wind energy prediction system (SWEPS) is developed by using four modules for four locations of Turkey, namely, Çanakkale, İstanbul, Balıkesir, and Manisa. These modules are consisted of mesoscale numerical weather prediction by using the Weather Research and Forecast (WRF) model, which is used to increase the resolution of wind energy field estimates of a global atmospheric model; micro scale wind energy prediction by using three diagnostic models, namely, WAsP, WindSIM, and AES RuzgarSIM; model output statistics by using artificial neural network method; and error analyses of modeled versus observation wind field and wind energy values. For four different wind farm locations, WRF model is run to produce 72 h wind field forecast for one-year period of 2010 with 1 km resolution by coupling 3 diagnostic models, the best of which has been chosen to include in SWEPS. After applying artificial neural networks, in order to determine the site-specific SWEPS confıguration, the overall performances of these models are tested. As a general result, the normalized RMSE is reduced to 20% for each wind farm location.
Poster title: A Quantitative Assessment of the Variability of Wind
Poster presentation: Darion Turner, Sales Trader, Trading SmartestEnergy Ltd, UK
Abstract: Using real, historic half-hourly wind generation data for an existing portfolio of wind projects across financial year 2012/13, the frequency and magnitude of error measures derived from a variety of persistence forecast bases and their associated costs at imbalance versus market value were assessed against expected values and against individual performance indicators such as capacity factor, various values of capacity credit (as derived from different definitions) and seasonal/time-of-day fractional generation output, highlighting key relationships. Total contracted value increases with total volume and capacity factor. Net imbalance value negatively correlates with capacity factor. Average contracted value is not dependent upon the total volume generated, but rather the time at which it is generated. The average net income is beneficiary of the above stated relationships – error magnitude and frequency are key suppressors of average net income and output delivery time and total volume key inflators of net income.
Poster title: Applying Model Output Statistics (MOS) in the new German Project EWeLiNE for enhanced wind forecasting for renewable power generation
Poster presentation: Gernot Vogt, Scientific Researcher Development of Meteorological Applications, Germany
Abstract: “Model Output Statistics (MOS) is a powerful tool for optimizing the direct output of numerical weather forecast models. By developing multiple linear regressions with predictors, derived from observations and model output at DWD (German Meteorological Service) a reduction of 50% of the error variance in the forecast has been achieved. Moreover, statistical post-processing yields numerous advantages in forecasting, e.g. down-scaling on point forecasts at observation stations with specific topographic and climatological characteristics, correction of biases and systematical errors produced by numerical models, the derivation of further predictands of interest (e.g. probabilities) and the combination of several models.
In the recently founded German project EWeLiNE (Simultaneous improvement of weather and power forecasts for the grid integration of renewable energies), which is fulfilled in collaboration of DWD and IWES (Institute on Wind and Energy Systems), one of the main goals is an adjustment of the DWD-system MOSMIX (combining the global models IFS and GME) to the demands of the transmission system operators. This includes the implementation of new predictands like wind elements in altitudes > 10m.
After the processes of converting raw data of acquired point measurements of observation masts and the implementation of their data into the MOS algorithms, studies have been accomplished investigating the fit of forecasts to observations by means of wind speed at 30m and 100m at selected locations, the choice of the predictors and the weighting of employed weather forecast models. By the implementation and pre-processing of the measured data, e.g. changing the length of training periods and the vertical interpolation of wind speed in heights of absent measurements, uncertainties develop, which require sensitivity studies, as the accuracy of the statistical forecast is affected. Amongst others these studies have been conducted by assessments of the RMSE.