Offshore wind farms are exposed to a dynamic environment where the interactions between wind and waves modify wind resources. By extracting energy from the flow, wind farms alter these interactions and traditional wind resource assessment usually simplifies or ignores the interactions between wind and waves. However, in recent years some progress has been made in properly considering these interactions. In this presentation, we will highlight the relevance of these wind-wave-wake interactions for wind energy applications by summarizing the findings of two studies.

First, we show which role these interactions play in tropical cyclone development. For that, we use Typhoon Megi, a category 5 cyclone that hit Taiwan in late September 2016, as an example. Second, we demonstrate the relevance of wind-wave-wake interactions for resource assessment. We use a statistical-dynamical downscaling approach to represent the 30-year climate in the German Bight and assess the impact of waves and wakes on wind resources (Fischereit et al. 2022).

For both studies, we use the Coupled-Ocean-Atmosphere-Wave-Sediment Transport Modeling System (COAWST; Warner et al. 2008, 2010). In this modelling system, we activate the atmospheric model WRF (Weather, Research and Forecasting model) and the wave model SWAN (Simulating WAves Nearshore model) . We apply the Wave Boundary Layer Model (WBLM; Du et al. 2017, 2019) to ensure that the exchange of flux and energy between these two models are consistent. To derive the range of possible wind farm effects, we use two different wind farm parameterizations (WFP) in the simulation, namely the WFP by Fitch et al. (2012) and the Explicit Wake Parameterization (EWP) by Volker et al. (2015).

The studies show that wind-wave-wake interactions influence wind resources, and that they also affect the intensity and track of a typhoon. This highlights that these interactions should be considered in wind energy applications.

How to cite: Fischereit, J., Imberger, M., Müller, S., Hahmann, A. N., and Larsén, X. G.: Wind-wave-wake interactions in offshore wind farms, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-158, https://doi.org/10.5194/ems2023-158, 2023.

Scanning lidars measure a wind field over a large area with a higher resolution than numerous anemometer installations would allow. The drawback is that a scanning lidar measures only the radial velocity – the projection of the actual velocity to the lidar's line of sight. Reconstructing the original wind field requires a retrieval procedure. Most of the retrieval algorithms were developed for homogeneous wind fields. While those algorithms estimate the free flow well, they do not preserve non-homogeneous structures such as wind turbine wakes.  
  
A recently developed 2D-VAR algorithm [1]  utilizes the property of conventional retrieval methods to estimate the homogeneous flow and uses it as an input to the cost-optimization function. The cost-optimization retrieves the wake structures more accurately. While the solution for a homogeneous flow is stable, the result of the cost optimization is affected by the initial guess field, e.g., a uniform wind field based on the FINO1 or SCADA time series.  
 
We perform a sensitivity analysis of the 2D VAR retrieval algorithm for lidar scans of the Alpha Ventus wind farm taken from the FINO1 platform during the OBLEX-F1 campaign. Partially covering September 2016, the dataset provides a range of wind speeds and directions to test the algorithm's performance under various conditions. The algorithm's sensitivity is evaluated for different initial guess fields.
 
The retrieval algorithm is evaluated based on the agreement with the reference data: FINO1 cup anemometer or SCADA time series. The retrieval accuracy is described by comparing the radial velocity calculated from the retrieved field to the original lidar scan. The results show that the retrieved field near the wind turbine tends to agree with the SCADA series better than with the FINO1 series. The agreement to the SCADA data is the best when the initial guess is based on the corresponding turbine series. The radial velocity residuals are slightly biased and are primarily localized in the near wake. Whether it is a positive or negative bias depends on the intensity of the wakes.

How to cite: Krutova, M., Bakhoday-Paskyabi, M., and Reuder, J.: Sensitivity analysis of the 2D VAR retrieval method in the application to the wind turbine wakes, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-242, https://doi.org/10.5194/ems2023-242, 2023.

Wind energy is one of the key aspects for renewable resources that contributes to climate change mitigation policies in national and international energy transition strategies. The availability of wind resources itself is also affected by climate change, due to possible expected changes in large-scale circulation patterns, with consequent impacts on wind regimes on the Italian territory. This study focuses on understanding whether, how and to what extent climate change can affect wind producibility in Italy, using the Euro-CORDEX regional climate models. Because wind turbines work within a defined wind speed range (within the cut-in and cut-off wind speed), a better estimate of wind speed values is able to significantly impact the estimation of wind producibility. For this reason, a bias correction is performed using the MERIDA meteorological reanalysis, for the 10 m wind speed variable of the climate models. The calculation of the producibility was obtained assuming a reference wind turbine that is widely used in the Italian wind farms (VESTAS V112 - 3000 kW). The study also analyzes the changes in the wind energy production compared to the reference period 1986-2005 for the short (2021-2050), medium (2051-2080) and long-term (2071-2100) scenarios, according to the RCP 4.5 and RCP 8.5 scenarios. Together with the variation of the climate signal of the ensemble mean, an uncertainty analysis is also performed to evaluate the reliability of the climate signal itself. If, on one hand, the results show a prevalently weak and not statistically significant climate signal for the RCP 4.5 scenario, on the other hand, a more pronounced and significant signal is highlighted for the RCP 8.5 scenario in the medium and long term, indicating a decrease in wind producibility. More specifically, the conclusions suggest that the future planning of wind producibility should mainly be targeted toward some specific areas of the eastern Italian coast and in the south-east Italian regions, mostly in the off-shore areas. In these regions, indeed, the RCP 8.5 scenario shows the lowest decrease in the overall annual producibility with respect to other Italian regions, while, for the RCP 4.5 scenario, the medium and the long term foresee a slight increase in wind producibility at the annual level.

How to cite: Bonanno, R. and Viterbo, F.: Climate change impacts on wind energy production for the Italian peninsula, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-92, https://doi.org/10.5194/ems2023-92, 2023.

How to cite: Tian, Y., Yang, X., Schicker, I., and Jung, A.: Wind to start the dishwasher? High-Resolution Wind Atlas for Finland, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-380, https://doi.org/10.5194/ems2023-380, 2023.

Europe's goal of becoming the first climate-neutral continent by 2050 implies increased investment in renewable energy sources at both continental and national level. The offshore wind is a rapidly maturing renewable energy technology that is poised to play an important role in future energy systems, as it is at the core of the European Green Deal. In this context, Spain and Portugal governments are working on legislation to facilitate the market in floating offshore wind energy. Most studies over the Iberian Peninsula (IP) are focused on the analysis of temporal and spatial changes of the wind resource, showing that some areas such as the western IP show high potential. However, there are important spatial planning conflicts for the deployment of floating offshore wind towers, related to wind infrastructure technologies and legislative limits. Moreover, the variable nature of wind power poses challenges for its use in the national electricity generation system. Increasing interconnections between regions of the IP can smooth the variability of local wind generation by exploiting their spatial complementarity.  In this work, we present an analysis of the geographical combination of potential floating offshore wind farms sites over the IP. Hourly wind speed at 105 and 150 meters hub heights (typical of present and future offshore wind installations, respectively) were extracted from the very high resolution (0.055º) COSMO-REA6 reanalysis for 1995-2018. Then, wind speed was translated into capacity factor using an adequate power curve for each hub height. In order to assess the spatial complementarity of potential Iberian wind offshore farms, around 15 gross locations were chosen based on the publicly available planning information given by the Spanish Maritime Spatial Plan and the Direção-Geral de Recursos Naturais, Segurança e Serviços Marítimos of the Portuguese government. Then, the coefficient of variation (CV) was calculated for each site, and the geographic aggregation of sites that minimizes the CV of the aggregated wind capacity factor was analysed considering annual and seasonal time scales. First results show that, at both hub heights, as more distant sites are added, the coefficient of variation decreases (~ 40%) more than the capacity factor (~ 15%).  This behaviour varies slightly by season, with the hourly variation decreasing the most in winter (~45%) and the capacity factor mean decreasing the most in summer (~ 24%). This ongoing analysis indicates that it is clearly more advantageous for the Iberian electricity system to build farms far apart than to concentrate wind farms in one or two highly productive areas. It also shows that a larger and more stable offshore wind resource can be obtained at higher heights. Thus, the capacity factor is around 11% larger and with less hourly and interannual variation (~12%) at 150 meters than at 105 meters.

How to cite: López-Franca, N., Sanchez, E., Gallardo, C., Gaertner, M. A., Molina, M. O., Ortega, M., and Gutiérrez, C.: Complementarity of potential Iberian offshore wind farms in allowed locations based on COSMO-REA6 high-resolution reanalysis, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-260, https://doi.org/10.5194/ems2023-260, 2023.

This study presents a high-resolution simulation of a low-level jet (LLJ) with explicit representation of wind turbines using the Weather Research and Forecast (WRF) model and the newly-developed Simple Actuator Disk model for Large Eddy Simulation (SADLES). LLJs are low-level wind maxima that can impact the power output and structural loads of turbines. Therefore, accurate prediction of the LLJ is essential for wind energy planning and management. The meso-to-micro downscaling approach is used to capture the LLJ dynamics, which involves nested domains with increasing spatial resolution from the mesoscale to the microscale. The WRF model is configured with five nested domains: mesoscale domains with grid spacing of 9 km, 3 km, and 1 km to capture the general dynamics structure of the LLJ, and two micro domains with grid spacing of 200 m and 40 m with large eddy simulation (LES) configuration. The SADLES model is used in the 40-m domain to enable the interaction between the LLJ and the wind turbines of the Alpha Ventus wind farm located in the North Sea. The simulation results are verified against observations at the FINO1 mast stations, including cup anemometers at various heights and LiDAR wind profiles. The WRF-SADLES system captures the LLJ dynamics and the interactions of turbines on the LLJ. The LLJ is well-reproduced in terms of its intensity, depth, and spatial extent, and the wakes of individual turbines are explicitly captured, significantly impacting the LLJ. The study provides insights into the LLJ dynamics and turbine-wake interactions, which can inform wind energy planning and management.

How to cite: Bui, H. and Bakhoday-Paskyabi, M.: Simulating wind turbine interactions with low-level jet using a meso-to-micro downscaling approach, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-455, https://doi.org/10.5194/ems2023-455, 2023.

The share of wind energy sources in Finland was reasonably expanded over last years. It is foreseen that wind power generation will exceed 30 % of the total electricity production in Finland by 2030. The growing share of wind resources, however, brings more variability to the electricity grid raising the risk of imbalances between electricity production and consumption. The larger wind share also has stronger impact on electricity prices in Finland that are set daily at the Nordic energy trading stock Nordpool. To estimate wind power generation for coming days a probabilistic wind power forecasting system has been developed at the Finnish Meteorological Institute (FMI).

The system is based on the Meteorological Ensemble Prediction System (MEPS) HARMONIE, running operationally by the group of Nordic countries under MetCoOp cooperation project. Information on wind farms is provided by the Finnish Wind Power Association (FWPA). FWPA data essentially enables wind calculations for every Wind Turbine (WT) location. MEPS-based wind speed is interpolated to the WT hub heights at their locations and further corrected by applying a wake propagation model. The key advantages of the method are 1) the method exempts from the necessity of calculating wake interaction between adjacent wind farms, 2) wind direction variations within large-scale wind farms are considered.

The method requires both power curve and thrust coefficient (C_T) curve of installed WT models for calculating respectively power and wake-related losses. Power curves for new WT models are typically not available in open databases, whereas C_T curves are generally not available for most installed WT models. A statistical solution was developed at FMI to approximate both power and C_T curves. The method well approximates power and C_T curves of most of the models installed by 2022. New WT models of 4 MW or larger capacity are often equipped with a system that reduces power generation at wind speeds exceeding c.a. 15-18 m/s towards cut-off to prevent abrupt shutdown of the rotor. This feature of WTs was addressed in approximating both power and C_T curves, with preliminary results demonstrating reasonable improvement, specifically for C_T curves.

The FMI power forecasting system is somewhat struggling to properly include all new WT installations, because of the rapidly expanding installed capacity in Finland.  Correction of the forecast is done using data on actual aggregated power generation and installed capacity provided by Fingrid, the Finnish Transmission System Operator. The overall agreement between forecasted and actual wind power production is satisfactory. However, a detailed analysis reveals some seasonal and diurnal behavior, which can be approximated as a function of month, day’s hour and forecast lead time. Power losses associated with downtime and icing can be accounted for by applying short-range correction.

How to cite: Atlaskin, E. and Lindfors, A.: Aspects of wind power forecasting over the Finnish wind fleet considered as a single wind farm, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-287, https://doi.org/10.5194/ems2023-287, 2023.

In the last few years there has been a significant increase in Belgian offshore wind energy production, with a 2.26 GW total installed capacity completed at the end of 2020. Storm events over the North Sea can impact many of these wind farms at roughly the same time, because they are situated relatively close together in a narrow zone in the North Sea. Each wind turbine has a characteristic cut-out speed, above which they will decrease production rapidly or shut down as a protection measure. In case of a major storm, many wind turbines can shut down simultaneously, which can lead to large imbalance risks in the electricity grid. To better understand and predict such events, the Royal Meteorological Institute of Belgium (RMI) was involved in the development of a dedicated storm forecast tool for Elia, the Belgian transmission system operator for high-voltage electricity. The aim is to forecast large storm events, several days up to seven days ahead, and associated cut-out events, a day ahead and up to two days, making use of weather models that generate wind speed forecasts at turbine height and location. Due to the uncertainty in the precise location, timing and intensity of a forecasted storm, and the fact that cut-out events are sensitive to whether or not a high wind speed threshold is exceeded, a probabilistic forecast approach was taken. Moreover, Elia also required high temporal resolution forecasts (output every 15 minutes), so that a combination of a high resolution deterministic model and lower resolution ensemble weather prediction model was used. This allows both detailed forecasts and a good estimation of the uncertainty in the forecasts, thereby helping end users in their decision making process. The storm forecast tool developed at the RMI makes use of the deterministic ALARO model (4 km resolution) combined with the ENS ensemble forecasts (18 km resolution) of the European Centre for Medium Range Weather Forecasting (ECMWF). The storm forecast tool has been operational at the RMI since November 2018. We give an overview of the current status of the storm forecast tool, together with its performance over the past years, and present some ongoing and planned future developments. These include ensemble calibration with historical wind speed measurements, the inclusion of wake effects using fast wake models and NWP wind farm parameterizations, and the prediction of ramping events.

How to cite: Smet, G., Van den Bergh, J., and Termonia, P.: Probabilistic storm forecasts for wind farms in the North Sea, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-505, https://doi.org/10.5194/ems2023-505, 2023.

Multiscale modeling is essential in atmospheric science as it allows for a better understanding of atmospheric processes at different scales, more accurate predictions of weather events, and specifically, it helps scientists and engineers in offshore wind to understand the complex flow physics involved in wind turbine operation. At the microscale, multiscale modeling for offshore wind applications may be based on Large Eddy Simulation of wind, which helps predict turbulent flow and mixing at small scales, and study the aerodynamic behavior and load characteristics of wind turbines. Simulations at small scales are further useful for optimizing the design of wind turbines for improved performance and efficient operations. At the mesoscale, multiscale modeling such as the Weather Research and Forecasting (WRF) provides capability to simulate a range of atmospheric phenomena from global climate patterns to regional weather episodes at coarse resolution. Since WRF at these coarse resolutions is not able to explicitly resolve the small-scale turbulence, the model-chain containing the WRF and LES combines the strengths of both models to resolve from several kilometers to several tens of centimeters.  

This study presents our recent development on offline nesting of the Weather Research and Forecasting (WRF) model with the Parallelized Large Eddy Simulation (PALM) model to improve simulation resolution in the area of Alpha Ventus offshore wind park, while maintaining reasonable computational costs. The WRF model simulates the entire domain covering the Southern North Sea, while the higher resolution PALM model is used to simulate the finest WRF nested domain within the area of interest. We use then the LES simulation results to investigate the wake dynamics behind the wind turbines and provide insights into the loads of downstream turbines. It is noted that we focus on the LES simulations of an Open Cellular Convection (OCC) event and the study of impacts of this transition event on the dynamics of wind, atmospheric turbulence in modulating wakes, loads, and in general flow patterns in the area of Alpha Ventus offshore wind park (OCCs can cause strong variations in wind speed and direction leading to significant modulation of the aerodynamic behavior of wind turbines and their power outputs).

How to cite: Bakhoday-Paskyabi, M., Bui, H., Mohammadpour Penchah, M., Krutova, M., and Ning, X.: Multiscale modeling of wind during an OCC event over an offshore wind park: Implication on wake and loads, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-547, https://doi.org/10.5194/ems2023-547, 2023.

The Copernicus European Regional ReAnalysis (CERRA) is a new limited-area reanalysis covering Europe at 5.5 km resolution, produced in the framework of the Copernicus Climate Change Service (C3S) and available in its Climate Data Store (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-cerra-height-levels).

Near-surface wind speeds are essential variables to simulate wind power production for wind energy applications or prospective studies of the European power system. The purpose of this study is to assess the quality of CERRA’s wind speeds (especially at the height level of 100 m) in respect to these studies and in comparison to other reanalyses such as ERA5.  

A first part of the study was to determine which forecast steps to use. CERRA provides an analysis every three hours and hourly forecasts up to 6 hours, therefore forecast streams are overlapping. For example, values at 04 UTC are provided by the step +4h of the 00 UTC analysis and also by the step +1h of the 03 UTC analysis. CERRA documentation leaves the choice to the user. For wind speeds, we recommend not to use the forecast steps close to the analysis as they exhibit too low values. This phenomenon was already observed in AROME NWP model (Jourdier, 2020). This wind deficit is more pronounced in summer, in Northern Europe and offshore. Therefore we chose to use steps +4 to +6h of each analysis.

In a second part, average 100 m wind speeds from CERRA were compared to other reanalyses:

• ERA5, ECMWF’s global 32-km reanalysis, which provides the boundary conditions to the CERRA system.
• COSMO-REA6, DWD’s European reanalysis which has a similar horizontal resolution but is based on older ERA-Interim global reanalysis.

Compared to ERA5, CERRA exhibits large differences (up to 6 m/s in average), much higher wind speeds over mountains (especially over the Scandinavian Mountains, the Alps and Dinaric Alps) and higher wind speeds in general over most of Europe either offshore or onshore (except Sweden and Finland). This was expected as ERA5 winds are known to be too low, especially in mountainous areas.

Compared to COSMO-REA6, CERRA exhibits smaller differences (up to 3 m/s). CERRA’s wind speeds are higher over Eastern Europe and much higher over some mountains (parts of the Scandinavian Mountains, the Alps). They are lower over the British Isles, most of Scandinavia and most parts of the Mediterranean Sea.

Finally, hourly wind power production series were simulated based on CERRA for all wind farms in France. Compared to observed power, CERRA shows less bias than ERA5 and similar hourly correlations. Still some bias remain, mainly positive in Western France and negative in Eastern and Southern France.

How to cite: Jourdier, B., Diaz, C., and Dubus, L.: Evaluation of CERRA for wind energy applications, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-311, https://doi.org/10.5194/ems2023-311, 2023.

To mitigate and adapt to the ongoing climate change, the decarbonization of the economy requires a radical change in our energy production and consumption patterns. A robust finding of existing studies is that renewable energy sources - and more specifically wind and solar power generation - are expected to represent a major share of the power mix in the future. Due to their higher dependency on meteorological variable a better understanding of the variability of VRE outputs across different temporal and geographical scales thus becomes critical. Of particular interest is the study of tight grid conditions with periods of coincidence between low wind and solar output because they will condition flexibility requirements.

Our work addresses these research questions based on an analysis of the variability of solar PV and wind power generation capacity factors provided by the Copernicus service C3S energy. The Copernicus C3S energy service deals with the transformation of climate variables (reanalysis, seasonal forecast and climate projection) into energy variables (wind, PV, hydro and energy demand principally). To take into consideration the dependance between the solar and wind resource the combined variability has been evaluated for different wind to PV share. In addition, the coupling of different kind of storage technologies has been anticipated by evaluating the variability on different time scales ranging from hour to month. Finally, focusing on extreme conditions, the results of the variability analysis was synthesized in a couple of boundary conditions that capture the impact of the meteorological variability on estimated joint wind and solar power output. The added value of the proposed methodology will be illustrated for a simple case in France.

How to cite: Saint-Drenan, Y.-M., Assoumou, E., and Haykal, R.: Integrating meteorological dependent variability of wind and solar power for resilient power systems, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-490, https://doi.org/10.5194/ems2023-490, 2023.

Solar radiation forecasting is essential for the efficient operation of solar energy plants. Accurate forecasting models can help optimize solar energy use, reduce costs, and minimize environmental impacts. The clear-sky index (CSI) is a common parameter used in forecasting models to estimate the amount of solar radiation reaching the Earth's surface. CSI is the ratio between surface solar radiation (SSR) and surface solar radiation in a clear sky situation (SSR_cs). The temporal variability of CSI is scale-dependent, meaning that the evolution of CSI is less predictable at smaller spatial scales than at larger scales. This variability can strongly impact the accuracy and quality of solar radiation forecasts.

Our study aims to investigate the temporal variability of CSI at different spatial scales (from 0.02˚ to 2˚) and its possible application to SSR forecasting models. The study uses cloudiness maps estimated from Meteosat SEVIRI Level 1.5 data by the HelioMont algorithm (Castelli et al., 2014) with a spatial resolution of 0.02˚ and a temporal resolution of 15 minutes over Switzerland. To quantify the temporal variability of CSI as a function of spatial scale, we draw on methods used by Venugopal et al., 1999 and Pulkkinen et al., 2019 for precipitation analysis.

We show that the temporal variability of CSI is indeed scale-dependent, with smaller spatial scales (down to ~0.02˚) exhibiting higher temporal variability than larger scales (up to ~2˚). We also show that Fourier decomposing the CSI field in space can help to track the different temporal evolutions for the different spatial scales. This finding is particularly important for the development of more accurate forecasting models, as it suggests that different models (or at least scale-dependent parameter values) may be required for different spatial scales.

In conclusion, this study provides valuable insights into the temporal variability of CSI at different spatial scales providing experimental correlation between the spatial scale and the variability of CSI's temporal evolution. The findings of this study have important implications for the development of more accurate forecasting models, which can ultimately contribute to the increased use of solar energy sources.

How to cite: Carpentieri, A., Folini, D., Wild, M., and Meyer, A.: Scale-dependent Temporal Variability of the Clear-Sky Index and its Relevance for Solar Radiation Forecasts, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-81, https://doi.org/10.5194/ems2023-81, 2023.

The increase in global deployment of PV technology and the estimation of its continuous growth for the near future requires an accurate solar resource assessment for the industry and the different stakeholders. While historically that assesment  was made with satellite or in-situ observations,  studies using climate models have shown that climate projections are necessary to account for climate variability in the near or mid-future. In recent years,  many studies have been conducted analyzing climate projections of solar resource, utilizing either global and/or regional climate models, considering different time horizons and emission scenarios worldwide. Some studies have used a single regional climate model for a specific area, while others have adopted a multi-model approach, incorporating simulations from coordinated experiments like Coupled Model Intercomparison Project (CMIP) for globals or Coordinated Regional Downscalling Experiment (CORDEX) for regionals. In certain cases, disparities between global and regional model projections have been identified, with corresponding analyses published.

The objective of this study is to conduct a comprehensive review of solar resource projections worldwide. The analysis will be based on six distinct regions, and the results will summarize the projected changes in solar resource and the underlying reasons. Additionally, the agreement or discrepancy between global and regional models, as well as the main sources of uncertainty reported in the literature for each region, will be examined. This work aims to synthesize the most up-to-date knowledge on solar resource assessment using climate models , in order to help the public sector and industrial experts working to incorporate climate impacts into energy sector decision-making processes around the world.

How to cite: Gutiérrez, C., Somot, S., Álvarez-García, F. J., and Cabos, W.: Impact of Climate Change on solar resource: a review of projections, uncertainties and perspectives, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-221, https://doi.org/10.5194/ems2023-221, 2023.

The shift towards renewable energy sources such as wind and solar poses major challenges for the entire energy sector. In order to ensure a secure energy supply in the future, accurate weather forecasts are becoming increasingly important. The majority of the operational numerical weather prediction models do not consider prognostic aerosol and aerosol-cloud radiation interaction. This repeatedly leads to erroneous weather forecasts during special weather conditions that involve aerosols, such as strong Saharan dust episodes or biomass burning events.

The project PermaStrom (May 2020 – April 2024), namely “Photovoltaic power prediction to better manage the influence of the atmospheric aerosol on the electricity grids in Germany and Europe” (translation from the German title), follows up on the developments of the predecessor project PerduS (March 2016 – February 2020), where global and refined regional forecasts of dust aerosol were established in a pre-operational version at DWD. Within PermaStrom, the model system is extended with more aerosol species (soot from biomass burning and sea salt) to further improve photovoltaic yield forecasts. Additionally, the treatment and parameterization of aerosol-cloud interactions are studied in a high-resolution LAM (limited area mode) setup along with ICON-ART ensemble predictions. The model developments are continuously validated against various ground- and satellite-based measurements of aerosol concentration and distribution, clouds and solar radiation. The dust forecasts are planned to become operational at DWD in its own data assimilation cycle in winter 2023/2024 with a global 26 km grid and a two-way 13 km nest covering Europe, North Africa and the North Atlantic.

This contribution will provide an overview of the research and model development activities within the PermaStrom project. Additionally, we will show validation results for several aerosol events with direct attention to photovoltaic power production in Germany.

How to cite: Filipitsch, F., Bachmann, V., Porz, N., Seifert, A., Förstner, J., Wagner, A., Hoshyaripour, A., Rohde, A., Kumar, P., Menken, J., Vogel, H., Doppler, L., Mattis, I., and Wacker, S.: Daily global aerosol forecasts with ICON-ART to reduce forecast errors for photovoltaic power generation caused by high loadings of aerosols, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-340, https://doi.org/10.5194/ems2023-340, 2023.

Understanding and characterizing long-term variability of surface solar irradiation (SSI) is essential for both climate studies and industrial applications. Recent studies on the long-term variability of SSI have shown that statistically significant trends -or more precisely long period waves- can be evidenced. These local "trends" vary spatially and temporally: coherent regions and periods with decrease (“dimming”) and increase (brightening”) in SSI have been revealed by the statistically analysis of worldwide observational networks [1]. The identification of brightening/dimming periods and the detection of the local trend is a difficult problem due to the entanglement of effects of the two unknowns (localization of the period and quantification of the trend) as well as the small order of magnitude of the trend with respect to the interannual SSI variability.

We propose to evaluate the potential of a change-point detection and time series decomposition algorithm called rBeast [2] to jointly isolate brightening/dimming periods and evaluate SSI local trends. The rBeast algorithm was selected because its formulation - assuming a decomposition of a time series in a trend, waves and seasonal and noise terms over different intervals - is particularly adapted to the long-term SSI analysis. In addition, its implementation which uses a Markov-Chain Monte-Carlo avoids any data preprocessing (e.g. application of Gaussian low-pass filter) and yields an uncertainty estimates of the local trends as well as of the transition instant between brightening and dimming periods.

The work presented is structured in two parts. A verification of the potential of the rBeast method to isolate brightening and dimming periods and to estimate trends has first been verified on synthetic but plausible data. The use of synthetic data allows to evaluate the sensitivity of the algorithm to different factors such as the data availability or the order of magnitude of the trends. Then, the algorithm has been applied to time series analyzed in previous papers on different locations equiped with long-term ground stations and the results are compared to the outputs of the rBeast algorithm. The results obtained for the Potsdam station agree with the periods highlighted in the literature and the uncertainty on the trends and change point given by the method represent a valuable information for the analysis of long-term variation of the SSI.

[1] Wild, M. (2009), Global dimming and brightening: A review,J. Geophys. Res.,114, D00D16, doi:10.1029/2008JD011470.

[2] Zhao, K., Wulder, M. A., Hu, T., Bright, R., Wu, Q., Qin, H., Li, Y., Toman, E., Mallick, B., Zhang, X., Brown, M. (2019) Detecting change-point, trend, and seasonality in satellite time series data to track abrupt changes and nonlinear dynamics: A Bayesian ensemble algorithm, Remote Sensing of Environment, Volume 232, 2019, 111181, ISSN 0034-4257, https://doi.org/10.1016/j.rse.2019.04.034.

How to cite: Saint-Drenan, Y.-M., Blanc, P., and Cave, J.: Application of a change-point detection and time series decomposition algorithm to the analysis of long-term interannual surface solar irradiation variations, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-499, https://doi.org/10.5194/ems2023-499, 2023.

In energy systems with a high share of renewable energy supply (RES), accurate forecasting methods are very important for a secure and economical energy supply. For applications such as demand-supply balancing, short-term forecast (nowcasting) of spatial RES power generation is particularly important. In the literature, most models addressing this need use satellite-derived SSI estimations as input and optical flow or block-matching algorithms to predict their motion. Recently, new families of algorithms based on deep learning have appeared with great potential for improving the performance of nowcasting systems. It is thus of relevance to assess and understand the potential and limitations of deep learning approaches. Furthermore, when maps of solar irradiance are predicted, the metrics used to evaluate the forecasts are usually computed pixel-wise and thus ignore important spatial features, such as the consistency of spatial variability and spatial resolution.

In this work, we investigate the potential of deep-learning algorithms for the forecasting of maps of 15-min solar surface irradiance (SSI) over short-term horizons, from nowcasts of SSI provided by CAMS radiation. We focus on a state-of-the-art deep learning model, designed for spatio-temporal processes: the convolutional long-term short-term network (ConvLSTM). We compare it to a “classic” forecasting model, based on an optical-flow algorithm (TVL1). We first perform a pixel-wise analysis of the models’ accuracy for forecasting horizons between 15 minutes and 3 hours. We then use Fourier spectral analysis to quantify the impact of each forecasting model on the spatial features of the SSI. Finally, we investigate the impact of the loss function used to train the convLSTM.

Our results show that convLSTM and TVL1 have similar pixel-wise performances for short time horizons (15 and 30 minutes ahead), whereas, for larger horizons, convLSTM has a significantly lower RMSE and higher correlation. Fourier analysis, however, reveals that this improvement in pixel-wise accuracy comes with a degradation of the spatial features of SSI. TVL1 forecasts indeed have realistic spatial variability for all tested horizons, but convLSTM produces increasingly smooth predictions: for horizons beyond 2 hours, and despite its higher accuracy, convLSTM acts as a low-pass filter and fully ignores high spatial frequencies. This shows that the gains in accuracy are obtained at the expense of the fine spatial structure, which is a well-known phenomenon in forecasting. However, highlighting and quantifying this effect is important because smoothing can be problematic in some applications such as variability or ramp forecasting.

Using hybrid loss functions penalizing the lack of variability in the forecasts indeed improves the spatial behavior of the deep-learning model without significantly reducing its pixel-wise performance. At large forecast horizons, however, such hybrid loss functions cannot prevent a substantial loss of spatial variability and convLSTM predictions remain overly smooth.

How to cite: Verbois, H., Saint-Drenan, Y.-M., and Blanc, P.: Assessing the Potential and Limitations of Deep Learning for Solar Irradiance Nowcasting across Large Geographical Areas, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-417, https://doi.org/10.5194/ems2023-417, 2023.

The Copernicus Atmosphere Monitoring Service (CAMS) provides open data access to solar irradiances through its CAMS Radiation Service (CRS). Data is accessible via both the CAMS Atmospheric Data Store and user-specific Python libraries as the PV-LIB.

Observations from the MSG meteorological satellites are combined with modelled aerosols, water vapor, and ozone from the CAMS Integrated Forecasting System. Irradiance time series in 1 min, 15 min, hourly and daily temporal resolution are produced 'on-the-fly' on user request - using the most recent method and input datasets at the desired location inside the domain of the MSG satellite view. In addition to the standard access as time series, a gridded dataset in 0.1° and 15 min for the years 2004-2022 is available for land surfaces of Europe and Africa. 

CRS v4.5 was released in 2022 and uses APOLLO_NG for deriving cloud information based on a probabilistic cloud detection. Its usage in Heliosat-4 follows a detection approach that is highly optimized for solar energy needs. The method is operationally applied for Europe, Africa and the Middle East, but has also been tested for Asia/Australia, and North and South America.

The derivation of the radiation at the earth’s surface was extended by a parameterization of circumsolar radiation and provides a more accurate validation of the direct radiation with pyrheliometers. Besides concentrating solar power technologies, accurate direct radiation is needed to provide direct/diffuse ratios e.g. for tilted irradiances at solar module planes.

CRS v4.5 switched to CAMS reanalysis as input. This allows improved accuracy for all years 2004 – 2020 compared to the usage of the CAMS IFS in its various versions over time. This accuracy gain is larger for years before 2020 with their older CAMS IFS versions used in the CRS before v4.5.

The bias correction methods used so far have compensated for offsetting errors due to aerosols and clouds, obscuring the opportunity for improvement in both the cloud and aerosol algorithms. Various method improvements have now eliminated the need for operational bias correction in the CRS.

Typical improvement from CRS v3.2 to CRS v4.0 (new cloud scheme, bias correction active) to CRS v4.5 (new clouds, new CAMS reanalysis, no bias correction active) will be discussed in the presentation. Most recent evaluations done for Himawari-8 Field of View will be added.

Furthermore, the operational CAMS integrated forecast system (IFS) provides radiation forecasts in hourly resolution for up to 5 days. These forecasts are evaluated against ground based observations of radiation and compared against the spatially higher resolved ECMWF HRES forecasts. 

How to cite: Schroedter-Homscheidt, M., Azam, F., Lezaca, J., Lefevre, M., Mantelier, N., and Saint-Drenan, Y.-M.: The new CAMS Radiation Service v4.5 – method improvements with a special focus on solar energy user needs, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-592, https://doi.org/10.5194/ems2023-592, 2023.

PV / solar power production is fostered to become one of the most powerful renewable energy sources in Central Europe to tackle our growing energy demands. While the just recently emerging PV power observations often record highly resolved real-time data for a relatively short time-horizon, solar irradiance offers long time-series using automatized weather station networks. Being closely linked, irradiance can act as a relevant input and estimate for PV forecasting. Particularly, long time series may be exploited for generating synthetic training datasets. However, both irradiance and PV can be challenging in forecasts using machine learning methods as they embrace a high degree of diurnal and seasonal variation.

Deep learning offers us new opportunities to generate highly resolved weather forecasts by learning relations in complex datasets. In this study, we investigate the suitability of several deep learning techniques for irradiance and PV nowcasts. Our main models investigated includes a sequence-to-sequence LSTM (long-short-term-memory; a type of artificial neural network) model using a climatological background model or NWP (numeric weather forecasting model) for post-processing, a Graph ANN (artificial neural network) model, and an analogs based deep learning method. Relevant input features include 3D-fields from NWP models (e.g.: AROME), satellite data and products (e.g.:  CAMS), radiation time series from remote sensing, and observation time-series.

Results for selected topologically diverse locations obtained by the developed method yield, in general, high forecast-skills, where we elaborate on interesting cases studies from a meteorological point of view. Different combinations of inputs and processing-steps are considered. We compare obtained forecast results to forecasts produced by traditional methods.

How to cite: Papazek, P. and Schicker, I.: Deep Learning Approaches for High-resolution Solar Irradiance and Solar Power Nowcasting, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-387, https://doi.org/10.5194/ems2023-387, 2023.

The presentation refers to DeepSky, a project aiming to develop an innovative and flexible on-site measurement system to fully address the needs of end-users of the meteorological, atmospheric and solar communities. DeepSky is based on the analysis of information retrieved by two head camera (an all-sky imager (ASI) and a thermal camera), radiative transfer in solar / thermal spectral bands and methods of computational optics and deep learning. The proposed system is able to reliably assess a range of geophysical variables (solar radiation, cloud cover percentage and type, aerosol optical properties, etc.).

For this scope, a bouquet of methodologies has been developed. Aerosol Optical Depth (AOD) is retrieved by the ratio of red to blue (RBR) color intensities, which are relevant to radiances at wavelengths 440 nm and 675 nm. A cloud type identification methodology is presented by exploiting the sky condition information: the proposed methodology uses a k-Nearest-Neighbor algorithm, considering as inputs specific information derived from the ASI such as color intensity, the cloud coverage, the saturated area around the Sun, the raindrop appearances and solar zenith angle. A methodology to retrieve precipitable water under cloud-free conditions using images from a thermal-infrared camera is also presented, by examining the relationship between PW and zenith-sky temperature.

Finally, this work is focused on modeling the global and diffuse horizontal irradiances (GHI and DHI) using deep learning techniques and ASI-derived information. The preliminary estimations underestimate GHI and DHI observations with systematic biases of –1.8 W m^–2 and –0.5 W m^–2, while the dispersion errors are 82.7 W m^–2 and 39.8 W m^–2, respectively. The correlation coefficient is high, approaching 0.95 and 0.85 for GHI and DHI.

Overall, a new on-site monitoring system is presented covering multiple needs in the areas of atmospheric physics and solar energy and providing parameters for which separate instruments would be needed.

How to cite: Kazantzidis, A., Salamalikis, V., Tzoumanikas, P., Logothetis, S.-A., Giannaklis, C., Tsourounis, D., Economou, G., and Theocharatos, C.: Estimation of atmospheric parameters and solar irradiance based on a sky imaging system, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-564, https://doi.org/10.5194/ems2023-564, 2023.

The transition to a fossil-free energy system requires the rapid installation of photovoltaic (PV) systems. For Germany, the government is targeting an installed PV capacity of 215 Gigawatts (GW) by 2030 (70 GW today). This target is associated with an increased installation rate of 22 gigawatts PV per year (around three times compared to the 2021 rate). In urban areas, the majority of systems will be installed on rooftops connected to the low-voltage grid. It is therefore likely, that the majority of suitable rooftops will be equipped with PV systems in the next 10 years. In parallel, significant changes in load patterns (e.g. e-mobility, heat pumps) and the integration of battery storages can be expected. 

The efficient integration of the additional PV systems into the electrical grid also requires a detailed understanding of the generation profiles at different levels from the household to the transformer. Therefore, the impact of very short-term solar irradiance variability on ramp rates and balancing effects should be investigated for scenarios with decentralized, but much denser PV generation than today. Since this variability is mainly caused by small scale cloud dynamics, high resolution information on temporal and spatial cloud cover and irradiance distribution is needed.

In northwestern Germany, DLR has installed and is operating Eye2Sky, a dense network of allsky imagers (ASI). At 30 different locations, high-resolution fisheye images of the sky are taken every 30 seconds. At 10 locations, the images are complemented by radiation and meteorological measurements. The Eye2Sky network covers about 100 km x 100 km centered at the city of Oldenburg. It has a low ASI density in rural areas and a high density in city of Oldenburg, thus providing an almost complete coverage of the city. 

Eye2Sky is used to study solar irradiance variability in the city of Oldenburg at a high spatial (50 meters) and temporal (30 seconds) resolution. This enables simulations of single rooftop PV systems. Compared to state-of-the art radiation data sources like satellite images or numerical weather models (NWP), the camera information in Eye2Sky resolves cloud details that cause solar irradiance fluctuations on small scales down to household level.

In this work, we would like to present a solar irradiance nowcasting validation from the ASI network in Oldenburg and its comparison with methods based on satellite (MSG) as well as NWP (ICON-D2) data. Emphasis will be made on the ability of the different methods to reproduce the spatio-temporal variability under different cloud conditions.

How to cite: Schmidt, T., Stührenberg, J., Blum, N., Lezaca, J., Hammer, A., von Bremen, L., and Vogt, T.: Solar irradiance nowcasting based on a network of all-sky imagers: the value of high-resolution data on variability information, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-577, https://doi.org/10.5194/ems2023-577, 2023.

The National Solar Radiation Database (NSRDB) provides global solar resource data at a high temporal and spatial resolution. This data is primarily used in solar energy modeling and is updated on a regular basis. The NSRDB uses a physical approach to satellite-based solar modeling. The underlying Physical Solar Model (PSM) computes cloud-properties using satellite remote sensing and subsequently solar radiation using radiative transfer models. The retrieved cloud properties include cloud-mask, cloud-type, cloud optical depth and cloud droplet size. The radiative transfer models require additional input parameters such as aerosol optical properties (AOD), preciptable water vapor, surface albedo, temperature and pressure to accurately model solar radiation. While cloud properties are obtained directly from the geostationary satellites other inputs are obtained from additional source such as the National Aeronautical and Space Administration (NASA) Modern Era Retrospective Analysis for Research and Applications version 2 (MERRA2), the Interactive Multisensor Snow and Ice Mapping System (IMS) model data from the U.S. National Ice Center and NASA’s polar orbiting satellites such as the Moderate Resolution Imaging Spectroradiometer (MODIS) instruments on the Aqua and Terra Platform.

 In 2022 the NSRDB was updated using the latest version of the underlying PSM. This update includes improved surface albedo and gap-filling of cloud properties. The inclusion of these updates reduced the uncertainty in the data compared to previous versions of the NSRDB. The Himawari and Meteosat Indian Ocean Data Coverage satellites were added to the Geostationary Operational Environmental Satellite and made our coverage global. While standard data from the GOES continues to be served at an hourly 4km x 4km resolution, full resolution data has also been made available to the user. The user is provided significant flexibility for downloading data depending on the amount of data required. Data can be downloaded using either the web-interface, an Application Programming Interface or directly from the cloud using Amazon Web Services. Services such as spectral data use on-demand computation and delivery.

Evaluation of the NSRDB was conducted for 18 stations and the the Mean Bias Error (MBE), Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) were computed for both GHI and DNI. The evaluation was conducted for the 1998-2021 period. Generally, the MBE lies within plus or minus ±5% for GHI and ±20% for DNI. The RMSE is less than 30% for GHI and 35% for DNI. 

Significant new updates are planned for 2023 including the use of the new FARMS DNI model under cloudy sky situations. There are additional plans to partition the NSRDB data using cloud fraction when evaluating the accuracy of the NSRDB. This presentation will provide users with the latest information about the NSRDB as well as plans for future development and updates.

How to cite: Sengupta, M., Habte, A., Xie, Y., Buster, G., Benton, B., and Yang, J.: New Capabilities in the National Solar Radiation Data Base, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-593, https://doi.org/10.5194/ems2023-593, 2023.

Climate-neutral societies rely on large-scale renewable energy expansion. With increasing renewable energy capacities, the impact of renewable energy forecasts errors on energy systems magnifies which strengthens the need to better understand and describe them. Renewable energy forecast errors are particularly interesting when studied in their spatiotemporal domain: Uncorrelated or even negatively correlated forecast errors can get smoothed out reducing the regional impact of forecast errors whereas correlated renewable energy forecast errors can result into largescale deviations between plannable and dispatchable renewable energy which can affect the entire energy system. Despite its importance, an analysis of the spatio-temporal characteristics of short-term renewable energy forecast errors is still lacking. This study aims to close this gap by studying the spatio-temporal characteristics of short-term (1 to 3 days forecast lead time) renewable energy NWP (ECMWF IFS HRES) forecast errors in Europe. The spatio-temporal relationship is described by deriving characteristic correlation lengths for each site whereas the effect of smoothing is calculated through the analysis of spatially convoluted forecast errors. We show that solar and wind energy forecast errors have fundamentally different behavior subject to site characteristics and forecast lead times. Furthermore, we identify regions which are prone to forecast smoothing or accumulation. Lastly, we illustrate that accumulated forecast errors for actual renewable sites in Europe are uncorrelated over large distances, yet forecast error clusters can be identified for wind and solar energy. This is particularly crucial for the case of planned wind offshore farms as we show for the EU wind offshore expansion plans until 2030. 

How to cite: Zech, M., Schroedter-Homscheidt, M., and von Bremen, L.: Spatio-temporal relationship of short-term renewable energy forecast errors, EMS Annual Meeting 2023, Bratislava, Slovakia, 4–8 Sep 2023, EMS2023-604, https://doi.org/10.5194/ems2023-604, 2023.