Overview
Toshiba Corporation has developed a technology that precisely forecasts the output of wind power generation, which fluctuates due to changes in the wind and other weather conditions. This technology combines meteorological predictions that account for the topography around wind power plants and Toshiba’s proprietary AI technology to obtain highly accurate forecasts of power generation, which were difficult to achieve in the past. This technology is expected to assist wind power operators in creating accurate power generation plans, thereby contributing to the expansion of wind power installations and the stable supply of renewable energy.
Toshiba Energy Systems & Solutions Corporation (Toshiba ESS) and Next Kraftwerke Toshiba Corporation were selected for the Ministry of Economy, Trade and Industry’s demonstration project, the FY2023 Renewable Energy Aggregation Demonstration Project, in which they verified this technology from December 2023 to January 2024. The results showed that compared with the average error in the previous year’s demonstration, the average prediction error(*1) for wind power generation at the time of the previous morning improved by more than 6% to 10.1%, thus demonstrating high prediction accuracy(*2). This technology will be incorporated in REBSet™, a renewable energy balancing system that was jointly developed by Toshiba ESS and Next Kraftwerke GmbH, and that is planned to be applied in services for power generators and aggregators.
Development background
The rapid deployment of renewable energy sources is necessary to achieve a carbonneutral society. Among these sources, approximately 5GW of onshore wind power has already been installed in Japan, and offshore wind power is also anticipated to expand with designated promotion zones being established(*3). Wind power operators are required to submit accurate power generation plans(*4) to ensure stable electricity supply. However, creating accurate plans is challenging because wind power output is heavily influenced by wind speed and other weather conditions. Discrepancies between planned and actual power generation can lead to an imbalance price being levied, which is a financial burden for the operators. Thus, there is a need for more accurate wind power generation forecasts.
Features of the technology
In response, Toshiba has developed a technology that precisely forecasts wind power generation with high accuracy. The technology achieves this by combining meteorological forecasting and AI technologies, and has the following three features.
The first is its proprietary weather forecasting system, which calculates wind speed data at various altitudes, taking into account the effect of topography near the wind turbine. The second involves AI-based learning from historical wind speed data and correcting the wind speeds near the nacelle(*5) according to the wind direction (wind speed correction AI). The AI also learns the power curve (the output characteristics of the wind turbine) from past wind speed observation data, which shows the relationship between wind speed and power output for each turbine (output characteristic learning AI). The power output is predicted from the wind speed output by the wind speed correction AI and the power curve is learned by the output characteristic learning AI. The third feature is that it creates an ensemble of predicted values from this power prediction model with those from other models, adjusting the mixing ratios of the forecasting methods while learning the trend of prediction errors on a daily basis in order to enhance forecasting accuracy.
With these features, the Wind Power Generation Forecasting AI has achieved high predictive accuracy.
This wind power generation forecasting technology was utilized in the FY2023 Renewable Energy Aggregation Demonstration Project. As a result, the average prediction error for wind power generation at the time of the previous morning was 10.1%. The conventional forecasting technology used in the FY2022 Renewable Energy Aggregation Demonstration Project had an average prediction error of 17.3%, showing a significant improvement in forecasting accuracy.
Toshiba ESS launched a renewable energy aggregation business in 2022 that provides power generation planning services on behalf of power producers(*6). This business bundles and operates renewable energy sources across all national areas of general transmission and distribution operators to achieve stable and economical power supply. The REBSet™ used in this aggregation business incorporates previously developed high-precision solar power generation forecasting technology(*7) and is already in practical use. With the development of this high-precision wind power generation forecasting technology, it is now possible to operate a broader range of bundled renewable energy sources than before.
Future developments
Toshiba plans to incorporate this technology into Toshiba ESS’s REBSet™ and anticipates launching the service within FY2024.
*1: Prediction error is calculated by normalizing the difference between the forecasted and actual generation amounts at 30-minute intervals according to the rated capacity of the generator.
*2: https://www.global.toshiba/ww/news/energy/2024/03/news-20240329-01.html
*3: Promotion zones for the development of marine renewable energy power generation facilities.
*4: A generation plan (kWh) is submitted at 30-minute intervals (48 frames) throughout the day.
*5: Comprising a gearbox, generator, brake system, rotor shaft, and main shaft. Located at the top of the wind turbine tower, and connected to the blades by the rotor shaft and hub.
*6: https://www.global.toshiba/ww/news/energy/2022/05/news-20220517-01.html
*7: https://www.global.toshiba/jp/technology/corporate/rdc/rd/topics/19/1907-02.html