High-Fidelity Solar Power Monitoring and Forecasting for Utility-Scale Solar Farms

This project introduced novel forecasting methods for the solar resource and solar power generation by addressing critical gaps in the current forecasting methods. The project focused on technologies that rely on solar irradiance such as concentrated solar power, concentrated photovoltaics and tracking photovoltaics. Power generation for these technologies relies on direct normal irradiance, which is the component of irradiance that requires the most specific development due to high sensitivity to cloud cover and aerosol content in the atmosphere. The critical need for accurate direct normal irradiance forecasting tools is evident by the relative scarcity of native forecasting algorithms for this irradiance component in the scientific literature and its absence from most numerical weather prediction models.
Researchers developed tools to monitor cloud cover that cause solar irradiance fluctuations across the solar plant field. The project team designed and used a new generation of low-cost solar instruments for monitoring and forecasting the solar resource for utility-scale solar farms. The team designed the instruments to provide reliable low-cost telemetry for real-time forecasts and covering detailed patches of the solar field nonintrusively.
The project team created several forecasting models depending on the forecast horizon (such as intrahour, day-ahead) and operational target (for example, reducing monthly imbalances) for two utility-scale solar farms in California. The project demonstrated that the forecasting accuracy for direct normal irradiance and plane-of-array irradiance across all time scales is improved using forecasting models that blend local and remote telemetry. The project results demonstrated the importance of having a rich set of input data to improve forecasting. This effect was especially important for the intrahour time scale, for which the use of multiple solar sensors, sky images, and high-resolution satellite images was found to be essential to reducing forecasting errors.
This report outlines the development and performance of the models. Furthermore, the report also addresses the applicability of the tools to other solar projects of different sizes and weather patterns.