High-Fidelity Solar Power Forecasting Systems for the 392 MW Ivanpah Solar Plant (CSP) and the 250 MW California Valley Solar Ranch (PV)
Improving forecasting methods at the Ivanpah Solar Thermal plant
The project demonstrated that the forecasting accuracy for DNI and POA irradiance across all time-scales (intra-hour, intra-day, day-ahead) can be 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 conclusion was more pronounced for the intra-hour time scale, in which multiple solar sensor data -- including sky images and high-resolution satellite images -- are essential to reduce forecasting errors. The tools developed were used in two very distinct testbeds (Ivanpah and CVSR) to predict irradiance and power generation. In both cases, the forecasting accuracy was improved relative to baseline models. The models were successfully applied to other locations, showing that they can be used in other grid-connected solar farms.
The purpose of this project is to develop and validate tools capable of monitoring and forecasting DNI and POA irradiance and the power generation accurately, from 5 minutes out to 72 hours in the future, at the Ivanpah Solar Thermal plant as well as at the California Valley Solar Ranch (CVSR) plant. The project also included the development of tools for predicting wind speed, which affects the heliostats' deployment, and the improvement of the power generation forecast via Resource-to-Power Model (RTP) for Ivanpah (CSP) and CVSR (Tracking PV) plants. The goal of this system is to reduce uncertainties associated with operation, regulation, and scheduling.
The project developed and validated models that lead to lower operation costs and consumer cost per solar kWh due to increased ability to absorb short-term ramps and maintain solar production, better utilization of ancillary generation, better integration of solar assets with utility and ISO operations, and decreased down-time tripping events due to solar variability. Researchers tested an alternative to CAISO’s centralized solar forecasting for CVSR and prepared the power plant for the CAISO real-time market. Results indicate a 67% reduction in the monthly imbalance and eliminates the forecasting fee of $0.10 per MWh.
The forecasting tools developed in this project will help economic development by reducing solar power plant operating cost by more than 10% and having the ability to substantially affect the effective solar capacity in California. These forecasting tools enhanced capacity of utility-scale CSP power plants due to improved prediction of resource and power output and contribute a better integration of solar assets with utility and ISO operations. Additionally, the project has created skilled jobs and prepared talented people with research skills for the California job market.
Environmental & Public Health
The project will result in GHG emission reductions by decreasing the uncertainty associated with solar power generation and diminishing the need of fossil fuel-based generation.
The tools developed in this project result in greater reliability by developing high-fidelity models that increase the accuracy of solar energy forecasting to decrease the number of forced outages and associated ancillary reserves. The forecasting models achieved significant improvement in accuracy for all time horizons. The DNI improvements were 10.2, 40.3 and 43.9 percent of intra-hour, intra-day and day-ahead models, respectively. The POA improvements were 34.1, 38.6, and 62.4 for intra-hour, intra-day and day-ahead models, respectively.