Development and Evaluation of Near-Real-Time Solar Power Forecasts

Growing concerns over rising fossil fuel costs, energy security, resource depletion, climate change and sustainable economic development have stimulated the Renewable Energy (RE) sector. Solar energy is on the rise in the United States, but its variable nature and relative unpredictability remains a major obstacle for integration onto the national energy grid. One problem is forecasting the behind the meter power production (e.g., by solar panels in distributed small- and medium-scale systems, such as those on rooftops), which reduces demand. More accurate and reliable solar energy forecasts will help the energy generation and transmission community to plan and manage short-term and day-ahead power generation better, effectively increasing the penetration potential of RE. We extend and apply our Solar Energy Forecasting System (SEFS) to estimate distributed solar energy production for a number of publicly reporting solar installations in the state of Massachusetts in the US, and validate it against their actual reported power production. First, we use a previous year's reported data to verify and calibrate site collection parameters, such as panel area, panel azimuth, and panel elevation, to closely match cloud field simulated versus actual power output for these sites. These reports have operationally derived data issues, e.g., reduced output due to non-meteorological effects such as maintenance outages, periodically inactive inverters, and persistent snow cover, which need be treated during processing. Then, we create a real-time forecast product which estimates the total power output of the aggregated collection of sites. This product estimates power production by hour from the current analysis time out to +48 hours, using our integrated CMG cloud analysis, CPF cloud forecast, ‘Insol' solar insolation model, and solved geometric panel models to estimate total power output. Finally, we use actual reported data to verify and evaluate power production forecasts after-the-fact. Success in this regard shows the practical applicability of these methods to forecast for ever-increasing geographically distributed power generation.