Day-ahead forecasting for commercial PV with battery storage

Scientists at the University of Pretoria in South Africa have developed a novel day-ahead forecasting method for commercial PV systems connected to battery storage.

The proposed methodology utilizes numerical weather prediction (NWP) models to forecast the PV system's output and then passes it through an optimal control strategy for battery usage.

“The problem to be addressed is to accurately forecast solar energy production to effectively manage solar power variability by integrating a battery storage system to improve the optimization and availability of solar PV energy during high demand levels in commercial sectors,” the researchers added.

For their method, the scientists used irradiance and ambient temperature with the data of a commercial PV system. With that, they calculated the cell temperature, using it to further forecast the output power. Then, a control strategy was applied for the predicted battery use.

“The control strategy first checks the building demand and calculates whether this load can be supplied only from PV,” they explained. “It allows for this if the PV supply can meet the demand. If the demand exceeds the PV supply, it is supplied by battery storage plus PV power. If there is still a deficit in the power supply, grid power will meet the demand. Any excess energy from PV is stored.”

As for the irradiance and ambient temperature, the academics used an unnamed open weather source, which combines the Weather research and forecasting (WRF) model with the Numerical weather prediction (NWP) model, and it is commonly used in solar forecasting, as it can simulate various atmospheric processes.

They tested the new methodology on a 16.8 kW with four strings, each with 14 modules with each an output of 300 W. The AC inverter was assumed to be 15 kW. They also compared the performance of this system with a reference array with the same characteristics in South Africa, for one winter month and one summer month.

Through these measurements, the scientists found that the root mean square error (RMSE) for the summer month was 425.79 W and for the winter it was 595.1 W.

“Furthermore, an excellent positive correlation exists between the predicted output power and the observed results, with R2 values over 90%,” the researchers said.

They presented the new approach in the study “Intelligent solar photovoltaic power forecasting,” published in Energy Reports. The group also comprises academics from South Africa's Tshwane University of Technology and the University of Sharjah in the UAE.