Evaluation of a Kalman predictor approach in forecasting PV solar power generation

Increased adoption of solar photovoltaic (PV) generation has motivated the investigation of improved forecasting techniques. This is due to the disruptive nature of variable energy sources on the existing electric utility infrastructure. This research develops a multiple-rate Kalman predictor that provides real-time forecasting of solar PV generation. The proposed method is evaluated on an operational 13 kW PV array installation at fifteen minute intervals using one-minute measurement data. Forecast values are posted once per-hour for a fifteen minute-ahead market scenario. Two designs are compared that contrast the trade-offs of steady-state variance versus transient following capability. Overall experimental results indicate that a multiple-rate Kalman predictor is a promising technique for improved solar PV forecasting.