Forecast-Guided Economic Dispatch of BESS in Grid-Export-Only PV Systems

Abstract

High short-term variability in photovoltaic (PV) output causes inverter clipping, curtailment, and revenue loss in grid-export-only solar plants. This paper presents a forecast-guided operational framework that integrates an hour-ahead Gradient Boosted Trees (GBT) model with a receding-horizon mixed-integer linear programming (MILP) dispatch strategy for Battery Energy Storage Systems (BESS). A dataset of 30,579 hourly samples, combining NASA POWER meteorological variables and AC-side PV measurements from March 2022 to September 2024, is used to train and evaluate the forecasting model. The GBT achieves strong performance with an RMSE of 0.152 kW and a MAPE of 6.65%. These forecasts drive the MILP scheduler, which incorporates inverter limits, state-of-charge dynamics, non-simultaneous charge/discharge constraints, and throughput-based degradation. A techno-economic comparison of BESS capacities from 250 to 1,000 kWh under fixed PPA and Time-of-Use tariffs shows that all capacities ≥ 250 kWh fully eliminate curtailment, while the TOU–750 kWh configuration maximizes economic returns. The results highlight the importance of moderate BESS sizing for cost-effective deployment in grid-export-only PV systems.