Data-driven hidden solar PV and energy storage capacity estimation from the net-load of active distribution systems

The proliferation of distributed energy resources (DER), particularly solar photovoltaic (PV) systems, has introduced challenges in managing active distribution systems. Due to their behind-the-meter installation, network operators often lack visibility in PV generation. Accurate net-load forecasting, which considers both load demand and DG output, is essential for ensuring grid stability and reliability. This research presents a data-driven approach to address these challenges. A novel method is proposed for estimating the capacity of DER, including PV and energy storage systems (ESS). Furthermore, a reinforcement learning-based ESS control strategy is devised to maximize the economic benefits of PV-battery integrated systems. A deep learning-based long short-term memory and Gated Recurrent Unit model is developed for net-load forecasting. Finally, to enhance model performance and reduce computational complexity, feature selection is implemented using the Shapley value technique. Simulation results demonstrate that the proposed approach achieves absolute percentage errors of 4.72 % and 47.87 % in PV and ESS capacity estimation, respectively. The proposed charging strategy increases the annual return of the PV-battery integrated system by an average of 1304 THB/kWp. Additionally, annual ESS utilization is reduced by an average of 2.91 % with the proposed strategy.