Decoupling based monthly net electricity consumption prediction model considering high penetration of distributed solar PV systems

Abstract

The large-scale introduction of distributed photovoltaic (DPV) increases the need for retailers to consider and quantify the differences in monthly electricity consumption of customers to maximize their interests in trading in the forward electricity market. For customers with DPV, retailers need to predict net electricity consumption (NEC), which is actual electricity consumption (AEC) minus DPV generation. However, the DPV is behind the meter and DPV generation data is invisible to retailers. Therefore, the issue of how to distinguish the transition of customers from no DPV to with DPV and their DPV installation information needs to be addressed. To better capture the additions of DPV timely under high penetration of DPV, a decoupling-based monthly NEC prediction model considering the DPV installation update is proposed. Firstly, the features are extracted from the hourly NEC data of known customers with DPV to distinguish other customers whether installing DPV. Secondly, an online update framework of DPV installation evaluated by two validations is proposed. Thirdly, based on the difference in the electricity consumption series before and after the installation of DPV, the NEC is decoupled into AEC and DPV generation. Finally, the monthly DPV generation prediction results are subtracted from the monthly AEC prediction results to obtain the final monthly NEC results. Different scenarios of DPV penetration are set in case studies to test the performance between the proposed model and other direct models. The results indicate the superiority of the proposed method under high penetration of DPV.