Study of a Aggregated Baseline Load Forecasting Method Based on Decoupling of Distributed PV-Containing Customers

In recent years, the distributed PV industry has developed rapidly, and more and more residential customers have started to install distributed PV power generation systems, while the unobservable characteristics of behind-the-meter distributed PV bring great challenges to the demand response aggregated baseline load (ABL) prediction. To improve the one-inch accuracy of ABL under high penetration distributed PV, a aggregated baseline load forecasting method based on decoupling of distributed PV-containing customers is proposed. Firstly, the abnormal load data is identified and corrected based on t-test in mathematical statistics, and the temperature data is corrected; secondly, a PV user identification method based on weather state characteristics is used to separate PV users and non-PV users; then, the aggregated baseline load of non-PV users is predicted by considering the differentiation of users' electricity consumption patterns; again, the aggregated baseline load for PV users is classified by typical weather types and a prediction model is built for each weather type using SVR for aggregated baseline load prediction; finally, the predicted values for both types of users are combined to obtain the final ABL prediction values. The simulation results show that the proposed method reduces the average absolute error index by 2.65 and the root mean square error index by 0.76 on average compared with the direct prediction method and other classification prediction methods.