A Baseline Load Estimation Method for Distributed Resource Cluster Users Based on Feature Identification and Net Load Decoupling

In the residential user clusters of distribution areas in China, the number of users purchasing electric vehicles or installing distributed photovoltaics is gradually increasing. The differentiated electricity usage characteristics of these distributed resources lead to inaccurate baseline load estimation for the clusters. To address this issue, a baseline load estimation method for clusters with distributed resources is proposed. First, the user type is identified; second, a wavelet convolutional neural network – long short-term memory (WCNN-LSTM) is used to decouple the net load curve and estimate the individual user's baseline load. Finally, the baseline loads of the user cluster are estimated by summing the individual estimations. By considering the characteristics of various distributed resources, multiple differentiated usage scenarios are constructed for baseline load estimation. The results show that the proposed decoupling method improves the performance of traditional baseline load estimation methods by 45% to 87%. In different datasets, the estimation performance deviation of this method does not exceed 30%. The above analysis indicates that this method has strong adaptability for different baseline load estimation scenarios and can effectively improve the accuracy of baseline load estimation.