Residential Customer Baseline Load Estimation Based on Conditional Denoising Diffusion Probabilistic Model

With the implementation of dynamic time-of-use (TOU) electricity tariffs and the adoption of smart household appliances, residential power consumption behaviors are becoming more complicated, which brings challenges to the estimation of customer baseline load (CBL). To improve the estimation accuracy, a novel deep learning model called the conditional denoising diffusion probabilistic model (CDDPM) is introduced in this paper. It can not only utilize the intrinsic load information of demand response (DR) participants, that is, the target customers who require CBL estimation, but also the load data of nonparticipants with similar load profiles. The experimental results show the superiority of the proposed model compared with other state-of-the-art methods.