PV- Load Decoupling Based Demand Response Baseline Load Estimation Approach for Residential Customer with Distributed PV System

Abstract

Due to increasing installation of distributed photovoltaic systems (DPVSs), load patterns of residential customers become more random, which makes customer baseline load (CBL) estimation harder. This paper proposes a PV-load decoupling approach to improve the CBL estimation accuracy in the presence of DPVSs. Firstly, K-means algorithm is used to divide the customers in control group into k clusters. Secondly, after calculating curve similarity index, each DR participant is matched to the most similar cluster based on the similarity between its load curve and cluster centroids during periods when the distributed photovoltaic (DPV) output power is equal to zero. Then the DPV output power in DR period can be obtained through the estimation model established based on DPV output power of non-DR periods in historical non-DR days and DR event day. Finally, CBL is estimated by the difference between actual load power and DPV output power. Four well-known averaging methods are compared with the proposed approach by using a real dataset of 300 customers in Sydney, Australia. The comparison result indicates the proposed approach shows better accuracy performance.