A Baseline Load Estimation Approach for Residential Customer based on Load Pattern Clustering

Abstract

Demand response (DR) is a key technology enabling reliable and flexible power system operation more economically and environment-friendly than conventional manners from supply side. Customer baseline load (CBL) estimation is an important issue in the implementation of DR programs for assessing the performance of DR programs and designing economic compensation mechanisms. The accurate estimation of CBL is critical to the success of DR programs because it involves the interests of multi-stakeholders including utilities and customers. Motivated by the inaccuracy of existing CBL methods, this paper proposes a residential CBL estimation approach based on load pattern (LP) clustering to improve the accuracy of CBL estimation. First, an adaptive density-based spatial clustering of applications with noise (DBSCAN) algorithm is proposed to extract typical load patterns (TLPs) of each individual customer in order to avoid the adverse effects from aggregating many dissimilar LPs together as the real TLP. Second, K-means clustering is utilized to segment residential customers into several different clusters based on the similarity of LPs. Finally, CBLs for DR participants are estimated based on the actual load of non-participants at the same cluster during DR event periods. The proposed methods are compared with some traditional methods on a smart metering dataset from Ireland. The results show that the proposed methods have a better performance on accuracy than averaging and regression methods.