Methodologies for Customer Baseline Load Estimation and their Implications

Demand response (DR) has gained increasing importance with the grid modernization and evolution of the smart grid. Customer baseline load (CBL) informs the expected consumption of the consumers and is an important factor in DR management. To develop forecasting methods for CBL that work satisfactorily is a major challenges faced by independent system operators (ISOs) and utilities. Consequently, it is critical that these methods are evaluated based on their performance. The major objective of this paper is to empirically evaluate the performance of CBL estimation from historical observations, for residential customers. In addition to some well established methods, we have used an artificial neural network regression for estimating CBL. We use the CER smart meter data from 6435 residential and small industrial customers collected between July 2009 to Dec 2010, for this evaluation. We performed the error analysis of different methods using both the accuracy and bias metrics.