Error estimation of statistical baseline load methods

Abstract

Customer baseline load (CBL) is widely used as an evaluation tool in demand response (DR) programs. Due to uncertainty of load, it is difficult to establish an accurate CBL model and the calculation error is inevitable. Therefore, calculation error of CBL must be estimated, so that DR programs can be evaluated effectively. Consumers' load changes with periodic pattern of production and life, meanwhile, it is also affected by some random factors. To reveal the relevance between CBL and the influence brought by the periodic trend and random factors of the load, an analysis is done in the paper. Firstly, discrete Fourier transform (DFT) is used to decompose the historical load into different periodic components. Then error of CBL calculation methods is analyzed. It can be seen that periodic components have a close relationship with the error limit of each method and give valuable information to determine the error range. Case studies demonstrate the validity of the analysis.