Improving the Performance of the Truncated Fourier Series Least Squares (TFSLS)Power System Load Model Using an Artificial Neural Network Paradigm

Abstract

Power System Load models have a wide range of application in the electric power industry including applications involving: (i) load management policy monitoring; (ii) assisting with the generator commitment problem; (iii) providing short term forecasts; (iv) aiding with system planning by providing long term forecasts. A method that has been utilized in the power systems planning community involves modeling the power system load (PSL) utilizing a truncated Fourier series. Presented herein is an innovative method based upon analyzing nine weeks of data and generating an optimum number of Fourier series terms included in model structure from a set of preselected heuristic basis functions for prediction. The resulting PSL model capable of providing high quality middle-long term forecasts and retain the shape prediction of the load curve out in time.