Short-term LMP forecasting using an artificial neural network incorporating empirical mode decomposition

摘要

A deregulated power market utilizes wholesale and retail transactions as two of its major business models. Locational marginal prices (LMPs) originating from power markets and system operations imply electricity values in a location or at a bus. Market participants can develop bidding strategies according to vital information of LMPs. The security coordinator can conduct market redispatch for congestion management based on LMP serving as a crucial indicator. This work presents a novel method using the empirical mode decomposition (EMD) incorporating correlation coefficients and neural networks using the back-propagation (BP) algorithm for undertaking short-term LMP forecasting. The correlation coefficient identifies coherent LMPs that act as inputs for the BP-based neural network to forecast the next LMPs. Additionally, validity of the proposed method is verified using historical LMPs in the PJM market in USA. Simulation results demonstrate that the proposed method can forecast short-term LMP values more efficiently than the traditional autoregressive integrated moving average, BP-based network without EMD, BP-based artificial neural networks with multi-resolution analysis of the wavelet transform and recurrent neural network. Copyright © 2014 John Wiley & Sons, Ltd.