Short-Term Power Load Forecasting Based on Empirical Mode Decomposition and Deep Neural Network

Abstract

Short-term load forecasting predicts the hourly load of the future in few minutes to one-hour steps in a moving window manner based on historical and real-time data collected. Effective forecasting is the key basis for in-day scheduling and generator unit commitment in modern power system. It is however difficult in view of the noisy data collection process and complex load characteristics. In this paper, a short-term load forecasting method based on empirical mode decomposition and deep neural network is proposed. The empirical modal number determination method based on the extreme point span is used to select the appropriate modal number, so as to successfully decompose the load into different timescales, based on which the deep-neural-network-based forecasting model is established. The accuracy of the proposed method is verified by the testing results in this paper.