Short-Term Residential Load Forecasting Based on Smart Meter Data Using Temporal Convolutional Networks

Abstract

Short-term residential load forecasting (STRLF) is the crucial step of customer side demand response (CSDR) that is mainly applied to peak cut through the adjustment of electricity price. Compared with load forecasting of high voltage level, STRLF is a more challenging task due to the high volatility and randomness of load. Most studies focus on STRLF using traditional machine learning and recursive neural network technology, which are difficult to maintain long-term load memory. Temporal Convolutional Networks (TCN), a deep learning method, is put forward to predict residential load which can not only keep the load memory longer, but also process the load information in parallel. Based on AMPds2 smart meter data set, experiments show that the proposed method has a great advantage over the state-of-the-art methods.