Short-term load forecasting of power system based on time convolutional network

Along with the deregulation of electric power market as well as aggregation of renewable resources, a sufficiently accurate, robust and fast short-term load forecasting (STLF) is necessary for the day-to-day reliable operation of the grid. To obtain parameter values that provide better performances with faster speed, this paper uses a temporal convolutional network (TCN), which constituted by the one-dimensional convolutional neural network, for short-term load forecasting. It's commonly thought that recurrent neural networks (RNNs) is the king of short-term load forecasting areas, but the TCN model has faster speed and less memory requirement because of the convolutional neural network structure. The simulation studies are carried out using an hourly power consumption dataset collected from Toronto. Compared with support vector regression machine (SVRM) model and long short-term memory (LSTM) model, the TCN model has the best performance in predicting the short-term electrical load.