A Novel Short-Term Intelligent Forecasting Method for Bus Loads Driven by Hybrid TCN-BiLSTM

Power load forecasting is a significant portion of power system operation and planning. With massive access to various new energy sources and distributed power sources, the non-linear characteristics of the load are significantly enhanced, which puts greater demands on the load forecasting model. However, most of the existing load forecasting models only extract load characteristics at a single time scale, ignoring the different trends in load data at different time scales. To this end, this paper uses a Temporal Convolutional Network (TCN) to capture load fluctuation features at multiple time scales, and feeds the extracted abstract synthesis vectors into a Bi-directional Long Short-Term Memory (Bi-LSTM) network to fully exploit the positive and negative order relationships of load data. Finally, the TCN-BiLSTM hybrid neural network proposed in this paper is validated using the power load data of Yangjiang City, Guangdong Province, China. The experimental results demonstrate that the network model presented in this paper exhibits excellent performance achieves a high level of accuracy in load forecasting, and has certain engineering application value.