Temporal inception convolutional network based on multi‐head attention for ultra‐short‐term load forecasting

Accurate load forecasting is essential for ensuring safe, stable, and economical operation of energy internet. Temporal convolutional networks (TCNs) have demonstrated superior performance, when compared to recurrent neural network models, since their introduction in electrical forecasting. However, the current TCN-based models are unable obtain a large receptive field strong long-time feature extraction capability owing specific kernel size 1D convolution structure. This paper proposes temporal inception based on multi-head attention (TICN-Att) ultra-short-term prediction. By introducing an structure into TCN, proposed model can extract multi-dimensional information from input features, through multiple hidden kernels different scales, without stacking layers depth-wise. Simultaneously, by mechanism, TICN-Att has long time-dependent capability, similar that short-term memory models. The generalization validity tested using global competition (GEFCOM2014) dataset, data city Jiangsu (China), PJM power system dataset. experimental results best prediction effect, other state-of-the-art