Power-Load Forecasting Model Based on Informer and Its Application

Worldwide, the demand for power load forecasting is increasing. A multi-step power-load model established based on Informer, which takes historical data as input to realize prediction of in future. The constructed abandons common recurrent neural network deal with time-series problems, and uses seq2seq structure sparse self-attention mechanism main body, supplemented by specific output modules long-range relationship time series, makes effective use parallel advantages mechanism, so improve accuracy efficiency. trained, verified tested using dataset Taoyuan substation Nanchang. Compared RNN, LSTM attention other models a cyclic network, results show that efficiency Informer-based 1440 steps have certain over models.