Research on a load forecasting model based on ACMD and BiGRU-Attention

To lessen the impact of fluctuation and randomness of the user-side load on the load forecasting accuracy， a BiGRU-Attention （bidirectional gated recurrent unit with attention） short-term load forecasting model based on ACMD （adaptive chirp mode decomposition） and the Attention mechanism is proposed. Firstly， ACMD is used to decompose the load time series into several relatively regular subcomponents； then， the BiGRU model is used to predict the subcomponents and sum them up to obtain the final prediction results. To highlight the influence of important information， the Attention mechanism is introduced into the BiGRU model to give corresponding weights to the implied states of the BiGRU network. The sparrow search algorithm is used for the optimal selection of model hyperparameters to reduce the impact of misselection of model hyperparameters. An open data set is used for example analysis that is compared with a single model and combined model respectively. The results show that the method is superior in prediction.