TCN-BiGRU power load prediction based on improved gray wolf optimization algorithm

To improve the accuracy of short-term power load forecasting, this paper proposes a TCN-BiGRU model based on an improved grey wolf optimization algorithm. In this framework, the input sequence is first processed by an enhanced temporal convolutional network (TCN) to capture long-term dependencies, and then by an improved self-attention-optimized bidirectional gated recurrent unit (BiGRU) to extract bidirectional dependencies. an auto regression (AR) module and an election mechanism are integrated within the model to enhance forecasting accuracy. Finally, the model parameters of the TCN-BiGRU are optimized using the improved grey wolf optimization algorithm to further boost its overall performance. Experimental simulations demonstrate that the proposed model achieves a mean absolute percentage error (MAPE) of 4.974%, mean absolute error (MAE) of 0.029, and root mean square error (RMSE) of 0.034, outperforming mainstream benchmark models and effectively enhancing load forecasting accuracy.