Short-Term Power Load Dynamic Scheduling Based on GWO-TCN-GRU Optimization Algorithm

With the continuous increase of global energy demand and the transformation and upgrading of energy structure, the stability and efficiency of the power system are facing unprecedented challenges. A short-term power load dynamic scheduling model on the basis of temporal convolutional neural network and gated recurrent neural network is designed, and improved through grey wolf optimization algorithm. It simulates the hunting behavior of grey wolf populations, conducts global parameter searches, and finds the optimal temporal convolutional neural network and gated recurrent neural network structures and their hyper-parameter configurations. Subsequently, based on the optimized model, short-term electricity loads are predicted to obtain high-precision load forecasting results. On this basis, combined with dynamic scheduling strategies, real-time scheduling and management of power loads are achieved to ensure the stable operation of the power system. From the results, the designed model converged with an accuracy of 0.98. The root mean square error value was 0.21 when the iteration reached 100. In dataset A, when the iteration was 40, the proposed model had a scheduling time of 2.3s, and its loss function value, accuracy, F1-value, and Micro F1 were 0.203, 0.897, 0.626, and 0.971, respectively. In dataset B, the proposed model had a scheduling time of 2.6s. Therefore, the proposed optimized artificial bee colony algorithm can effectively predict and schedule power loads, providing a solution for scheduling problems in power systems.