An ISSA-TCN short-term urban power load forecasting model with error factor
Chaodong Fan, Gongrong Li, Leyi Xiao, Lingzhi Yi, Shanghao Nie
Abstract
Abstract Under the general trend that all industries try to reduce energy consumption and carbon emissions, urban power load forecasting directly affects energy planning and power system management decisions. In this regard, this paper proposes an improved salp swarm algorithm (ISSA) optimized temporal convolutional network (TCN) load forecasting model with error factor. The model first introduces the SSA and improves it through adaptive leader scale adjustment, position update modification, and boundary adjustment strategy, aiming to improve the algorithm’s pre-global search and post-local search ability; then, apply the improved algorithm to hyperparameters optimization for TCN, aiming to search for the optimal hyperparameters combination for different power load data, and the forecasting main model is constructed; Finally, to further enhance the accuracy and stability of power load forecasting, the paper considers the forecasting error of the main model and designs a specialized error auxiliary model. Test comparisons with different algorithms on the CEC2017 benchmark function were conducted to verify the effectiveness of the algorithm improvement. In addition, load forecasting experiments were conducted using data from three cities, and the forecast error values were lower than those of other forecasting models under three identical error metrics, showing good short-term load forecasting performance.