Short-Term Power Load Forecasting in City Based on ISSA-BiTCN-LSTM

Accurate short-term power load forecasting is crucial for the stable operation of power systems. In this paper, we propose an advanced forecasting model that combines the Salp Swarm Algorithm (SSA), Bidirectional Temporal Convolutional Network (BiTCN), and Long Short-Term Memory (LSTM). The model first exploits the parallel fusion of BiTCN and LSTM (BiTCN-LSTM), taking full advantage of BiTCN's strength in parallel processing of local features and the LSTM's ability to capture long-term dependencies through its gating mechanisms. Subsequently, the Improved Salp Swarm Algorithm (ISSA) is enhanced through adaptive leader ratio adjustment, dual-food design, and food lure follower strategy. Finally, the hyperparameters of the BiTCN-LSTM model are optimized using ISSA to improve the model performance. In the short-term load forecasting experiments, electric load data and weather data from Los Angeles, Tetouan, and Johor were used to compare the proposed model with eight existing models. The evaluation metrics included root mean square error (RMSE), mean absolute error (MAE), normalized root mean square error (NRMSE), and mean absolute percentage error (MAPE). The experimental results showed that the model achieved lower error values than the comparison model in most cases in different seasons, working days, and rest days in different cities. In particular, the error values of RMSE, MAE, NRMSE, and MAPE were 925.11 kW, 732.63 kW, 0.019, and 1.034% for the rest days in the city of Tetouan, respectively. Compared with other algorithms, ISSA demonstrates stronger optimization capability and shorter optimization time. Additionally, model structure analysis was conducted through optimization comparison and ablation experiments, further demonstrating the proposed model's strong predictive performance.