Power load forecasting based on Time convolution network and bilstm optimized time series algorithm

This paper discusses the optimal time series algorithm based on time convolution network (TCN) and bidirectional long-term and short-term memory network (bilstm), in order to improve the accuracy of power load forecasting. Through comparative experiments, we evaluated three different models: a separate time convolution network, a bidirectional long-term and short-term memory network, and a bidirectional long-term and short-term memory network optimized by a time convolution network. The experimental results show that the optimized model performs best on the training set, and its root mean square error (RMSE) is 22.26, which is better than using bilstm (RMSE is 24.12) and TCN (RMSE is 29.08) alone. In the test set, the optimized model also performed well, with RMSE of 27.03, while the RMSE of TCN and bilstm were 29.22 and 33.59, respectively. These results show that the bidirectional long-term and short-term memory network optimized by time convolution network has significant advantages in capturing the complexity and dynamics of power load time series. In addition, the evaluation parameters of the model further confirm the superior performance of the proposed model in power load forecasting. It can not only predict the change of power load more accurately, but also better understand and simulate the behavior mode of power system. Therefore, this study not only provides an effective algorithm for power load forecasting, but also provides a new perspective and method for time series analysis in related fields.