关键词: 时间序列预测, 变分模态分解（VMD）, 小波卷积（WTC）, Informer模型

Abstract: With the large-scale integration of renewable energy sources such as wind power and photovoltaic power into the power grid, the uncertainty and volatility of the power system operation have significantly increased, making it difficult to extract load sequence features and improve the accuracy of short-term power load forecasting. To address this issue, a short-term power load forecasting model based on the combination of wavelet transform convolution(WTC) and Informer model is proposed. The improved variational mode decomposition(VMD) is used to decompose and denoise the input data before feeding it into the wavelet convolution module for multi-level wavelet convolution, achieving multi-scale feature extraction of complex time series and reducing sequence complexity, thereby improving the forecasting accuracy. To verify the validity of the model, multiple sets of experiments are conducted. The experimental results show that the Mean Absolute Percentage Error(MAPE) of the model proposed in this paper is 1.893 1%. Compared with the methods of using the Informer model alone or only using GSWOA-VMD-informer, it is reduced by 1.059 4% and 0.504 8%, respectively, verifying the effectiveness of this model.

Key words: time series forecasting, variational mode decomposition(VMD), wavelet transform convolution(WTC), Informer model