Short-term Electricity Load Prediction and Carbon Emissions Based on Synchrosqueezing Wavelet Transform and Nested Long Short-Term Memory

Accurately predicting that short-term electricity load and carbon emissions are essential for optimizing energy structure, reducing carbon emissions and achieving sustainable development. This article proposes a short-term electricity load carbon emission prediction method based on Synchrosqueezing Wavelet Transform (SWT) and Nested Long Short-Term Memory (NLSTM) neural network. First of all, by introducing SWT to improve the model forecasting accuracy, the electricity load data is decomposed. Then enter the decomposed data input to NLSTM neural network for training and prediction, solving the difficulty of extraction of ordinary Long Short-Term Memory (LSTM) neural networks for the periodic features of the original electricity load, and a problem of randomness in model parameter selection. Finally, combined with the local electricity carbon emissions factor, the predicted value of carbon emissions is finally obtained. The experimental results show that this method has higher predictive accuracy compared to variational mode decomposition (VMD) and single neural network model, providing an effective means for electricity load prediction and refined management of carbon emissions.