Electricity Load Forecasting Based on SVI Bayesian Uncertainty Quantification

Electricity load forecasting is crucial for transmission planning and security warning in real-life Electricity systems. In this work, we address the problem of how to achieve highly accurate and reliable load forecasting by proposing a bidirectional time convolution network model incorporating an attention mechanism based on the variational modal decomposition method and quantifying the uncertainty of load forecasting in conjunction with stochastic variational inference. The model achieves accurate Electricity load forecasting and provides decision support for the forecasting results through uncertainty. It is demonstrated through experiments on realworld electricity load dataset that the model proposed in this work performs well in various benchmark experiments and can effectively improve the prediction performance and reliability.