Abstract:

With the goal of carbon peaking and carbon neutrality， renewable energy， represented by photovoltaic solar energy， has attracted a lot of attention. However， due to the increasing complexity of urban power systems， the short-term forecasting of time-of-use electricity load for microgrids is not high. To address this issue， a short-term forecasting model that integrates various empirical mode decomposition methods and neural network models is proposed. By grouping， normalizing， and performing empirical mode decomposition on the initial data signal， multiple sets of intrinsic mode function （IMF） subsequences are obtained. Furthermore， all subsequences and the original data are respectively fed into a Long Short-Term Memory network （LSTM） and a Backpropagation Neural Network （BPNN） to obtain the forecasting model. Finally， an empirical test is conducted using a large distributed photovoltaic power station in Shanxi Province as an example. The results indicate that， compared to single or combined methods， the proposed forecasting model based on Ensemble Empirical Mode Decomposition （EEMD） and LSTM demonstrates higher accuracy in short-term time-of-use load forecasting for microgrids， showing significant superiority. Therefore， significant application value in time-of-use load forecasting for microgrids is demonstrated and can also be applied in fields such as wind power generation to assist enterprises and managers in making informed decisions.

Key words: microgrid, short term load forecasting, empirical mode decomposition, long-short term memory, back propagation