Research on Medium and Long-Term Trading Power Prediction Based on Marine Predator Algorithm to Optimize LSTM

In order to solve the problems of overfitting, low prediction accuracy and efficiency in the current medium and long-term trading electricity forecasting methods, a new mediumand long-term trading electricity forecasting model based on Long Short-Term Memory (LSTM) was proposed. Firstly, the influencing factors of medium and long-term trading electricity are determined and the correlation analysis is carried out. The Marine Predators Algorithm (MPA) was used to automatically optimise the hyperparameters of the hidden layer neurons, training batches and initial learning rate of the LSTM network, and five influencing factors of GDP, secondary industry GDP, urban resident population, per capita disposable income of urban residents and monthly average temperature were selected as input variables, and the monthly transaction electricity was used as the output variables. The MPA-LSTM model was applied to the monthly trading electricity forecasts in different years. Finally, the MPA-LSTM model and other three traditional methods proposed in this paper are used to predict the monthly trading electricity in a region in 2022, and the error index is used to evaluate the prediction results. The analysis of the prediction results shows that the prediction accuracy of the LSTM model optimised by MPA has been effectively improved, which provides a reference for improving the prediction accuracy of regional trading electricity.