Optimization Of LSTM Neural Network For Photovoltaic Power Prediction Based On Golden Jackal Algorithm

In order to further promote the construction of photovoltaic power generation market and improve the market competitiveness of power grid companies,the prediction of photovoltaic power generation power is very important in the planning and maintenance of power grid.Therefore,it is necessary to further study and explore the photovoltaic power prediction.To process sample data,appropriate analysis and correction are required to exclude abnormal data.Due to the different dimensions of the influential factors of photovoltaic power,the prediction effect may be poor if directly input into the prediction model.Therefore,it is necessary to normalize the sample data to eliminate the influence of different dimensions on the prediction of photovoltaic power.Artificial neural network has the characteristics of elastic topology,high redundancy and nonlinear operation,and can quickly find the optimal solution,which plays an important role in the prediction.Therefore,neural network is used to predict the photovoltaic power generation.However,the key parameters of neural network need to rely on the empirical selection of researchers.In order to solve this problem,Golden Jackal Optimization(GJO)algorithm was introduced to automatically optimize its key parameters and find a group of optimal model parameters.In this thesis,GJO-LSTM(Long Short-Term Memory)model is proposed to predict photovoltaic power generation.This method is mainly based on golden jackal optimization algorithm to optimize LSTM neural network parameters,which can be divided into three steps: Firstly,the sample data was cleaned,and variables highly correlated with photovoltaic power were selected by Pearson correlation coefficient as input features,and photovoltaic power as output.Secondly,the key parameters of LSTM neural network were optimized by using the global optimization capability and internal parallel computing capability of GJO optimization algorithm.The photovoltaic power prediction model based on GJO-LSTM was obtained by training.Finally,the error evaluation criteria are used to compare the prediction effect of the model.In order to verify the best prediction effect of the model proposed in this thesis,LSTM,CNN,GRU,GJO-LSTM,GJO-CNN and GJO-GRU models were designed to predict the photovoltaic power generation,and RMSE,MAE and MAPE were used as error evaluation criteria.The results show that the combined model has better prediction effect than the single neural network,indicating that the optimization effect of GJO algorithm is significant.The RMSE,MAE and MAPE values of the GJO-LSTM model proposed in this thesis are all the smallest,indicating that the GJO-LSTM model has the best prediction effect.