Based On EEMD And LSTM-ARIMA Model Research On Power Load Prediction

For modern power system,load forecasting is an important prerequisite to realize economic and safe operation.If the result based on load forecasting has a higher accuracy,it can promote the power grid to complete the dispatch of power resources in a more reasonable way,and can also provide some reference for the development of bidding strategy of electricity selling companies,so that they have a stronger profitability.Therefore,from the current point of view,it is very necessary to study the short-term power load forecasting technology,which has a relatively great significance in reality.Therefore,this paper is aimed at improving the accuracy of power load forecasting.The work of this thesis is embodied in the following three aspects:1.In view of the problem that the traditional load forecasting method cannot make full use of the correlation of data,which leads to low prediction accuracy,the ensemble empirical mode decomposition(EEMD)is established to decompose the time series signal into several intrinsic mode function(IMF)components and trend components,which better reflects the physical significance of the signal.The problem of mode aliasing caused by EMD algorithm decomposition is solved.Through the real load data,EMD and EEMD models were decomposed respectively,and it was found that the modal components obtained from EEMD had more significant regularity.2.As the time interval increases,some models,such as the RNN model,will lose the ability to learn information connected far away.Therefore,the Long and Short Time Memory Network(LSTM)is introduced for load prediction.In order to further improve the prediction accuracy,the autoregressive differential moving average model(ARIMA)was used to correct the prediction errors,and the ARIMA(p,d,q)model with the highest prediction accuracy was selected through permutation and combination.3.The load prediction model of EEMD-LSTM-ARIMA was used to simulate the load data of Queensland,Australia,and the prediction results were tested by using root mean square error and mean absolute relative error,and then compared with the results obtained by EEMD-LSTM model and BP neural network.It is found that the results obtained by this model have higher accuracy.Finally,the load data of different months in the region are selected to train the model and then predict it.The results show that the error of the prediction results obtained from different data is not large,which proves that the model has a certain stability.