Study Of Short-term Power Load Forecasting Based On Combined EEMD-SSA Model

With the rapid development of the national economy,people have put forward higher requirements for the power system.Short-term power load forecasting has very important guiding significance in all work of the power system.Accurate load forecasting results can be used not only in the formulation of electricity prices,but also in the dispatch of power grids,and can also be used in the power generation planning and arrangement of power plants.In recent years,many excellent deep learning algorithms have been widely used in power grid load forecasting,but each algorithm has its own advantages and disadvantages.In order to make a clearer comparative analysis,this thesis first builds a variety of single algorithm prediction models based on some excellent intelligent algorithms,and draws conclusions through comparative analysis of examples: Back Propagation Neural Network(BPNN)and Long Short-term Memory Recurrent Network(LSTM)It has good forecasting effect in load forecasting work.In the power system,due to the influence of many factors,the power load data presents strong volatility and instability.These characteristics will greatly affect the short-term load forecasting of the power grid.In response to this problem,this paper adopts a data processing method based on ensemble empirical mode decomposition(EEMD),using EEMD algorithm to decompose the original data into several empirical mode components and a residual component to reduce the impact of data fluctuations in the forecast work.At the same time,in order to verify the advantages of the EEMD algorithm in the load forecasting work,two prediction models of EEMD-BPNN and EEMD-LSTM were built,and the comparison and analysis of examples showed that after the data is decomposed by EEMD,the prediction effect of the model has been significantly improved.In order to integrate the advantages of different models in load forecasting,this thesis combines BPNN and LSTM on the basis of EEMD algorithm,and proposes a combined forecasting method based on EEMD.This method first decomposes the original data by EEMD to obtain several empirical mode(IMF)components,and then calculates the sample entropy of each IMF component sequence based on the sample entropy theory,and assigns different models according to the sample entropy for training prediction.At the same time,considering that the input of the sample features and the parameters of the neural network model will have an important impact on the overall prediction results,this thesis uses the salp swarm algorithm(SSA)to determine the number of neurons in each single model on the basis of the combined model.And the lagged terms of input variables are optimized,and the final EEMD-SSA combined forecasting model is obtained.In addition,ude the historical power load data in a southern city of china for comparative analysis.The experimental results show that the combined prediction model based on EEMD-SSA has a better prediction effect than a single prediction model and an unoptimized combined prediction model.