Research On Short-term Power Load Forecasting Based On Improved Sparrow Search Algorithm To Optimize Long Short-term Memory Network

In order to further improve the construction of power marketization and improve the market competitiveness of power grid companies,short-term power load forecasting plays a key role in the planning and maintenance of power grids.Therefore,it is required to conduct more in-depth research and exploration on short-term power load forecasting.Perform corresponding analysis and processing on sample data,and correct abnormal data.In load forecasting,the dimensions of different influencing factors must also be considered.The difference in dimensions also has a certain impact on the final forecast results.Therefore,the sample data is normalized to eliminate the short-term power load of different dimensions.The impact of forecast results.When performing load forecasting,the least squares support vector machine(LSSVM)and long short-term memory(LSTM)neural network models have the following shortcomings: the key parameters are mainly selected based on the researcher’s experience.In order to solve this problem,Sparrow Search Algorithm(SSA)is introduced to optimize its key parameters and find the optimal model parameters.However,in the process of optimizing SSA,there are still the shortcomings that it is easy to fall into the local optimum,and the population lacks diversity in the later stage of the iteration.Therefore,SSA is improved accordingly,the population is initialized by Sin chaotic map,and the location update idea of the discoverer and the joiner in the bird search algorithm(BSA)is applied to the location update of the discoverer and the joiner in the SSA.On the basis of the foregoing,the introduction of Cauchy mutation and reverse learning strategy to improve it to obtain an improved sparrow search algorithm(ISSA).A simulation comparison experiment based on the benchmark test function is designed to compare the optimization performance of ISSA and SSA.The results show that the convergence speed and convergence accuracy of ISSA are better than SSA.In order to improve the prediction accuracy,this paper proposes the ISSA-LSSVM model and optimizes the parameters of the LSSVM model to obtain a better set of parameters in the model.By designing a simulation comparison experiment,ISSALSSVM is compared with LSSVM and SSA-LSSVM,and the average absolute percentage error(MAPE)is used as the evaluation standard.The results show that the model has high prediction accuracy.In order to further improve the prediction accuracy,on the basis of the previous,the ISSA-LSTM model is proposed,and the model is compared with the three models of LSTM,SSA-LSTM and ISSA-LSSVM.The results show that the ISSA-LSTM model has better performance than the other three models.Higher prediction accuracy.