Research On Short-term Load Forecasting Method Based On Jumping Dual Strategy Drosophila Optimization Algorithm

High-precision short-term load forecasting can bring scientific and effective power dispatching scheme to the power grid,so as to achieve power supply and power consumption balance.It will minimize the loss caused by uneven distribution and reduce the operation cost,which can bring stable economic gain to the society.Due to the existence of various influencing factors,such as time,weather and electricity price,load forecasting is a complex nonlinear problem.Short-term load forecasting is to predict the trend of future load based on the change rule of historical load data and its influencing factors.The accuracy of traditional prediction method is not ideal,and it is only suitable for coincidence load series with strong periodicity.According to the present research status,many scholars use the regression characteristics of support vector machines to construct prediction models and obtain good results.However,there are two problems in the prediction method of standard support vector regression machine.The model precision can be directly affected by the model parameter selection and the model input.In this paper,leap double strategies fruit fly optimization algorithm is proposed to realize automatic selection of SVM parameters,and the optimal variables of prediction model are selected by feature selection.And then build a short-term load forecasting model.The work of this paper are as follows:1.Aiming at the two deficiencies of drosophila optimization algorithm,such as poor population diversity and local best,this paper proposes leap double strategies fruit fly optimization algorithm(LDFOA).Two different strategies are introduced to increase the population diversity of drosophila optimization algorithm and leap factor is added to enhance its ability of escape from local optimal.The test results of 15 classical functions show that the improved scheme can improve the performance of the algorithm.2.This paper takes the leap double strategies fruit fly optimization algorithm as the core search engine of feature selection,and proposes an feature selection method based on leap double strategies fruit fly optimization algorithm(LDFOAFS).The leap double strategies fruit fly optimization algorithm is used to selected the training set according to the correlation,and the selected subset is evaluated by the evaluation function.Finally,the optimal subset is obtained.12 UCI data sets are used to test it,and the results showed that feature selection could improve the effect of data classification,and the classification performance and subset selection ability of an feature selection method based on leap double strategies fruit fly optimization algorithm are better.3.This paper proposed a support vector regression prediction model based on feature selection and leap double strategies fruit fly optimization algorithm.The parameter optimization of support vector regression machine was realized by leap double strategies fruit fly optimization algorithm,and the optimal variables of prediction model are selected by feature selection.Taking the load of New South Wales from May 4 to May 13,2007 as the research object,the experimental results show that the performance of the proposed prediction model is good.