Application And Research Of GASA-BP Neural Network In Forecasting Industrial Electricity Consumption In Anhui Province

In recent years,with the national "12th Five-Year Plan" and "13th Five-Year Plan",the industrial economy of various places has developed rapidly.With the rapid development of industrial economy,the trend of industrial electricity consumption has been greatly increased,and non-renewable energy sources such as coal used for industrial power generation are also failing.If the contradiction between industrial electricity consumption and industrial economic development can not be solved properly,it will lead to a series of social problems.To study the relationship between industrial economy and industrial electricity consumption is helpful to promote social progress and development.If the industrial electricity consumption can be predicted scientifically and reasonably by the relevant indexes of industrial economy,it can not only help to promote the development of local industrial economy,but also make rational use of electric energy and save energy.In this paper,the principles of BP neural network,genetic algorithm(GA)and simulated annealing algorithm(SA)are introduced.The prediction model of BP neural network and the prediction model of GA-BP neural network are introduced,and the prediction model of GASA-BP neural network is put forward.Then,the prediction models based on BP neural network,GA-BP neural network and GASA-BP neural network are studied by using the relevant data indexes of Anhui industrial economy and the monthly data samples of industrial electricity consumption from June 2012 to April 2017.In addition,according to the selection of the related parameters of the three prediction models,the related experiments are designed.The forecasting effect of three forecasting models on the sample of industrial electricity consumption data in Anhui Province is compared and analyzed.In the BP neural network prediction model,four groups of experiments were designed to determine the optimal number of hidden layer neurons.The optimal number of neurons in the hidden layer is determined by comparative analysis.In the prediction model of GA-BP neural network,the population size of genetic algorithm is determined by comparing and analyzing the error of experiment.In the prediction model of GASA-BP neural network,two groups of experiments are designed for the initial temperature and the cooling rate.Finally,the optimum initial temperature and cooling rate are determined by comparing and analyzing the experimental errors.The prediction results of BP GA-BP and GASA-BP neural network are comparedand analyzed.The experimental results show that GASA-BP neural network inherits the characteristics of GA-BP neural network and can jump out of the local region by simulated annealing algorithm when it falls into local optimum.Then a global search for the optimal solution is carried out.Through the improvement of GA-BP neural network,the accuracy of GASA-BP neural network in predicting industrial electricity consumption in Anhui province is improved,and the weight and threshold of BP neural network are optimized better and the accuracy of prediction is improved.