Research On Energy Consumption Prediction Of Forging Line Based On Support Vector Machine

Industrial energy consumption accounts for 70%of China's energy consumption structure.For the increasing severe situation of energy demand,energy conservation in industry is not only the requirements of the national " 13th Five-Year" energy conservation and emission reduction planning,it is the inevitable choice of any industrial enterprise to improve the comprehensive strength.Energy consumption modeling and prediction of production line is an important research direction of energy saving in industrial enterprises,As the Machine learning algorithms become more mature,it offers a way to build an energy consumption model with good performance.First the paper introduces the related theories of production line system and energy consumption prediction,according to the features of modern manufacturing system,the research methods of energy consumption in equipment layer,process layer and system layer are summarized.Based on the application of energy consumption prediction in different fields,the whole energy consumption prediction modeling framework of the forging line is designed.Considering practical energy consumption of a certain company,based on the analysis of four aspects of process flow,production scheduling,equipment status and logistics,the paper selects 11 influence factors of the forging line energy consumption.The data of energy consumption and related factors of the enterprise in 2015-2016 were collected,and 7 main influencing factors of energy consumption of the production line were screened out by removing the outlier sample data,the secondary sample data and the redundant sample data.And then the mathematical models of energy consumption of 8 kinds of products in the forging production line are established by multiple regression method.There are various factors of energy consumption in forging production line,and there are also many unstable factors in the production process.In order to solve the problem,Machine learning algorithms can be used to simulate high dimensional and complex systems.In this paper,support vector machine had been used and through many experiments,it is confirmed that RBF is the optimal kernel function,the method of trial and error and 5 fold cross validation are used to analyze the penalty parameter C and the kernel parameter g of the support vector machine.Most of the energy consumption of forging line comes from the medium frequency heating furnace,leave-one-out method is used to design SVM energy consumption predicting model.The global optimization ability of particle swarm optimization algorithm is improved by combining the self-adaptive mutation and the cross validation.In this way,the estimated performance of the forging line energy consumption model is improved.The practicability and advantage of the method are verified by the comprehensive evaluation results of different models.