Calculation Model Of Fuel Consumption Of Vehicle Based On BP Neural Network Optimized By Genetic Algorithm

With gradual energy lack and the competition between transport enterprises getting fiercer and fiercer, the managers begin to pay close attention to the car fuel consumption. As a result, the demand of monitoring system focus on vehicle precise fuel consumption is becoming stronger and stronger. There are many researches on monitoring the fuel consumption of vehicle at home and abroad, but volume method is most popular in practical application. In the monitoring system designed based on volume method, the job of counting the oiling quantity of vehicle is one of the difficulties. By observing the monitoring data of vehicle in actual operation, mining the nonlinear relationship between actual oiling quantity of vehicle and monitoring data has been identified as the goal of the design of calculation model.The BP neural network proved outstanding in approximating nonlinear functions has been a research hotspot in recent years. But the standard BP neural network has many disadvantages, such as easy to fall into local optimum, slow speed of search and so on. Many researchers have proposed a variety of optimization scheme, such as genetic algorithm optimization, particle swarm optimization, etc. The author chose the genetic algorithm optimization scheme, by using genetic algorithm to optimize the initial weights and threshold of BP neural network. When designing the BP neural network model optimized by genetic algorithm, the author learned about that the genetic algorithm itself also has certain defects. Although the adaptive genetic algorithm put forward by Srinivas solved the problem brought by the fixed crossover rate and mutation rate, the new algorithm has shortcomings also. The search speed is slow in the initial stage of population evolution and the individual diversity of population is not good.Based on the study of Srinivas' s theory, this paper advances an improved adaptive genetic algorithm, in order to overcome the weakness and improve the performance of the algorithm. This paper improve the algorithm mainly from three aspects. Firstly, besides of the fitness of individual and population, the stage of population evolution is regarded as a major parameter to adjust the adaptive variables of genetic operations. In the initial stage of population evolution, the bigger adaptive variables can promote the fitness of population fast. Then with the population evolution going, the adaptive variables decrease step by step to protect excellent individuals from destroy. Secondly, the calculation method of adaptive crossover and mutation rates is improved, enhancing the individual diversity of population in some degree. Thirdly, the amount of mutation point is adaptive according to the fitness of individual, the fitness of population and the current stage of population evolution.Finally, through the analysis of the experimental data this paper verified that the improved adaptive genetic algorithm is superior to Srinivas' s adaptive genetic algorithm and standard genetic algorithm. By comparing the calculation errors of standard BP model, GA-BP model, AGA-BP model and IAGA-BP model, it further illustrates that the proposed improvement scheme can reduce the probability of BP neural network learning process into local optimal value probability of fall into local optimum of BP neural network sharply, give full play to the local search ability of BP neural network and improve the calculation accuracy and the generalization ability of the model.