GA Fuzzy Neural Network Based Automatic Train Operation Optimization

Automatic train operation system not only can realize automatic drive and reduce thedriver’s working load, but also can ensure the safe operation of the train on the basis ofimproving train of various performance indicators, so the automatic train operation systemhas become a hot research topic. The traditional PID algorithm is good for linear system, anddoes not have the adaptability, train operation process has the characteristics of complex,multi-objective, nonlinear, a precise mathematical model is hard to establish, so the trainautomatic driving system which based on the traditional PID algorithm is in the real line, oncethe influence by the line and the environment and so on，the control effect will be verydififcult to achieve satisfactory results. GA fuzzy neural network algorithm is the combinationof fuzzy logic and neural network, then lets genetic algorithm learn the weight of BP neuralnetwork. So not only the complementary of the three methods in the aspects of knowledgeacquisition, reasoning, learning ability, but also make extensive mapping ability of neuralnetwork and has fast convergence and reinforcement learning ability. This algorithm not onlyimproves the control precision of the system and reduces the error rate in the process ofcontrolled object, but also does not have high requirements on the control object modelaccuracy. Thus the GA fuzzy neural network algorithm is applied to automatic train operationsystem, and the line of inlfuence factors on the train with compensation function of fuzzycontrol to cancel, the simulation results show that the GA fuzzy neural network algorithm isapplied to automatic train operation system is able to achieve good control effect.Firstly, This thesis introduces the background, signiifcance, the domestic and foreignresearch status of automatic train operation system, and explains the structure, function andthe ifve performance indicators of automatic train operation system, then studies automatictrain operation strategy. Secondly，introduces the train movement model and line model andbuild a train target curve, then studies the GA fuzzy neural network algorithm detailed, anddesigns speed controller based on GA fuzzy neural network algorithm. Finally, in theMATLAB simulation software platform, simulation of them which compares to the controleffects respectively based on traditional PID algorithm and the GA fuzzy neural networkalgorithm, Comparing the simulation result, the control effect of fuzzy neural network speedcontroller is signiifcantly superior to the traditional PID algorithm, which proves thatintroduce GA fuzzy neural network algorithm in the train operation system feasibility andeffectiveness.