动态多子群QPSO算法及其在机车粘着优化控制中的应用

Dynamic multiple sub-population QPSO algorithm and its application in optimized adhesion control of locomotive

针对列车重载和高速运行时轮轨间粘着存在极限状态以及此时最优粘着利用能否获得的问题,利用所提出的动态多子群QPSO算法训练神经网络,并基于训练好的神经网络设计了机车粘着智能优化控制器,通过对电机转矩的动态调整,实现了轮轨间粘着的最优利用。仿真研究中,利用典型测试函数对所提出的动态多子群QPSO算法进行性能测试,证明该算法具有相对较高的寻优精度和效率,能有效提高神经网络的收敛速度和学习能力,将该算法应用于机车粘着优化控制中,得到了良好的控制效果。

Wheel-rail adhesion often reaches its limit states when heavy-haul and high-speed trains traveling. In order to obtain the maximum adhesive force utilization, this paper put forward a dynamic multiple sub-population QPSO algorithm to evolve a neural network, and designed the intelligent optimization controller based on the network to implement the optimized adhesion control of locomotive. By dynamically adjusting the motor torque, it achieved the optimal wheel-rail adhesion force. In simulation study, this paper used a typical test function to test the performance of the dynamic multiple sub-population QPSO algorithm. The simulation results demonstrate the relatively high accuracy and efficiency of the algorithm, and prove that the algorithm can improve the convergence speed and learning ability of the neural network, at the same time, in optimized adhesion control of locomotive, the intelligent optimization controller can also get a good control effect.