DPICSA优化的城轨交通超级电容FNN控制研究  

Study on FNN Control of Supercapacitor of Urban Rail Transit Optimized by DPICSA

作　　者：徐凯[1] 何周阳 徐文轩 吴仕勋[1] XU Kai;HE Zhouyang;XU Wenxuan;WU Shixun(School of Information Science and Engineering,Chongqing Jiaotong University,Chongqing 400074,China;School of Electrical Engineering,Chongqing University,Chongqing 400044,China)

机构地区：[1]重庆交通大学信息科学与工程学院,重庆400074 [2]重庆大学电气工程学院,重庆400044

出　　处：《铁道学报》2021年第10期43-52,共10页Journal of the China Railway Society

基　　金：重庆市教委科学技术研究项目(KJQN202000703);重庆市研究生教育教学改革重点项目(YJG172004);城市轨道交通车辆系统集成与控制重庆市重点实验室开放基金项目(CKLURTSIC-KFKT-201805)。

摘　　要：在城市轨道交通中超级电容储能系统用于吸收列车再生制动能量,以实现节能并抑制直流接触网的电压波动。在超级电容储能系统双闭环控制中,采用遗传算法优化的电压外环PI控制易引起输出振荡,而采用常规模糊控制又存在量化、比例因子整定困难,自适应性差且模糊推理速度慢等缺陷。针对上述问题,提出一种双种群免疫克隆选择算法(DPICSA)优化的城轨列车超级电容模糊神经网络(FNN)控制策略。采用DPICSA综合协调优化主模糊控制器的隶属度函数与量化、比例因子;在此基础上,设计一个模糊参数自校正器对量化、比例因子进行在线调节;采用两个RBF神经网络分别记忆主模糊推理与参数自校正模糊推理,利用神经网络高速并行分布式计算能力,加快模糊推理速度。通过3种不同场景下的仿真实验研究,验证了该策略在抑制网压波动和节能方面均优于遗传算法优化的PI控制和常规模糊控制。The utilization of a supercapacitor energy storage system(SCESS) to store regenerative braking energy in urban rail transit system effectively improves the energy saving rate and suppresses the voltage fluctuation of the DC catenary system. Based on the double closed-loop control strategy of SCESS, the outer-loop voltage PI controller optimized by genetic algorithm may cause the output oscillation, while the conventional fuzzy controller has the problems of quantification factor, scaling factor setting difficulty, poor adaptability and slow fuzzy inference speed. In response to the above-mentioned problems, a fuzzy neural network(FNN) control strategy was proposed, which was optimized by double-population immune clonal selection algorithm(DPICSA). First, the membership functions and initial values of the quantification factor and scaling factor of the main fuzzy controller were synthetically optimized by DPICSA. Then, a parameter self-correction fuzzy controller was designed to adjust these initial values online. Eventually, two RBF neural networks were used respectively to memorize the fuzzy inference of main controller and parameter self-correction controller. The fuzzy inference speed was accelerated by the parallel and distributed computing capabilities of the neural networks. The simulations conducted in three different scenarios proved better voltage stabilization and energy-saving effects of the proposed control strategy, compared with the GA-PI control and the conventional fuzzy control.

关 键 词：城轨交通 超级电容储能系统 控制策略 双种群免疫克隆 模糊神经网络 

分 类 号：TM53[电气工程—电器]