基于预充电模型与改进粒子群算法的直流支撑电容器参数辨识方法  

Capacitance estimation method for DC-link capacitors based on pre-charging model and improved PSO algorithm

作　　者：孙潞 吴宜霖 伍珣 SUN Lu;WU Yilin;WU Xun(School of Traffic and Transportation Engineering,Central South University,Changsha 410075,China)

机构地区：[1]中南大学交通运输工程学院,湖南长沙410075

出　　处：《铁道科学与工程学报》2025年第3期1243-1253,共11页Journal of Railway Science and Engineering

基　　金：国家自然科学基金资助项目(52202428)。

摘　　要：直流支撑电容器是变流器的关键部件之一。由于温度、湿度、电压以及谐波等因素影响,其老化速度加快,不利于变流器的可靠运行。因此有必要对直流支撑电容器进行状态辨识,根据其健康状态选择不同的维护策略,进一步保障系统可靠性。目前已有较多学者对电容器状态辨识开展研究。但是,噪声对直流支撑电容器状态辨识的干扰问题尚未得到良好解决,容易导致显著的电容器参数辨识误差。针对上述问题,提出一种基于预充电模型与改进粒子群算法的直流支撑电容器状态辨识方法。在不加装额外传感器的前提下,建立等效预充电模型,采集预充电过程的电压信号,利用改进粒子群算法在参数变化区间内搜索最优值,满足其损失函数值最小,进而对电容量进行计算。该方法原理简单,计算量小,可以有效避免算法收敛于局部最优解,在一定程度上解决了噪声干扰的问题。使用Matlab/Simulink平台进行仿真实验,结果表明该方法在较低采样频率下对噪声干扰下的电容状态辨识有较高精度,当采样频率为100 Hz和50 Hz时辨识误差少于1.5%,采样频率为20 Hz和10 Hz时少于3%。此外,该方法具有较好的抗噪声性能,在噪声方差σ2=3时辨识误差仍少于5%。随后,使用3组地铁实车数据验证了算法的可靠性,每组数据连续进行2次辨识,电容量辨识误差均少于1.8%。DC-link capacitor is one of the key components of converters.However,due to factors such as temperature,humidity,voltage,and harmonics,their aging rate accelerates,which is detrimental to the reliable operation of the converter.Therefore,it is necessary to perform parameter evaluation on DC-link capacitors and choose different maintenance strategies according to their health status to further ensure the reliability of the systems.Currently,many scholars have conducted work on parameter evaluation of capacitors.However,the problem of noise interference in the capacitance estimation of DC-link capacitors has not been well resolved,which can easily lead to significant estimation errors of capacitor parameters.To address the above issue,a method for DC-link capacitor parameter evaluation based on pre-charging model and improved particle swarm optimization was proposed.Without installing additional sensors,an equivalent pre-charging model was established,and the voltage signal during the pre-charging process was collected.The improved particle swarm optimization was used to search for the optimal value within the parameter range,satisfying the minimization of its loss function,and then calculate the capacitance.This method had the advantages of simple principle,small computational load,and can effectively avoid the algorithm converging to local optimal solutions,thereby solving the problem of noise interference to some extent.Simulation experiments were conducted using the Matlab/Simulink platform.The results show that the method has high accuracy in capacitor parameter evaluation under low sampling frequency.When the sampling frequency is 100 Hz and 50 Hz,the evaluation error is less than 1.5%,and when the sampling frequency is 20 Hz and 10 Hz,it is less than 3%.In addition,the method has good noise resistance,and the evaluation error is still less than 5% when the noise variance is equal to 3.Subsequently,the reliability of the algorithm is verified using real vehicle data from three sets of subway experiments,with

关 键 词：电容器 状态识别 预充电模型 粒子群算法 牵引变流器 

分 类 号：TM464[电气工程—电器]