发电机励磁碳刷运行故障检测方法

2022,30(6):53-58
刘刚, 赵健龙
浙江浙能嘉华发电有限公司
摘要:针对发电机励磁系统中碳刷结构故障检测困难,故障定位不准确的问题,本文根据实际碳刷运行过程建立碳刷结构检测系统。通过建立组件对象模型(Component Object Model COM)动态检测方案将碳刷结构模型化,使系统能够更为精确的检测到故障原因。设计数字式光纤传感器(BF5R)检测电路将碳刷故障过程图像化,缩短系统检测时间。通过改进维格纳威尔分布(Wigner Ville distribution WVD)故障定位算法精准定位碳刷故障位置,采用合理方式进行维修。通过Proteus软件仿真检测系统运行过程,实验表明本设计对碳刷故障检测具有明显效果,在15KW发电机环境中,碳刷故障定位时间为3.5min,信号幅值为13V,结果精确度为96.4%,证实了本设计的可行性;通过仿真对比三种不同系统信号检测幅值电压和检测准确度曲线,由此验证了本研究的优越性。
关键词:发电机励磁系统;碳刷故障检测;COM动态检测方案;BF5R检测电路;改进WVD算法

Fault Detection Method of Generator Excitation CarbonBrush Operation

Abstract:Aiming at the difficulty of detecting the carbon brush structure fault in the generator excitation system and the inaccurate fault location, this paper establishes a carbon brush structure detection system based on the actual carbon brush operation process. By establishing a Component Object Model COM (Component Object Model COM) dynamic detection program to model the carbon brush structure, the system can detect the cause of the failure more accurately. Design a digital optical fiber sensor (BF5R) detection circuit to visualize the carbon brush failure process and shorten the system detection time. By improving the Wigner Ville distribution WVD (Wigner Ville distribution WVD) fault location algorithm to accurately locate the carbon brush fault location, and use reasonable methods for maintenance. Through the Proteus software to simulate the operation process of the detection system, the experiment shows that this design has obvious effects on carbon brush fault detection. In a 15KW generator environment, the carbon brush fault location time is 3.5min, the signal amplitude is 13V, and the result accuracy is 96.4% , Confirmed the feasibility of the design; through simulation and comparison of three different system signal detection amplitude voltage and detection accuracy curve, thus verifying the superiority of this research.
Key words:Generator excitation system; Carbon brush fault detection; COM dynamic detection scheme; BF5R detection circuit; Improved WVD algorithm
收稿日期:2021-12-07
基金项目:
     下载PDF全文