基于改进BP神经网络的轨道交通不间断电源设备故障检测系统设计

Design of Fault Detection System for Uninterruptible Power Supply Equipment in Rail Transit Based on Improved BP Neural Network

摘要:

轨道交通不间断电源设备的工作状态复杂多变,其中存在大量非线性关系,使得电源设备的工作状态变得更加复杂多变,不同因素之间相互影响,导致故障检测的难度增加。为保证轨道交通不间断电源设备的运行安全,设计并开发了电源设备故障检测系统。改装电源工作温度、电流/电压传感器,调整信号采集器、转换器和处理器的内部组成结构,实现硬件系统的优化。从物理和逻辑两个方面,构建系统数据库,为数据存储提供空间。模拟轨道交通不间断电源设备工作过程,确定设备故障的检测标准。利用硬件系统中的传感器设备,采集电源设备的工作数据,利用改进BP神经网络提取电源设备工作特征。采用特征匹配的方式得出轨道交通不间断电源设备的故障类型检测结果,以可视化的形式输出故障检测结果。通过系统测试实验得出结论:与传统故障检测系统相比,优化设计系统的故障类型错误检测率降低约30%,故障量检测误差减小约4.1KWH。

Abstract:

The working status of uninterrupted power supply equipment in rail transit is complex and variable, with a large number of nonlinear relationships, making the working status of power supply equipment more complex and variable. Different factors interact with each other, leading to an increase in the difficulty of fault detection. To ensure the safe operation of uninterrupted power supply equipment in rail transit, a power supply equipment fault detection system has been designed and developed. Modify the working temperature, current/voltage sensors of the power supply, adjust the internal composition structure of the signal collector, converter, and processor, and achieve hardware system optimization. Build a system database from both physical and logical aspects to provide space for data storage. Simulate the working process of uninterrupted power supply equipment in rail transit and determine the detection standards for equipment failures. Utilize sensor devices in the hardware system to collect working data of power supply devices, and extract working characteristics of power supply devices using an improved BP neural network. Using feature matching to obtain the fault type detection results of uninterrupted power supply equipment in rail transit, and outputting the fault detection results in a visual form. Through system testing experiments, it has been concluded that compared with traditional fault detection systems, the optimized design system reduces the error detection rate of fault types by about 30% and the detection error of fault quantity by about 4.1KWH.