随着矿区整体开采设备自动化水平逐步提升，大量非线性电力电子设备的应用对煤矿供电系统的稳定性与安全性造成严重的影响。对煤矿供电系统中发生的电能质量扰动成因追本溯源，将繁杂的电能质量问题分离出本质特征，并进行识别分类，可为后期电能质量问题的解决以及整个系统经济运行的可靠性与安全性提供有效的信息与多样的方案思路。本文针对煤矿供电系统特定工况环境的电能质量扰动检测与识别分类方法进行研究。

首先，本文对煤矿供电系统结构中主要负荷设备工况进行分析，结合实际环境确定存在的14种单一、复合电能质量扰动类型。根据数学模型在Matlab中对扰动信号进行模拟仿真并生成相关电气数据。

其次，对比分析周期滑动有效值法(RMS)、卡尔曼滤波(KF)及无迹卡尔曼滤波(UKF)三种检测方法性能，选用UKF检测处理14种扰动类型；筛选最优特征指标后，选取基波、3次谐波、5次谐波、7次谐波以及振荡幅值的最小值和峰值指标作为最优特征向量集成为分类器的输入；比较KF、UKF与神经网络(BPNN)、极限学习机(ELM)、核极限学习机(KELM)两两组合的检测分类法效果，仿真结果表明，采用UKF-KELM对14种扰动类型识别平均成功率为98.14%，具有较强的抗噪性与稳定性。

最后，对电能质量扰动检测识别搭建实验平台，利用微机继电保护测试装置模拟煤矿供电系统扰动信号，用以STM32H743IIT6芯片为核心的信号采集传输单元将扰动数据传输至上位机，采用UKF-KELM检测分类法对扰动信号跟踪检测分类，最终识别出电能质量扰动类型。实验结果表明，本文采用的检测识别方法识别成功率达98%，具有一定的有效性。

With the gradual promoting of the automation of integral mining equipment, it have seriously affected the stability and safety of the coal mine power supply system that large numbers of nonlinear power electronic devices. Tracing the causes of power quality disturbances in the coal mine power supply system, separating the essential characteristics from the complex power quality problems, then it can be used to solve the problem of power quality in the later period that identifying and classifying characteristic information. It provides effective information and various scheme ideas for the reliability and security of the whole economic operation of system meanwhile. The detection, identification and classification methods of power quality disturbance in specific working conditions of coal mine power supply system are studied in this paper.

Firstly, the working conditions of primary load equipment in the structure of coal mine power supply system is analyzed, and 14 kinds of single and compound power quality disturbance types are determined combining with the actual environment. According to the mathematical model, the disturbance signal is simulated in Matlab and the relevant electrical data is generated.

Secondly, after comparing and analyzing the performance of the three detection methods of a cycle of RMS, Kalman filter (KF) and unscented Kalman filter (UKF), UKF is selected to detect and process 14 types of disturbances. After screening the optimal feature index, the minimum and peak indexes of fundamental wave, third harmonic, fifth harmonic, seventh harmonic and oscillation amplitude are selected as the input of the classifier. The detection classification effect of KF and UKF with the combination of neural network (BPNN), extreme learning machine (ELM) and kernel extreme learning machine (KELM) are compared. The simulation results show that the average success rate of UKF-KELM in identifying 14 disturbance types is 98.14%, which has strong anti-noise and stability.

Finally, the experimental platform is built for the detection and identification of power quality disturbances. The disturbance signal of coal mine power supply system is simulated by microcomputer relay protection testing device. The signal acquisition and transmission unit based on STM32H743IIT6 chip transmits the disturbance data to the upper computer. The UKF-KELM detection classification method is used to track and classify the disturbance signal, and ultimately the power quality disturbance type result is identified. The experimental results show that the recognition success rate of the detection and recognition method adopted in this paper is 98%, which has certain effectiveness.