当前电力行业中高压断路器的在线感知数据积累严重不足,处理难度较大,并且故障试验成本高昂且危险性高,现场故障时设置的故障单一,获得的数据具有局限性,这些问题导致高压断路器的机器学习算法的置信度低,难以实际应用到设备的在线监测...当前电力行业中高压断路器的在线感知数据积累严重不足,处理难度较大,并且故障试验成本高昂且危险性高,现场故障时设置的故障单一,获得的数据具有局限性,这些问题导致高压断路器的机器学习算法的置信度低,难以实际应用到设备的在线监测中。本文通过Matlab搭建故障模拟平台,可根据需要调整断路器操动机构的各个参数,模拟断路器在不同运行状态下的各种机电故障,用软件提取断路器分合闸线圈电流信号和动铁芯直线行程位移信号的时域特征,并对混杂着各种故障状态下的合闸、分闸以及拒动过程训练数据进行处理研究,导出故障诊断模型训练规范的数据文件,有效地解决了现有数据量不足的问题,为构建断路器人工智能故障诊断模型提供数据基础。The accumulation of online perception data for high-voltage circuit breakers in the current power industry is seriously insufficient, and the processing difficulty is high. In addition, the cost of fault testing is high and the danger is high. The single fault set during on-site faults has limitations in the data obtained. These problems result in low confidence of machine learning algorithms for high-voltage circuit breakers, making it difficult to apply them in practical online monitoring of equipment. This paper builds a fault simulation platform through Matlab, adjusts the parameters of circuit breaker operating mechanism according to needs, simulates various electromechanical faults of circuit breaker under different operating states, and uses software to extract the time domain characteristics of circuit breaker switching coil current signal and moving core linear stroke displacement signal. And the training data of the closing and opening processes under various fault states, as well as the refusal to operate, are processed and studied, and the data file for training the fault diagnosis model is extracted, effectively solving the problem of insufficient data, providing a data basis for building an artificial intelligence fault diagnosis model for circuit breakers.展开更多
文摘当前电力行业中高压断路器的在线感知数据积累严重不足,处理难度较大,并且故障试验成本高昂且危险性高,现场故障时设置的故障单一,获得的数据具有局限性,这些问题导致高压断路器的机器学习算法的置信度低,难以实际应用到设备的在线监测中。本文通过Matlab搭建故障模拟平台,可根据需要调整断路器操动机构的各个参数,模拟断路器在不同运行状态下的各种机电故障,用软件提取断路器分合闸线圈电流信号和动铁芯直线行程位移信号的时域特征,并对混杂着各种故障状态下的合闸、分闸以及拒动过程训练数据进行处理研究,导出故障诊断模型训练规范的数据文件,有效地解决了现有数据量不足的问题,为构建断路器人工智能故障诊断模型提供数据基础。The accumulation of online perception data for high-voltage circuit breakers in the current power industry is seriously insufficient, and the processing difficulty is high. In addition, the cost of fault testing is high and the danger is high. The single fault set during on-site faults has limitations in the data obtained. These problems result in low confidence of machine learning algorithms for high-voltage circuit breakers, making it difficult to apply them in practical online monitoring of equipment. This paper builds a fault simulation platform through Matlab, adjusts the parameters of circuit breaker operating mechanism according to needs, simulates various electromechanical faults of circuit breaker under different operating states, and uses software to extract the time domain characteristics of circuit breaker switching coil current signal and moving core linear stroke displacement signal. And the training data of the closing and opening processes under various fault states, as well as the refusal to operate, are processed and studied, and the data file for training the fault diagnosis model is extracted, effectively solving the problem of insufficient data, providing a data basis for building an artificial intelligence fault diagnosis model for circuit breakers.
