矿用功率变换器微小故障演化预测方法研究

The reliable operation of the equipment of coal mining and safety monitoring can be directly threated by the failure of mine power converter. For the problem of the insensitivity of characteristic parameters of mine power converter coupled with signals and variable working conditions, and the complexity of small faults evolution under harsh environments, the project deeply analyzes the fault signals’ features, decouples circuit monitoring signals based on the technique of nonlinear blind source separation. The impact of the working condition can be eliminated through putting forward the equivalent analysis method and determines the circuit-level characteristic parameters which is only sensitive to the faults under the variable working conditions, then the laws that reflect the small fault evolution of mine power converter are uncovered. A variety of prediction algorithms are optimized, and the error feedback calibration technique can be adapted and declines the accumulation of errors. The multi-parameter self-tuning prediction algorithm is proposed based on the error feedback, and the scheme of small fault evolution of mine power converter method for high-precision prediction is presented, then mine power converter accurately track and small fault with the precision of prediction can be achieved. The study provides the necessary theoretical basis and technical support for fault prognosis and health management of mining electrical and electronic equipment, which is of great significance for preventing and reducing the occurrence and expansion of coal mining accidents.

矿用功率变换器故障直接威胁煤矿生产及安全监控设备的可靠运行。针对矿用功率变换器信号、工况关联耦合下故障特征参数不敏感及恶劣环境下微小故障演化过程的复杂多变性问题，深入探析矿用功率变换器故障信号特点，利用非线性盲源分离解耦电路观测信号，提出当量分析法消除工况影响，确定复杂环境下仅对故障敏感的电路级故障特征参数，揭示矿用功率变换器微小故障演化规律；优化匹配多种预测模型，利用误差反馈校正方法降低误差积累，构造基于误差反馈的多参数自校正预测算法，给出矿用功率变换器微小故障演化高精度预测方法体系，实现矿用功率变换器微小故障演化的准确跟踪与高精度预测。项目研究为矿用电子电气设备的故障预测与健康管理提供必要的理论基础和技术支撑，对预防和减少煤矿生产事故的发生与扩大意义重大。

复杂环境下高可靠功率变换器蕴藏着巨大的潜在应用价值，工作于高温、高湿、强电磁干扰等复杂环境下的矿用功率变换器故障率高，直接威胁所在电气设备的可靠运行。针对复杂环境下功率变换器信号、工况关联耦合下故障特征参数不敏感及微小故障演化过程的多变性问题，开展了故障信号的预处理方法、多故障诊断方法及健康状况预测方法三方面关键技术及相关的理论知识的研究。给出了基于压缩感知的电力电子电路故障信号预处理方法，降低电力电子电路的信号数据量以实现在线快速故障诊断；研究了基于变分模态分解的单通道信号盲源分离方法；针对电力电子多软故障、多硬故障的故障特征表现较相似，难以进行正确诊断难题，从故障特征提取与故障分类方法两方面研究，给出了基于JADE-SAE、基于WPE-ELM的两种适用于电力电子电路的多故障诊断方法，取得较好的诊断效果；为消除工况变化对故障特征参数的耦合影响，分析了工况对不变工况下有效故障特征参数的影响规律，提出当量分析法，基于BPNN 等智能算法建立标准工况下的故障特征参数与实际工况、实际工况下故障特征参数的关系，将统一在标准工况下的故障特征参数当量值作为变工况下的电路级健康特征参数，所提取的健康特征参数不再受工况影响且仅与电路性能退化状况有关，能够反映电路的健康状况；拓展了LSTM网络在功率变换器故障预测领域的应用，验证了LSTM网络在功率变换器状态预测方面可行有效。项目研究给出了工作于煤矿等复杂环境下功率变换器健康状况预测方案，形成了一套适用于矿用功率变换器故障预测的技术成果。项目研究成果为优化功率变换器的可靠性设计、改善工作效率和性能、系统健康管理提供理论依据与技术支撑。