【作者】 张伟；

【导师】 孙曙光；

【作者基本信息】 河北工业大学 ， 工程硕士（专业学位）， 2021， 硕士

【摘要】 万能式断路器在低压配电系统中起着控制和保护的作用,它的机械寿命是影响其寿命的主要因素,通过开展断路器机械寿命预测的研究,对其状态进行在线监测,不仅能及时掌握其健康状态,同时也可以降低由断路器机械故障引起的事故率。大量试验证明,断路器动作过程中产生的振动信号包含了重要的状态信息,可以作为其机械状态的信息载体。基于此,本文研究万能式断路器寿命周期内的退化过程,利用振动信号对断路器的机械寿命进行在线预测评估。首先,对断路器的机械特性监测常规参数进行分析,面向在线寿命预测,提出利用断路器合、分闸时的振动信号实现断路器状态在线监测;在此基础上,通过构建断路器机械寿命试验系统,实现了对断路器动作过程的控制,以及测量信号的采集与存储,为断路器机械寿命预测研究提供了平台基础。其次,为了从高噪声背景下,准确提取动作过程中的振动信号,提出基于变分模态分解和奇异值分解的联合降噪方法。对联合降噪方法中关键参数进行研究,分别利用中心频率法确定变分模态分解中的模态个数,以及奇异值差分谱确定奇异值分解中的有效秩阶数。通过仿真信号对降噪效果分析可得,联合降噪后的信号SNR得到大幅提升,验证了降噪方法的有效性。然后,在对断路器的机械特性参数进行分析的基础上,提出基于振动检测在线提取断路器时间特性参数的方法。对振动信号进行联合降噪后,利用基于短时能量+双门限法从振动信号中在线提取分闸过程中的机构动作时间参数;并将该机构动作时间作为反映断路器机械状态退化的主要特征参数以及机械寿命的失效判据。最后,针对断路器机械寿命预测模型,提出基于粒子群优化的支持向量回归预测方法。在此,为了更加全面地反映断路器运行状态,进一步对所提取动作时间内的振动信号进行分析,获取关键时域指标作为动作时间参数的有效补充,从而构成支持向量回归寿命预测模型的多参量输入。断路器剩余机械寿命预测实验表明,相较于动作时间单参量输入,多参量预测具有更小的预测误差,同时与其他方法对比预测效果最佳。更多还原

【Abstract】 The conventional circuit breaker plays the role of control and protection in the low-voltage power distribution system.Mechanical life is the main factor which affects the circuit breaker’s life.By carrying out the research on the prediction of the circuit breaker’s mechanical life,and online monitoring of its condition,not only the health status can be obtained in time,but also the accident rate caused by the mechanical failure of the circuit breaker can be reduced.A large number of tests have proved that the vibration signal generated during the operation of the circuit breaker contains important status information and can be used as an information carrier for its mechanical status.Therefore,this paper studies the degradation process in the life cycle of the circuit breaker,and uses the vibration signal to online predict of the circuit breaker’s mechanical life.Firstly,the traditional parameters of the mechanical characteristics of the circuit breaker are analyzed.Aiming at the online prediction of the remaining life,it is proposed to use the vibration signal in the switching operation to realize the online monitoring of the circuit breaker status.On this basis,by constructing the mechanical life test system of the circuit breaker,the control of the action process of the circuit breaker and the collection of measurement signals are realized,which provides a platform foundation for the research on the mechanical life prediction of the circuit breaker.Secondly,in order to accurately extract the vibration signal during the action from the background of high noise,a joint denoising method based on variational modal decomposition and singular value decomposition is proposed.By studying the key parameters of the joint denoising method,the center frequency method is used to determine the number of modes in the variational modal decomposition,and the singular value difference spectrum is used to determine the rank order in the singular value decomposition.Through the analysis of the denoising effect of the simulated signal,the SNR of the signal after joint denoising has been greatly improved,which verifies the effectiveness of the denoising method.Then,based on the analysis of the mechanical characteristic parameters of the circuit breaker,a method of extracting the time characteristic parameters through vibration detection is proposed.After the joint denoising of the vibration signal,the short-term energy and double threshold method is used to extract the mechanism action time parameters in the opening process from the vibration signal online.The action time of the mechanism is taken as the main characteristic parameter reflecting the degradation of the circuit breaker’s mechanical status,and as the failure criterion of the mechanical life.Finally,for the mechanical life prediction model of the circuit breaker,a support vector regression prediction method based on particle swarm optimization is proposed.At the same time,in order to reflect the mechanical status of the circuit breaker comprehensively,the vibration signal during the extracted action time is further analyzed.The key time domain indicators are obtained as an effective supplement to the action time parameters,thereby the multi-parameter input of the support vector regression life prediction model has been formed.The prediction experiment of the remaining mechanical life shows that compared with the single parameter input of the action time,the multi-parameter prediction has a smaller prediction error,and the prediction model is the best compared with other models.更多还原