为实现电气设备状态检修,提出了一种检测设备状态变化的建模与分析方法。首先,根据广义伏安特性,建立了基于电气量测信息的电压型和电流型电气设备端口模型,并通过偏最小二乘回归算法对模型参数进行辨识;其次,考虑环境和随机量测误差影响,使用非参数核估计方法得到检测时间窗口内设备模型参数的概率密度函数;最后,根据不同检测时间窗口内模型参数的概率密度分布差异,基于条件概率公式计算设备状态变化的概率,并埘电气设备状态变化进行告警。绕制了一单相双绕组变压器,并改变子绕组连接模拟匝间短路和绕组变形,测量端口电气信息,模型参数辨识结果表明:参数b1虚部b1i能够反映变压器状态变化;在3组实验中,由b1i计算得到的设备状态变化概率均大于门槛值,表明该方法具有较高准确性和灵敏度。
To realize condition-based maintenance for electric equipment, we proposed a new model and its related me- thod for detecting state changes of electric equipment. Firstly, according to generalized volt-ampere characteristic, we built voltage and current port models, in which the parameters were identified by the partial least squares regression algo- rithm based on measured voltage and current. Secondly, to estimate the effects of environment and random measurement error, we obtained a probability density function of the parameters using the non-parameter Kernel estimation method. Finally, according to the differences between probability distributions of the parameters in different time windows, the state change probability was obtained through conditional probability calculation, and a threshold was set to alarm the state change. Moreover, a single-phase two-winding transformer is made and tested with its turn-to-turn short circuit and winding deformation simulated through different connections of sub-windings. Based on the data obtained from the tests, the model parameters are identified. The results show that bli,namely, the imaginary part of bb is able to represent the condition changes, and the state change probability calculated using bli is larger than the threshold value in all three tests, which indicates the high accuracy and high sensitivity of the proposed method.