基于神经网络响应面模型的有载分接开关弹簧储能故障的识别

An identification method for spring energy storage fault of on-load tap changer based on neural network response surface model

为有效识别有载分接开关的弹簧储能故障,提出一种基于神经网络响应面模型的有载分接开关弹簧储能故障的识别方法。首先,采用有限元法建立有载分接开关的故障仿真模型;然后,基于仿真试验和均匀试验设计生成响应面模型的训练样本,通过样本训练构建神经网络响应面模型;最后,采用意愿函数构造的多目标识别算法对表征弹簧储能不足的力学参数进行识别,通过仿真对UCL型有载分接开关弹簧储能不足故障的识别结果进行验证。研究表明,基于神经网络响应面模型能够有效识别弹簧储能不足故障,识别结果与参考值的最大相对误差为3.93%,验证该方法的有效性。

In order to precisely identify the spring energy storage failure in an on-load tap changer(OLTC), an identification method is developed for spring energy storage failure of the OLTC based on the neural network response surface model. Firstly, the fault simulation model of the OLTC was established through the finite element method. Then, the training samples of the response surface model were generated from the uniform experiments and simulations, and the neural network response surface model was therefore constructed by training these samples. Finally, the mechanical parameters of the spring energy storage deficiency were identified using the multi-objective identification algorithm constructed by desirability function, and the identification results of spring insufficient energy storage faults of the UCL type OLTC was validated by simulation. The Results show that the fault of spring insufficient energy storage can be identified accuratelyvia the neural network response surface model. The maximum relative error between the identified result and the reference data is 3.93%, which can verifie the effectiveness of this method.