基于加速退化数据和现场实测退化数据的电机绝缘剩余寿命预测模型

doi:10.19595/j.cnki.1000-6753.tces.221227

Abstract
Figure/Table
References (32)
Related Citation (15)

Download: PDF (2322 KB) HTML (1 KB)
Export: BibTeX | EndNote (RIS)

Abstract

Insulation system is the weakest part of motor reliability. Monitoring its condition and realizing accurate remaining life prediction is an effective means to ensure the reliability and safety of motor operation. Aiming at the disadvantages of mainstream remaining useful life (RUL) prediction models, including machining learning model, stochastic process model and stochastic filtering model, a motor insulation RUL prediction model based on accelerating degradation data and field state monitoring data under thermal stress, which combines extended Kalman filtering (EKF) with support vector machine(SVM) model and stochastic process model, is proposed. This model is mainly oriented to the RUL prediction problem of motor main insulation with thermal aging as the main failure mode. First, the Arrhenius model is used as the acceleration model, and the mapping relationship between the thermal stress level and the Wiener model drift coefficient and diffusion coefficient is constructed based on the Wiener process. Taking the residual breakdown voltage as the state variable, a prediction model of motor insulation life based on accelerated degradation data under actual working conditions is established, and it is used as the state equation of the Kalman filter model. Secondly, the expression of the maximum partial discharge is deduced by the breakdown voltage estimation equation, and the observation equation of the Kalman filter model is constructed based on the accelerated degradation data of the maximum partial discharge and on-site monitoring data; Then, in order to solve the problem of insufficient prediction accuracy caused by the inability to obtain new monitoring information and the inability to update the covariance matrix of the EKF model in life prediction, this paper takes time as the input variable and the maximum partial discharge as the input variable. Based on the support vector machine, a prediction model of the maximum partial discharge is established to realize the continuous update of the covariance matrix. Finally, for the 6650 polyimide film commonly used in motors, an accelerated degradation test is designed, and the test data of insulation resistance, insulation capacitance, dielectric loss tangent, maximum partial discharge and residual breakdown voltage with aging time are recorded. Based on the accelerated degradation data at 290℃, 300℃, 310℃, and 320℃, the maximum likelihood estimation method is used to construct the state equation. The observation equation was constructed by fusing the accelerated degradation data and the partial discharge data of the material sample at 240℃. The model was verified based on the measured aging data of the sample at 240 °C for 60 hours. The results showed that the model prediction error was within 4%. Then, the prediction accuracy of the EKF model and the stochastic process model were compared, and the comparison results showed that the prediction accuracy of the Kalman filter model was higher, which verifies the effectiveness and engineering application value of the proposed model in improving the remaining life prediction accuracy.

Key words： Motor insulation life prediction Wiener process extended Kalman filter support vector machine

Received: 28 June 2022

PACS:

TM306

	Service

	E-mail this article
	Add to my bookshelf
	Add to citation manager
	E-mail Alert
	RSS
	Articles by authors
	Zhang Jian
	Zhang Qin
	Huang Xiaoyan
	Fang Youtong
	Tian Jie

Cite this article:

Zhang Jian,Zhang Qin,Huang Xiaoyan等. Motor Insulation Remaining Useful Life Prediction Method Based on Accelerating Degradation Data and Field Degradation Data[J]. Transactions of China Electrotechnical Society, 2023, 38(3): 599-609.

URL:

https://dgjsxb.ces-transaction.com/EN/10.19595/j.cnki.1000-6753.tces.221227 OR https://dgjsxb.ces-transaction.com/EN/Y2023/V38/I3/599