Comparison of Kalman Filter and Least Squares Regression-based RUL Estimation of Capacitors in Variable Speed Drives

Variable speed drives (VSD) are widely used in many industrial and consumer applications due to their ability to control electric motors efficiently. Failure of these variable speed drives may adversely affect the production in industries and comfort for consumers in applications like electric vehicles (EV) and air conditioning and mechanical ventilation (ACMV). As the capacitors are one of the most failure-prone components in the variable speed drives, condition monitoring systems for capacitors are necessary to ensure reliable operation. In this paper, the remaining useful life (RUL) estimation methods for capacitors are analysed based on their accuracy and computation complexity using the data from accelerated ageing tests. The analysis showed that the computational complexity of Kalman filter based method is lower compared to least squares regression. However, it is also shown that in applications where the degradation model is non-linear and there is no prior knowledge regarding the model parameters, the least squares regression provides better RUL estimates than the extended Kalman filter.