Engineering Science and Technology, an International Journal (Feb 2021)
Multiple linear regression-based impact analysis of impedance network design on life expectancy of DC-link capacitor in q-ZSI fed motor drive
Abstract
Most literature works carrying out capacitor remaining useful life (RUL) prognosis in a three-phase inverter fed Adjustable Speed Drive (ASD) system focus on analyzing the effect of different environmental and operating conditions. But it is the effect of inverter design parameters on capacitor RUL that is of greater interest to the inverter topology designers. This paper performs a parametric investigation on the impact of two design parameters primarily impedance (Z) network inductance (LZ), and shoot through duty ratio (Dsh) on capacitor RUL in a quasi-Z-source inverter (q-ZSI) fed ASD system. To perform this analysis, the online condition monitoring (OCM) of a capacitor placed in the Z network of q-ZSI topology is carried out, and its RUL is predicted for varying values of LZ and Dsh. The technique/procedure to monitor the operation of the system under study and collect its data for preventive maintenance without interrupting its operation is known as Online Condition Monitoring (OCM). The online method suggested in this paper uses data from the dynamic simulation (DS) model based on non-linear equations of q-ZSI. The use of the DS model increases the accuracy of OCM since it eliminates the use of data collected from sensors that are considered the cause of error in RUL prediction. A dataset of RUL predictions corresponding to multiple combinations ofLZ and Dsh is obtained through sensitivity analysis using Monte Carlo simulations, and the same dataset is used to derive a Multiple Linear Regression (MLR) model for RUL prediction. The sensitivity analysis data provides inferences on how the change in the selected design parameters affect the capacitor RUL in q-ZSI, and MLR model provides mathematical expression correlating LZ, Dsh and capacitor lifetime, with a 95% confidence level. The significance test statistics of the MLR model prove that LZ and Dsh are highly significant parameters that impact capacitor RUL in the system under study.
