Intelligent Health Monitoring of Capacitor Using Reduced Experimental Input Data

Usage of the electrolytic capacitor is a must for all power electronic converters. The reliability of electrolytic capacitors is critical for the product to be more reliable. In the past, there have been significant steps in the academic study dedicated to the condition monitoring of electrolytic capacitors to predict the state of capacitor health. However, preventive maintenance may be the required feature for industrial applications to have reliable products. However, the implementation of evolved state monitoring methods is limited in industrial applications due to the cost constraints, difficulty, and other operational problems. Thus, a summary of the past study on the condition monitoring is required to explain the needed resources to implement and understand each critical technique's performance. Therefore, this paper has presented the study of capacitor condition monitoring and proposed an artificial neural network (ANN) based capacitance condition monitoring system for estimating the capacitance. The training data required for ANN is obtained through an experimental setup. The ANN is generated by training the network with experimentally measured data in MATLAB to estimate the LLC converter input DC-link capacitor value. Training and estimation accuracy are analyzed at different loading combinations.