Power Quality Improvement and Fault Diagnosis of PV System By Machine Learning Techniques

This paper employs the newly proposed time delayed switching filter paradigm for active power quality improvement for grid-connected Photovoltaic (PV) systems. Thereafter fault diagnosis scheme for the same system has been recommended using machine learning technique. The main novelty of this paper work can be enumerated as (i) Proposed time delayed switching filter paradigm for active power quality improvement and (ii) fault diagnosis scheme for the same system using machine learning technique which can speeds up fault detection time and detect the fault location 95-99% accurately. The Cascaded Hybrid Multilevel Inverter (CHMI) used here for core inverter comprises of number of switches which in turn, increases the power losses. The Kalman filter controller is utilized to predict the state and to improve power sharing injected by renewable energy resources. But in the practical case, it is also assumed that the measurement noise of the filter are not accurately known. To estimate the states properly under these proposed circumstances, this work suggests adaptive estimation based Kalman Filter. Again, due to the switching of MIs, the state equation of the system has been changed and time delayed is present in the output. This problem deals with to use of switching Time delayed Adaptive Kalman Filter (TAKF). To enhance the reliability, a fault diagnosis technique has been planned here for CHMI. This paper presents a Machine learning based fault diagnosis technique. The proposed scheme can diagnosis the continuous and intermittent faults for open circuit. The efficacy of the scheme, proposed here is authenticated by the simulation study of a PV system.