The objective of this work is to select a Machine Learning Technique (MLT) to develop a fault diagnosis scheme for the power switching devices of the High Frequency link (HF link) in a grid-tied Photovoltaic (PV) system, without increasing the total number of sensors, and being capable to operate online. Artificial Neural Network (ANN), Support Vector Machine (SVM), Probabilistic Neural Network (PNN) and Naive Bayes (NB) algorithms are considered to solve the problem of fault classification. These four MLTs are compared using the specificity and sensitivity indexes. The inputs of the models are obtained from the mean value of the signals given by the Discrete Wavelet Transform (DWT) of the dc link voltage and the power extracted from the PV panels. Support vector machine algorithm is chosen as the most suitable classifier to diagnose single and simultaneous open circuit faults with lower computational effort. Simulation and real-time hardware-based experimental tests demonstrate that the MLTs are suitable and reliable to diagnose open circuit faults in a wide range of irradiance levels, ranging from 200 W/m2 to 1000W/m2, even under 6 % and 12 % measurement errors, without increasing the overall system cost.

An open circuit fault diagnosis scheme for the power switches of the output inverters in a cascaded H bridge multilevel converter is proposed in this research work, which is designed to be used as part of a large-scale photovoltaic generation system with a rating of 1 MW and 13.2 kV. A two stage classifier is the main core of the technique, in which two different approaches are combined, 1) a signal processing algorithm and 2) a machine learning method, having an artificial neural network classifier as its main model. The general fault diagnosis scheme is developed in a per-phase basis of the converter, using only three neural network classifiers for the three phase configuration. The method is suitable to diagnose open circuit faults under different irradiance conditions and parametric variations, and is specified to keep the same level of complexity even if the general structure of the topology is changed. The total sensor count is not incremented since the proposed algorithm uses the individual dc bus voltages and the output currents already required to execute the control system. The performance of the fault diagnosis scheme is tested at a simulation level using Matlab - Simulink / Simscape, considering a wide range of irradiance conditions, even under an irregular distribution of this parameter, and also assuming single open circuit faults in the power switches of one phase.

A fast detection algorithm for sags, swells and interruptions is developed in this paper. The proposed method is based on the combination of the digital RMS technique and the Kalman filter. The union of these methods allows using the main advantages of Kalman filter and RMS techniques and helps rejecting the extreme drawbacks of Kalman filter. The proposed approach can be applied to dynamic voltage restorers, dynamic sag correctors, uninterruptible power supplies, and other compensation topologies for power quality improvement

A fast detection algorithm for sags, swells and interruptions is developed in this paper. The proposed method is based on the combination of the digital RMS technique and the Kalman filter. The union of these methods allows using the main advantages of Kalman filter and RMS techniques and helps rejecting the extreme drawbacks of Kalman filter. The proposed approach can be applied to dynamic voltage restorers, dynamic sag correctors, uninterruptible power supplies, and other compensation topologies for power quality improvement