| Work Detail |
Indian scientists have proposed using a multi-layer neural network to detect line-to-ground, line-to-line and bypass diode faults in strings of photovoltaic modules. They tested the new method on a 22.5 kW solar field and obtained “competitive” precision results.
A group of researchers from India has proposed a novel fault detection technique in photovoltaic strings that uses multilayer neural networks (MLNN), a machine learning technique that can handle complex relationships by learning their hierarchical representations.
“With the help of the proposed technique, line-to-ground (LG) and line-to-line (LL) faults are detected, classified and localized,” the researchers explain. “The proposed MLNN technique only requires the installation of a current sensor for each branch. However, it can identify problems in PV arrays of any size or degree of mismatch.”
The research group trained the fault detection technique on varied data sets with various environmental conditions. Parameters such as temperature, irradiance and maximum power were taken into account.
“In the case of nonlinear classification problems, multiclass deep neural networks are implemented in the extraction process,” the academics explain. “The multilayer perceptron falls under nonlinear classification, that is, complex nonlinear data is acceptable for the calculation process. Each layer is interconnected with existing hidden units. Each hidden unit processes the weights with the help of the bias function.”
To test the MLNN detection method for faults in photovoltaic strings, the researchers simulated a 22.5 kW solar array composed of four parallel strings and 10-series modules. In the simulation, they obtained information about when the current dropped to zero and the current differences in the top and bottom modules of each string. They then processed those measurements into failures and discrepancies and compared them to those presented by the detection model.
The researchers defined accuracy as “the fraction of the total number of accurate predictions made of the possible outputs, divided by the total number of predictions made by the matrix.”
The proposed MLNN achieved an accuracy of 98.76% in detecting LL faults, LG faults, and bypass diode faults. This compares to an accuracy of 96.5% achieved by probabilistic neural network (PNN), 92.1% by radial basis functions (RBF), and 90% by convolutional neural network (CNN), as reported. cited in previous scientific literature.
“The proposed MLNN technique can solve any complex non-linear calculations, handle a large amount of input data from defective panels and quickly predict failures,” the researchers conclude.
Their findings were presented in “ Photovoltaic string fault optimization using multi-layer neural network technique ,” published in Results in Engineering . The research team consisted of academicians from Marri Laxman Reddy Institute of Technology and Management, Andhra Pradesh National Institute of Technology and CVR College of Engineering. |
