Training data optimization for ANNs using genetic algorithms to enhance MPPT efficiency of a stand-alone PV system

Abstract

Maximum power point tracking (MPPT) algorithms are used to force photovoltaic (PV) modules to operate at their maximum power points for all environmental conditions. In artificial neural network (ANN)-based algorithms, the maximum power points are acquired by designing ANN models for PV modules. However, the parameters of PV modules are not always provided by the manufacturer and cannot be obtained readily by the user. Experimental measurements implemented in the overall PV system may be used to obtain the ANN dataset. One drawback of this method is that the generalization ability of the neural network usually degrades and some data reducing the effectiveness of the network may exist. A genetic algorithm can be used to automatically select the important data among all the inputs, resulting in a smaller and more effective dataset. In our study, a genetic algorithm is used to improve the MPPT efficiency of a PV system with induction motor drive by optimizing the input dataset for an ANN model of PV modules. A variable frequency volts-per-Hertz (V/f) control method is applied for speed control of the induction motor, and a space-vector pulse-width modulation (SV-PWM) method is used to operate a 3-phase inverter. Both simulation and experimental results are presented to demonstrate the validation of the method. © TU©BİTAK.