Artificial neural network based maximum power point tracking technique for PV systems

The dependency of photovoltaic (PV) arrays on temperature and irradiance levels shapes their known nonlinear behavior; hence maximum power point tracking (MPPT) is mandatory. Traditional MPPT techniques, like Perturb and Observe (P&O) and Incremental Conductance (IncCond), offer acceptable performance with a trade-off between accuracy and fast operation. Moreover, moderate operation is remarked at rapidly changing environmental conditions. On the contrary, off-line trained artificial neural network (ANN) is considered as accurate, fast and robust estimation technique. In this paper, a two-stage off-line trained ANN based MPPT technique is proposed where two cascaded ANNs are utilized. The first estimates the temperature and irradiance levels from the array voltage and current signals while the other network determines the optimum peak operating point from the temperature and irradiance, estimated by the first ANN. The proposed technique offers enhanced performance even under rapidly changing environmental conditions, no need for temperature/irradiance measurement, in addition to reduced required training sets because of the presented ANN cascaded structure.