Comparison of several neural network perturb and observe MPPT methods for photovoltaic applications

In this article, artificial neural network approaches are proposed to achieve an adaptive Maximum Power Point Tracking (MPPT) strategy. Several neural controllers are applied to a boost converter supplied by a standalone PV system. These neural controllers combine two kinds of technique to achieve an adaptive MPPT strategy: A Perturb and Observe P&O technique is combined to a Multilayer Perception's (MLP) learning capabilities. According to this principle, three neural structures are proposed. If their inputs are electrical and environmental parameters; the outputs are the predicted output power at a given instant or the duty cycle to control the boost converter. Several simulation tests allow comparing the performances of the proposed neural P&O controllers with a conventional method. The MPPT with the neural controllers provide low oscillations around the point of maximum power and a fast tracking response to changing conditions.