On-line soft-sensing model based maximum power point tracking for photovoltaic generation systems

Maximum power point tracking (MPPT) is mandatory for improving the power-conversion efficiency of photovoltaic (PV) systems. Although classical MPPT techniques are widely used by virtue of their simplicity and ease of implementation, the tracking performance is degraded compared with artificial-intelligence based techniques under rapidly changing environmental conditions. This paper proposes an on-line soft-sensing model (OLSSM) based MPPT method, which applies on-line radial basis function neural network as soft model to predict the locus of maximum power points (MPPs). The Euclidean distance between the estimated and measured duty cycle at MPPs is used to check whether the new sample will be used for model updating. Simulation and experimental results show that the OLSSM based MPPT offer enhanced tracking performance especially during transient operations.