A Mechanical Sensorless MPPT Algorithm for a Wind Energy Conversion System based on a Modular Multilayer Perceptron and a Processor-in-the-Loop Approach

Abstract This paper proposes a novel mechanical-sensorless maximum power point tracking (MPPT) algorithm for a 400 W wind energy conversion systems (WECS). The algorithm is based on a permanent magnet synchronous generator (PMSG), which is connected to a rotor side power converter via a three-phase diode rectifier (3P-DR). In order to achieve a practical implementation and capture the maximum wind energy from a wide range of wind speed values by means of a cost-effective MPPT method, the novel MPPT algorithm is developed via the combination of an intelligent modular multilayer perceptron (MLP) approach and a simplified model of WECS. The rotational speed of the wind turbine (WT) is predicted by the modular MLP. This speed is used by the simplified model to compute the optimal current reference, which is used by a control interface for the rotor side power converter to track the maximum power point (MPP) of WECS. The modular MLP, which includes three feedforward MLP structures, is designed by considering 266,001 data recorded from each conduction modes of the 3P-DR, namely: 2/0, 2/3, and 3/3 modes. Each MLP structure consists of two inputs, five neurons in the hidden layer and an output layer with one neuron, obtaining a considerable reduction in its computational complexity. The output voltage and current of the 3P-DR are the inputs to the modular MLP, while the rotational speed is the target to learn. Results based on the Processor-in-the-Loop (PIL) approach, clearly show that the efficiency of the MPPT process is 99.95% when WECS is subjected to random noise conditions and wind gusts at very high speed (in average above 7 m/s).