On-line wind speed estimation in im wind generation systems by using adaptive direct and inverse modelling of the wind turbine

This paper presents a neural network (NN) based wind estimator and related Maximum Power Point Tracking (MPPT) technique for high performance wind generator with induction machine. The target is to develop an MPPT system, embedding an adaptive virtual anemometer which is able to correctly estimate the wind speed even in presence of variations of the wind turbine characteristic, caused by aging or even damages. This paper proposes the use of the adaptive properties of feed-forward neural networks to address the on-line estimation of the wind speed even in case of slowly time-varying wind-turbine parameters. The method is inspired to the inverse adaptive control but it is used for parameter estimation and not for control purposes. Once the wind speed is estimated, the machine reference speed is then computed by the optimal tip speed ratio. For the experimental application, a suitably developed test setup has been used, with a back-to-back configuration with two voltage source converters, one on the machine side and the other on the grid side.