In this paper, a neural identifier is derived, based on the work of [34] , however, no sliding techniques are used, and a nonlinear dynamics is proposed. This neural identifier is not used to estimate unknown parameters like in Ref. [34] , neither to estimate the wind speed directly, but to approximate the unknown scaled mechanical
torque signal Tsm. Another difference with the previous cited works, is that, training stage is not necessary as it can learn in real-time the optimal weight vector.