To achieve load reduction and power optimization, wind turbine controllers design requires the availability of reliable control-oriented linear models. These are needed for model-based controller design. Model identification of wind turbine while operating in closed loop is an appropriate solution that has recently shown its capabilities when linear time-invariant controllers and complicated control structures are present. However, the collective pitch control loop, one of the most important wind turbine loops, uses non-linear controllers. Typically, this non-linear controller is a combination of a linear controller and a gain scheduling. This paper presents a new algorithm for identification in closed-loop operation that allows the use of this kind of non-linear controllers. The algorithm is applied for identification the collective pitch demand to generator speed of a wind turbine at various operating points. The obtained models are presented and discussed from a control point of view. The validity of these models is illustrated by their use for the design of a linear fix robust controller. The performance based on simulation data of this linear controller is similar to that obtained with simulations based on a linear controller with gain scheduling, but its design and implementation is much simpler. Copyright © 2012 John Wiley & Sons, Ltd.