Int. J. Appl. Math. Comput. Sci., 2018, Vol. 28, No. 2, 247–268 DOI: 10.2478/amcs-2018-0018 DATA–DRIVEN TECHNIQUES FOR THE FAULT DIAGNOSISOF A WIND TURBINE BENCHMARK SILVIO SIMANI a, ∗ ,SAVERIO FARSONI a ,PAOLO CASTALDI b a Department of Engineering University of Ferrara, Via Saragat 1/E, 44124 Ferrara, Italy e-mail: {silvio.simani,saverio.farsoni}@unife.it b Department of Electronics, Computer Science and Systems University of Bologna, Via Fontanelle 40, 47100 Forl` ı, Italy e-mail: paolo.castaldi@unibo.it This paper deals with the fault diagnosis of wind turbines and investigates viable solutions to the problem of earlier fault detection and isolation. The design of the fault indicator, i.e., the fault estimate, involves data-driven approaches, as they can represent effective tools for coping with poor analytical knowledge of the system dynamics, together with noise and disturbances. In particular, the proposed data-driven solutions rely on fuzzy systems and neural networks that are used to describe the strongly nonlinear relationships between measurement and faults. The chosen architectures rely on nonlinear autoregressive models with exogenous input, as they can represent the dynamic evolution of the system along time. The developed fault diagnosis schemes are tested by means of a high-fidelity benchmark model that simulates the normal and the faulty behaviour of a wind turbine. The achieved performances are also compared with those of other model-based strategies from the related literature. Finally, a Monte-Carlo analysis validates the robustness and the reliability of the proposed solutions against typical parameter uncertainties and disturbances. Keywords: fault diagnosis, analytical redundancy, fuzzy systems, neural networks, residual generators, fault estimation, wind turbine benchmark. 1. Introduction The increased level of wind-generated energy in power grids worldwide raises the levels of reliability and sustainability required of wind turbines. Wind farms should have the capability to generate the desired value of electrical power continuously, depending on the actual wind speed level and on the grid’s demand. As a consequence, the possible faults affecting the system have to be properly identified and treated before they endanger the correct functioning of the turbines or become critical faults. Megawatt-class wind turbines are extremely expensive systems. Therefore, their availability and reliability must be high, in order to assure the maximisation of the generated power while minimising the operation and maintenance (O & M) services. Alongside the fixed costs of the produced energy, mainly due to the installation and the foundation ∗ Corresponding author of the wind turbine, the O & M costs could increase the total energy cost up to about 30%, particularly considering the offshore installation (Odgaard, 2012). These considerations motivate the introduction of a fault diagnosis system coupled with fault tolerant controllers. Currently, most of the turbines feature a simply conservative approach against faults that consists in the shutdown of the system to wait for maintenance service. Hence, effective strategies coping with faults have to be studied and developed for improving the turbine performance, particularly in faulty working conditions. Their benefits would concern the prevention of critical failures that jeopardise wind turbine components, thus avoiding an unplanned replacement of functional parts, as well as reduction in the O & M costs and an increase in the energy production. The advent of computerised control, communication networks and information techniques brings interesting challenges concerning the development of novel real-time monitoring © 2018 S. Simani et al. This is an open access article distributed under the Creative Commons Attribution-NonCommercial-NoDerivs license (http://creativecommons.org/licenses/by-nc-nd/3.0/).