FTC of LPV systems using a bank of virtual sensors:Aapplication to wind turbines

Abstract

Ahstract-In this paper, an FTC strategy using Linear Parameter Varying (LPV ) virtual sensors is proposed and applied to the IFAC wind turbine case study. The novelty of the proposed strategy consists in that virtual sensors are applied to the FTC problem in a new original fashion. Instead of hiding the fault, the virtual sensors are used to expand the set of available sensors. Then, the state observer is designed using LPV techniques based on Linear Matrix Inequalities (LMIs) taking into account a varying parameter that is introduced in order to select which sensors are used by the observer among the physical and the virtual ones. In this sense, the proposed approach can be considered as a multisensor fusion strategy that integrates data provided by various sensors in order to obtain a better estimation.In order to reduce the cost of energy generated by wind turbines, it is of high importance to increase their reliability. For this reason, research in Fault Detection and Isolation (FDI) and Fault Tolerant Control (FTC) with application to wind turbines has become a subject of interest in research. Until a few years ago, there were no FTC applications to wind turbines in the literature, the common approach being to monitor the turbine conditions and shut it down in the event a fault were detected [l]. In [2], a solution to this problem based on the design of passive and active FTC was proposed for a 4.8 MW variable-speed, variable-pitch wind turbine model with a fault in the pitch system. In [3], a multiobserver switching control strategy for robust active fuzzy FTC has been proposed for variable-speed wind energy conversion systems subject to sensor faults. Recently, a benchmark model of a wind turbine containing sensors, actuators and system faults has been presented, in order to test different detection, isolation and accommodation schemes [4]. Remarkable results were obtained in both the FDI [5] and the FTC parts [6]-[9]. The FTC approach considered in this paper uses the idea of virtual sensors. Initially developed for linear systems [10], virtual sensors have been successfully extended to piecewise affine [11], Hammerstein-Wiener [12] and LPV [13]-[15]systems. The interest in this latter class of systems lies in that LPV techniques have consolidated as an efficient solution to analysis and synthesis problems for nonlinear systems. Since their introduction by Shamma in his Ph.D. thesis [16], the LPV approach has become a standard formalism in systems and control [l7], because gain-scheduling of nonlinear systems can be performed according to the LPV paradigm, where the non-linearity is embedded in the varying parameters that depend on some endogenous signals.In this paper, the idea of virtual sensors is applied to the FTC problem in a new and original way. Instead of hiding the fault, the virtual sensor is used to expand the set of available sensors before the state observer is designed. A parameter is introduced to select the sensors to be used by the observer among the physical and the virtual ones. The pr...

A machine-learning approach to synthesize virtual sensors for parameter-varying systems

“…Inferring the value of ρ k in real time from input/output data can be useful for several reasons. In predictive maintenance and anomaly/fault detection [4,5,6], detecting a drift in the value of ρ k from its nominal value or range of values can be used first to detect a fault and then to isolate its nature. In gain-scheduling control [7,8], ρ k can be used instead to decide the control law to apply at each given time instant.…”

Robust Sensor Fault-Tolerant Control Scheme for Wind Turbines with Hydrostatic Transmission

“…In Sami and Patton (2012a), a passive sensor fault-tolerant control strategy is implemented using a sliding mode controller for the partial-load region that tolerates generator speed sensor faults and generator torque offset faults. In Rotondo et al (2013), an FTC strategy using Linear Parameter Varying (LPV) virtual sensors is proposed and applied to the benchmark model (Odgaard et al (2009)). Instead of hiding the fault, the virtual sensor is used to expand the set of available sensors before the state observer is designed.…”