Sensor Fault Estimation And Compensation Of PMSG Control Systems In Direct-Drive Wind Energy Applications

With respect to the problems of exacerbating energy security and environmental pollution,wind energy resources are rapidly accounting for an increasing proportion of renewable energy power generation because of the sustainable and renewable characteristics.Since direct-drive permanent-magnet synchronous generator(PMSG)based wind turbines have the advantages of high conversion efficiency,full controllable power and gearless transmission,the total installed capacity and new added capacity have steadily increased in recent years.However,the characteristics of intermittent wind,high density and complex electromagnetic environment lead to frequent component faults of control systems,and the anual failure rate and faulty downtimes are much higher than that in mechanical equipment such as wind turbines.Sensor faults in PMSG control systems cause deviation between measurements and real system states.These measurement errors will be introduced into the control loop though feedbacks.Due to the tracking error elimination by the feedback controllers,the fault features of sensor faults are not so significant.Furthermore,the fault propagation though control loop will cause multiple fault signatures or even secondary faults.During the faulty operation,these sensor faults may cause waveform distortion and power quality degradation or even system shutdown,which may jeopardize the safety and stability of the power grid.The research on sensor fault diagnosis,location and fault-tolerant control of PMSG control systems is of great significance to ensure the continuous and stable operation of grid-connected systems.This dissertation investigates the observer based multiple sensor fault diagnosis and compensation methods due to the fault propagation characteristics and error elimination effect.Observer based estimation methods comprehensively utilize the control input and measurements,and provide a possible solution to the problem caused by aforementioned effects.The main work and research contributions are organized as follows.In order to estimate multiple sensor faults simultaneously in balanced PMSG control systems,the system model is augmented by introducing the mechanical equation in the stationary reference frame.Aiming at the model nonlinearity caused by the introduction of rotor speed and position in augmented model,the linear parameter-varying(LPV)modeling is presented to transform the nonlinear system into a convex polytopic uncertain system.The rotor position measurement is selected as the gain-scheduled parameters.With the investigation of phase current sensor and rotor speed sensor faults including drift,gain variation and disconnections,the control system model of PMSG under various kinds of sensor faults is estabilished in stationary reference frame.Then,observer-based sensor fault estimation methods are discussed further for both current sensors and mechanical sensors.The challenging problems of multiple sensor fault diagnosis and estimation based on conventional system model are summarized with respect to fault propagation and deviation elimination of the closed loop.According to the control modes of three phase current feedback and two-phase current feedback,two kinds of robust multiple current sensor fault diagnosis and location methods are proposed based on the system control loop configurations.The former achieves online location of three phase current sensor faults in abc reference frame while the latter locates both the phase current faults and rotor speed sensor fault.By introducing PMSG mechanical dynamics,the system model is augmented in stationary reference frame to solve the problem of simultaneous estimation of multiple phase current sensor fault.The fault detection and fault estimation observers are designed based on the common Lyapunov function method and Bounded Real Lemma(BRL).The gain matrices of observers at each vertices of the convex hull are calculated by the designed LMI.In the control configuration of three-phase current measurements,the proposed sensor fault estimation algorithm realizes online location of three phase current sensor faults.The multi-sensor fault estimation method under two-phase current measurement operation is designed to realize simultaneous estimation of multiple faults of phase a,phase b and speed sensor.Aiming at simultaneous estimation of multiple faults of speed and current sensor,the system model is established in the stationary reference frame by introducing the mechanical torque dynamics.The nonlinearity of the rotational speed and phase current in dq reference frame is decoupled by the proposed system model.The parameterdependent observer is designed for estimation of multiple sensor faults including the current sensor and rotor speed sensor.The vertex gain matrices of the proposed observer are designed by the extension of BRL for LPV system and disk pole assignment.The gain-scheduled observer is calculated online with rotor position measurement of the PMSG.The fault compensation strategy designed in this dissertation realizes multi-sensor FTC without reshaping the main control loop.The proposed method can simultaneously estimate mechanical torque,speed sensor fault and current sensor faults of PMSG control system,which provides a promising FTC solution for simultaneous fault cases of mechanical sensor and current sensor.In order to enhance the applicability and applicability of the fault estimation observer,an observer-based fault estimation and compensation scheme is designed for multiple current sensor faults without rotor position measurement.This method is presented to solve the problem that gain-scheduled observer design depends on accurate rotor position measurement in stationary reference frame.The rotor position estimation is calculated by utilizing mechanical torque dynamics and rotor speed estimation in fault estimation observer.The position measurement is replaced by the position estimate.Then,the system model consisting of rotor position estimation error is transformed into a linear parameter uncertain model.The multi-sensor fault estimation method is proposed with uncertain scheduling parameter by taking rotor position estimation as the observer gain-scheduling parameter.The designed observer can simultaneously estimate multi-current sensor faults and wind turbine mechanical torque,which is suitable for multi-sensor fault compensation of the applications with varying mechanical torque.