Research On Model Predictive Control Method For DFIG Converter

Model predictive control(MPC) method has the property of fast dynamic response and high parameter robustness. So MPC can be used in MIMO and non-linear system directly. MPC could restrict state variable and output variable to acquire high performance under the effect of evaluation function. However, now widely used proportion-integral controller has the problem of slow current response, high rotor and stator current harmonic contents and weak parameter robustness. These problems should be resolved by looking for other control methods. So this paper researches on doubly fed induction generator(DFIG) MPC control method.The main research content of this paper is as follows: research on DFIG MPC control method, DFIG MPC cascade control method, DFIG parameter identification method based on maximum likelihood estimation.DFIG MPC control method is the main research content of this paper. Firstly, difference form of DFIG augmented model is established. DFIG model is discretized. Controllability analysis is completed. In order to reduce model sensitivity to disturbance noise and simplify MPC control derivation, discreet model should be transformed to augmented model which is embedded integral action to support the research of DFIG MPC control. DFIG MPC control system is established based on MPC theory and DFIG augmented model. Closed-loop system model is completed by using input-output relation. The effect of weighting coefficient and value range are analyzed. Fast dynamic response and stability are testified by simulation.DFIG MPC cascade control method and improvement method are proposed. In order to reduce complexity of MPC system and current prediction error, the system order is reduced and optimized. Firstly, predictive step is chosen to be the minimum step that could offset time delay problem. To reduce the difficulty of managing six order system, voltage\power outer loop and current inner loop are obtained. In order to reduce current predictive parameter sensitiveness, close-loop current predictive model is derived. Grid connection and power control are completed. Theoretical study and simulation are conducted. Simulation result indicate that adopting cascaded pattern and improved method could reduce the complexity of DFIG MPC method and parameter sensitiveness. High control precision and dynamic performance are obtained.Parameter of DFIG will drift after long running. DFIG parameter identification method based on maximum likelihood estimation has been researched. At every sampling point, joint probability function is established which is composed of current estimated parameter and input\output value. Likelihood function is obtained by conduct partial derivative to the outside disturbance variance. The real value of DFIG parameter should satisfy that likelihood function is maximum. Identification result will converge to real value of plant. This paper adopts iteration form of maximum likelihood estimation which is easy to programme and debug. DFIG stator and rotor voltage function are used to obtain stator and rotor parameter. Experimental results show that the result of DFIG maximum likelihood estimation fluctuates little. Power step change does nothing to identification result.In order to verify the effectiveness of proposed DFIG model predictive control strategy and maximum likelihood estimation of DFIG, experiments have been conducted at DFIG wind power experiment platform.