Research On Variable-pitch Control Of Wind Power System Based On Intelligent Sliding Mode Variable Structure Control

Because of the influence of the randomness of wind and the effect ofpneumatic, wind power system has many characteristics, such as strongnonlinearity and fast dynamic process, the large fluctuating wind powergrid-connected power fluctuates largely, which has a negative influence tostability and operation of power grids. With the scale of wind powerdevelopment, the capacity of single generator is growing. The traditionalcontroller is hard to satisfy with the request of well dynamic and steadyperformance. It is necessary to adopt advanced control strategies toensure the high reliable quality of electric power and the stability ofpower system operation.Based on the study of the principle of wind energy generation andthe variable-pitch system, a mathematical model of wind turbines isestablished and analyzed. Then the sliding model control(SMC) isintroduced to eliminate the influence of system parameters perturbationand external disturbances. This thesis overcomes the chatteringphenomenon by combining SMC with neural network theory and supportvector machine(SVM) theory.The below is the main contents：The strategies for pitch angle control are studied based on operatingprinciples of wind turbine. Submodule of wind power system isestablished on Matlab software，the whole engine system simulationmodel is obtained through assembling all submodules. It offerscornerstone for latter pitch-controlled system research.A sliding mode variable structure controller using RBF neuralnetwork is designed based on analyses of the features of wind turbinegenerator system and pitch control strategy. The network gets the initialcenters through a FCM algorithm, and gets the weights through recursiveleast squares(RLS) in off-line training. The sliding mode error isintroduced in the adaptive law to improve the performance of the systems.The strategy mentioned above has such advantages as strong ability ofrejecting chattering, well robust to the variation of parameter and fast response. Coupled with neural network control it also effectively reducesthe chattering of only using a sliding mode variable structure controlsystem.SVM has strong learning ability and better generalizationperformance. The combination of SVM and SMC can not only strengthenthe self adaptability of the conventional SMC，but also provide methodsfor the systematic, integrated and complicated process. To improve thedynamic performance in the operation region of constant power output, asliding mode controller combining with SVM(SVM-SMC controller) isproposed. Conventional sliding mode controller is adopted to generatedata samples for getting the structure and initial parameters of SVM-SMCcontroller in the prophase. After that, we switch to SVM-SMC controllerwith using online self-learning mechanism to realize the adaptive controlfor variable pitch system. The simulink model of variable pitch system isestablished with Matlab/Simulink and simulation results also werecompared. The result shows that the SVM-SMC controller has strongadaptability, good robustness and dynamic performance.