Application Of Stochastic Predictive Control In Wind Power System

Stochastic systems are widespread in practice.Wind turbine is a nonlinear system with multivariables.In the process of linearization,various uncertainties are caused by wind fluctions.Therefore,to obtain the maximum power and to extend the life of the blade,the choice of control strategy is essential.In this paper,a model with multiplicative uncertainty is established firstly.And impose probabilistic constraints that allow for violations.Then based on the multiplicative uncertainty,the additive uncertainty is introduced to establish the related model due to the dependence of aerodynamic torque on wind speed and pitch angle.And suitable control algorithms are proposed for these two stochastic models.Stochastic predictive control is a class of control algorithms that can systematically consider the uncertainties of objects and satisfy probabilistic constraints.A formulation of the expected performance index function is derived based on the multi-step feedback law designed algorithm for the stochastic model with multiplicative uncertainties.The stochastic coefficient matrix of the state space model is described by probabilistic multicells,and the probability distribution ellipse sets of states are calculated,which can satisfy the probabilistic constraints.Propose the simplified algorithm in order to reduce the computational conplexity,the offline part is responsible for calculating the required matrix variables,and optimize the perfomance index during the online part.For the stochastic system with multiplicative uncertainty and additive uncertainty,this causes that the perfomance index converges to a nonzero limit,and need modify the performance index function.In this paper,we propose an algorithm based on the autonomous state space description,and introduce the concept of probability invariance.We use the confidence ellipsoids in the state space to handle probabilistic invariance in the former researches.The work discussed here propose a significant improvement through the use of the vertices of polytopic sets.The state transition probabilities between these sets are constrained from the current moment to the next moment and form the basis of the Markov chain model.Probabilistic invarinance and Markov chain model form the framework of probabilistic constraint processing.Simulation results show the effectiveness of the proposed algorithm.