Research On Nonlinear Model Predictive Emergency Voltage Control In Power Systems

In order to improve the control capability of the power system voltage stability and toenhance spatial and temporal coordination of voltage control means, it is essential to establishthe model of emergency voltage control that can globally mobilize reactive power support andvoltage control potential.Focus on the long-term voltage stability of power system, the paperintroduce nonlinear MPC into emergency voltage control, settle the problem that how toestablish the model of emergency voltage control and how to solve the optimization modelstably and efficiently.Based on quasi-steady-state approximation, the nonlinear model predictive controlmethod is applied to design emergency voltage controller. The receding dynamic optimizationmodel is established to maintain long-term voltage stability by coordinating different kinds ofvoltage control means. In this receding dynamic optimization problem, the objective functionsynthetically considers both voltage deviation and control cost, and equality constraintsinclude continuous and discrete time differential-algebraic equations. To solve this recedingmodel efficiently, Radau collocation method is used to convert this model into a nonlinearprogramming problem which can be solved using the nonlinear primal-dual interior pointmethod.For solving receding dynamic optimization model, in order to improve computationalefficiency and to ensure the computing convergence, the paper proposes the feasibilityrestoration algorithm that can be used to solve large-scale nonlinear programming problems.The Radau collocation method is used to convert this dynamic optimization model into alarge-scale NLP directly, then solve the NLP using the line search filter method, that is satisfythe objective function and constraint violation respectively. when the line search filter methodencounters convergent difficulty during solution of the this nonlinear programming problem,the optimization process will be switched to reconstruct feasibility of original NLP, and try tofind a new iteration point so that the optimization computation can continue from this point.The algorithm is a balanced optimization program between numerical stability andcomputational efficiency.Based on detailed models of power system, the receding optimization model oflong-term voltage stability control is established under framework of model predictive control.In order to improve the computational efficiency and reduce feedback delays，nonlinearprogramming sensitivity algorithm is proposed to solve receding optimization model. Thebasic idea of this approach is converting this model into a nonlinear programming problem using Radau collocation method firstly, calculating the NLP program according to thepredicting state variable values of next period and evaluating the sensitivity of the KKTsystem with respect to model parameters。In the next control period, it calculates the value ofcontrol variables based on sensitivity information according to the true state variable values.The proposed method can improve computational efficiency significantly which creates thecondition for the emergency voltage control application to large-scale systems.