Research On Reduced Space Algorithm For Model Predictive Transient Voltage Stability Emergency Control

Model predictive control strategy had fast response to the sudden changes of the systemstructure and random outside interference because of the receding horizon optimization andfeedback correction technology. Therefore, model predictive control was an effective strategyfor power system transient voltage control online, but the nonlinear character of the powersystem predictive model and the computation efficiency for online control has not been solvedefficiently. The computation efficiency of the receding horizon optimization was the criticalbottleneck of the existing model predictive voltage stability control especially for transientvoltage stability emergency control.The receding horizon optimization model was established for power system transientvoltage stability emergency control. In this model, the objective function considered bothvoltage deviations in predictive horizon and control cost in control horizon, and equalityconstraints consisted of the differential and algebraic equations which reflected the dynamicprocess on power system, and inequality constraints were bounds on variables. The dynamiccharacteristic of the induction motor loads which had a significant impact on the transientvoltage stability progress was considered in this paper. The power system transient voltagestability emergency control was implemented by means of regulating reactive power injectionat each bus and reference voltages of automatic voltage regulator (AVR) in premise of no loadshedding. Therefore, the receding horizon optimization model was a dynamic optimizationproblem which consisted of a large amount of differential algebraic equations (DAEs) andbounds on variables.Radau collocation method was applied to convert the dynamic optimization to nonlinearprogramming (NLP) which had high dimension but relatively few degrees of freedom.Considering the few degrees of freedom in this optimization, the reduced space sequentialquadratic programming (RSQP) methods was applied to enhance the computation efficiencyof the high dimensional nonlinear programming. Several key steps of the reduced spacedapproach including space decomposition, computation of cross term and reduced hessian,selection of the independent variables were discussed in detail.Compared with the standard sequential quadratic programming (SQP), interior pointmethod based on BFGS (IPM-BFGS) and interior point method based on limited-memoryBFGS (IPM-LBFGS) in Matlab’s optimization toolbox, test results on IEEE3-machine9-busand IEEE10-machine39-bus system verified the efficiency of the reduced space sequentialquadratic programming (RSQP) from the aspects of computation time and convergence. The time domain simulation model of the model predictive transient voltage stabilityemergency control was established in PSAT (Power System Analysis Toolbox), the simulationresults on the two systems above verified the proposed control strategy could enhance thepower system transient voltage stability efficiently.