Research On Model Order Reduction Method And Predictive Control Algorithm Of Grid Voltage Control System

Recently,a modern energy system,which is low-carbon clean,safe and efficient,and sustainable development is being built in China.It highlights the diversity and dynamics of grid operation mode.So it is a huge challenge to maintain normal operation and security and stability of power system,and voltage stability problem becomes one of the core issues.In the voltage regulator,the update period of reference trajectory is too long to reflect the changes of the power system on time.At the same time,due to the use of steady state power flow equations,it is difficult to describe and predictive the dynamic behavior of power system,and the voltages may lose its stability in the transition process from the failure to the stable operating point.It is necessary to use the model which can reflect system dynamics for voltage control.Model predictive control is an open-loop optimization and colsed-loop control method based on a dynamic model.There are two problems to be solved applying this method to power system.The one is how to bulid a predictive model,and the other is how to solve the rolling optimization problem quickly to obtain the optimal control.When there is a high order predictive model,the last problem is more prominent.Given the analysis above,the grid voltage control system model reduction method and predictive control algorithm are investigated systematically and thoroughly based on power system voltage control,predictive control,and Gramian balancing reduction method in this thesis,in order to lay a theoretical foundation for realizing power voltage predictive control.There is a strong feasibility for realizing grid automatic voltage safety predictive control.The main work is summarized as the following,(1)A two-level voltage hierarchical predictive control structure contained trajectory update control sub-level and voltage predictive control sub-level is presented.The trajectory update control sub-levelstarts the computation according to the change of the system at any time for voltage predictive control sub-level.The voltage predictive control sub-leveluses multi-step prediction to predictive system dynamics to make AVR track the trajectory calculated trajectory update control sub-level in order to realize optimal control of the whole system.(2)The problems caused model predictive control applied to voltage hierarchical predictive control structure are analyzed and the solution is proposed.Firstly,a new control mode which is handling an optimal calculation and sending the optimal solution flexibly after sampling many times is presented.Then,a blocking technology is used to describe the change of states and inputs of the system.Thirdly,forming a multi-step prediction and rolling optimization model which can reflect the global basic dynamics of the system and be suitable for applying Gramian balancing reduction.What's more,an implicit expression which ishigher convergent precisionand stability is used to predictive the trajectory of states.Moreover,the historic sampling values are employed to build a grey dynamic model to provide the initial values for optimization iteration.Lastly,the warm start technology and reasonable iteration convergent condition are used to decrease the iteration number.(3)The model reduction of power system linear dynamic model is studied by Gramian balancing reduction.The relationship between research problems and reduction objects is analyzed.The goal-oriented dual low rank cholesky factor alternationg direction implicitmethod is presented to solve high-order Lyapunov functions in order to decrease the time of model reduction.The Gramian balancing reduction is applied to multi-machine power system excitation predictive control and power system voltagehierarchical predictive control based on linear predictive model respectively.(4)The model reduction of power system nonlinear dynamic model is studied by empirical Gramian balancing reduction.The perturbation infliction solution considering the characteristics of the power system is presented to form contain sample data set contain abundant dynamic information to caculate effective empirical controllability and observability covariances.The Gramian balancing reduction is applied to multi-machine power system excitation predictive control and power system voltagehierarchical predictive controlbased on nonlinear predictive model respectively.(5)A toolbox used to study power system model predictive control and balancing reduction is set up,based on power system toolbox inmatlab.The advantages of this toolbox are that the programmes of predictive control and balancing reduction embedded PST have no influence on the original function of PST,and that it can extend more algorithms based on the open source code toprovide convenience for studyingpower system model predictive control and balancing reduction.Some example simulations including single generator-to-infinite bus system,IEEE 4-machine system,IEEE 10-machine system,IEEE 16-machine system,IEEE 50-machine system,andgrid in southern Hebei province to verify the feasibility and validity of methods and schemes proposed in this thesis.