Research On Stable Nonlinear Model Predictive Control In Nuclear Reactor Control System

China is currently in the context of industry structural adjustment and energy structural adjustment.Nuclear power generation has played an important role in ensuring the security of national energy supply,implementing the structural reform of energy supply side and realizing the adjustment of energy structure due to its characteristics of high energy density,self stability and small amount of fuel transportation.The nuclear reactor control system is an important part of the control system of the nuclear power plant.It can automatically track the changes of the power system load in real time by adjusting the water level of the steam generator and the power of the nuclear reactor.The balance between nuclear power,thermal power and load power of nuclear power plant is very important for the safe and efficient operation of nuclear power unit.With the increasing proportion of nuclear power installation and the increasing capacity of nuclear power unit,the "load following" mode must be adopt by the nuclear power unit to improve the stability and power quality of the grid.In the "load following" mode,the frequent power change of the nuclear reactor arouses the essential nonlinearity of the system,which brings new challenges to the control of the nuclear reactor control system.Thus,it is of great significance to study the reasonable optimal control method of nuclear reactor control system under load following mode for the safe and efficient operation of nuclear power unit.Based on analysis of the dynamic characteristics of the nuclear reactor control system,the control problem of nuclear reactor control system of nuclear power plant in the "load following" model is investigated with the nonlinear model predictive control algorithm and numerical simulation.The main contributions of this thesis are as follows.(1)A quasi-min-max fuzzy model predictive controller is constructed for the water level control problem.The approximate fuzzy augmented state model of steam generator water level is established.Then a robust predictive controller is designed based on the quasi-min-max strategy and fuzzy Lyapunov function.To reduce the online computational burden,a bank of ellipsoid invariant sets together with the corresponding feedback control laws are obtained by off-line solving some linear matrix inequalities.Based on the system states,the online part is simplified to a constrained optimization problem with a bisection search for the corresponding ellipsoid invariant set.The stability of the controller is analyzed under the well-known framework of the "three ingredients".(2)A water level soft constraint model predictive controller is constructed to remedy the unstable water level phenomenon caused by the hard constraint on water level of the steam generator.The reason that the unstable zero is reversed to the pole is analyzed.Then the idea of softening water level constraint is adopted into the controller design,and two slack variables are introduced to relax the water level constraint and terminal constraint set.The terminal constraint set is computed off-line with only satisfying terminal constraints set of feed water valve constraints and hence the burden of online computation is reduced.The relationship between the two performance indexes of water level penalty and reference tracking and the coefficient of water level penalty is analyzed from the point of view of multi-objective optimization.The performance of the soft constrained controller is compared with that of unconstrained and hard constrained.The inherit robustness of proposed controller is also discussed.(3)The "flux tilt" phenomenon may occur in the large pressurized heavy water reactor.A fuzzy predictive controller with decentralized structure is designed for spatial control of reactor power.First,the multipoint kinetics modeling method is applied to divide the reactor into 14 regions.Each region uses a set of differential equations to describe the power dynamics.Then,fuzzy modeling is used to approximate the nonlinear dynamic at different regions.The online optimization problems are solved in parallel based on fuzzy model predictive control and"quasi-min-max" strategy.The plant-wide stability is achievable by the asymptotically positive realness constraint for this decentralized control structure.(4)The power control problem of a small reactor under wide range load variation is studied.First,the reactor power control problem is converted to a lexicographic optimization problem with different object priorities.Then,the sequential solution strategy is applied to solve the nonlinear programming problem.The reason why the multi-time scale characteristics of reactor system bring difficulties to online solving of sequential quadratic programming is analyzed.Finally,the continuous control rod velocity set is restricted to the pre-selected discrete velocity value,and the particle swarm optimization is used to efficiently solve the reactor power optimization control problem.