Nuclear power plant load tracking method and system based on hybrid time domain model predictive control

Abstract

The application discloses a nuclear power plant load tracking method and system based on a hybrid time domain model predictive control, and belongs to the technical field of nuclear power plant operation optimization. The method is used for unified representation of complex multi-time scale dynamics under the multi-physical field coupling of a nuclear power plant, by constructing a hybrid time domain prediction model based on limited time domain accurate prediction and unlimited time domain equivalent approximation. On the premise of strictly guaranteeing the operation safety and operation constraints of the nuclear power plant, the method aims at minimizing the load tracking error in the full time domain range, so as to realize the fast and smooth tracking of the grid load instruction. The hybrid time domain model predictive control designed in the application can accurately predict the fast and slow dynamic responses of the nuclear power plant at the same time, and implements closed-loop optimal control through online rolling optimization. The hybrid time domain model predictive control not only is beneficial to guaranteeing the safety and stability of the load tracking operation of the nuclear power plant, but also effectively improves the operation flexibility of the nuclear power plant participating in the grid peak shaving.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of nuclear power plant operation optimization, and particularly relates to a nuclear power plant load tracking method and system based on a hybrid time domain model predictive control. BACKGROUND

[0002] The nuclear power plant object has complex characteristics, adopts various new devices and process innovations, and has close coupling between multiple nuclear steam supply system modules. Mechanism modeling analysis is the main technical means for implementing current nuclear power plant optimization control. As a large-scale strongly coupled complex process system, the nuclear power plant mechanism model has multiple time scales, multiphase flow, distributed parameters, and nonlinear complex characteristics coexisting, and its operation is extremely strictly constrained by nuclear safety regulations. Therefore, realizing safe, stable, and accurate load tracking control faces a series of key challenges:

[0003] First, in view of the multi-physical field coupling of neutron physics, thermal-hydraulic, and energy conversion involved in the operation of the nuclear power plant, a control strategy capable of uniformly processing fast and slow varying dynamics needs to be designed; second, it is necessary to ensure that key variables such as reactor power, primary and secondary loop working fluid temperature and pressure during the operation of the nuclear power plant are always maintained within the safety boundary, and the robustness of the load tracking control is guaranteed; finally, the control strategy is required to quickly and smoothly respond to the time-varying grid load instruction. The traditional predictive control method has limitations in handling the multi-objective collaborative optimization of safety, stability, and dynamic performance.

[0004] Therefore, in view of the above, there is an urgent need for a new predictive control framework that can fundamentally coordinate the handling of multi-time scale dynamics, naturally take into account safety constraints, and have online adaptability, to support advanced nuclear power plants to be safely, stably, and efficiently integrated into the grid operation. SUMMARY

[0005] The present application provides a nuclear power plant load tracking method and system based on a hybrid time domain model predictive control, which is directed to the technical field of nuclear power plant operation optimization, by constructing a finite-infinite hybrid time domain predictive model of the nuclear power plant, and respectively introducing safety boundary constraints and operation constraints set according to the nuclear power plant operation safety regulations and the control requirements of the nuclear power plant, establishing a hybrid time domain optimization objective function and boundary conditions, and obtaining an optimal control sequence accordingly, to overcome the limitations of existing predictive control methods in coordinating the handling of multi-time scale dynamics and strictly ensuring safety constraints online, and to realize accurate, flexible, and stable nuclear power plant load tracking operation.

[0006] A nuclear power plant load tracking method based on a hybrid time domain model predictive control, comprising the following steps:

[0007] Divide the full time domain of the nuclear power plant load tracking operation scenario into a finite time domain and an infinite time domain, and construct a hybrid time domain predictive model containing the finite time domain and the infinite time domain;

[0008] Setting safety boundary constraints of state variables and operation constraints of control variables in load following operation as operation constraints according to safety procedures followed by load following operation and control requirements of nuclear power plant respectively;

[0009] Constructing a hybrid time domain optimization objective function for load following;

[0010] Obtaining real-time state observation values and steady-state state set values through a distributed control module of the nuclear power plant and an upper-level scheduling module of the nuclear power plant respectively, and constructing two-point boundary conditions based on the real-time state observation values and the steady-state state set values for constraining a predictive control optimization process;

[0011] Integrating a hybrid time domain predictive model, operation constraints, a hybrid time domain optimization objective function, and two-point boundary conditions to construct and solve a rolling optimization problem, obtain an optimal control sequence at the current time, implement a closed loop, and update iterations to complete load following of the nuclear power plant.

[0012] Preferably, the full time domain of the load following operation scenario of the nuclear power plant is divided into a finite time domain and an infinite time domain, including:

[0013] Based on the dynamic multi-time scale coupling characteristics of the nuclear power plant, fast-varying dynamics and slow-varying dynamics that need to be processed in the predictive model are identified;

[0014] The length of the finite time domain is determined according to the dominant time constant of the fast-varying dynamics, and is used to completely cover the main transient response process from nuclear fission energy release to generator electric energy output;

[0015] The period from the end of the finite time domain to time tending to infinity is defined as the infinite time domain, which is used to represent and process the long-term asymptotic behavior of the slow-varying dynamics of the nuclear power plant.

[0016] Preferably, the infinite time domain is processed to achieve mathematical description and calculation, specifically including:

[0017] A Sigmoid type time scale transformation function is used to map the original infinite time domain to a transformation domain with a finite length, the chain rule is used to calculate the continuous time dynamic characteristics of the nuclear power plant in the transformation domain, and an equivalent approximate model for describing the dynamic behavior in the infinite time domain is constructed based on the continuous time dynamic characteristics using integral transformation.

[0018] Further preferably, the Sigmoid type time scale transformation function is a hyperbolic tangent function.

[0019] Preferably, the hybrid time domain predictive model containing the finite time domain and the infinite time domain is constructed, including:

[0020] Based on the mechanism of nuclear power plant, a continuous-time model in the form of differential-algebraic equations is established to describe the evolution of nuclear power plant under load following operation scenario;

[0021] The continuous-time model is discretized to obtain a discrete-time model describing the evolution of nuclear power plant at sampling period;

[0022] The discrete-time model is applied in finite time domain to construct a truncated model for multi-step forward state prediction, taking the current state of nuclear power plant and future control sequence as input, and outputting the state at the end of finite time domain; the state at the end of finite time domain is taken as the input of the equivalent approximate model to realize the connection between finite time domain and infinite time domain, and obtain a hybrid time domain prediction model.

[0023] Further preferably, the differential-algebraic equations form includes the following types of equations: mass, energy and momentum conservation equations describing the basic conservation of nuclear power plant, neutron kinetics equations describing nuclear fission chain reaction, thermal hydraulic equations describing the heat transfer and flow process of primary and secondary loops, property calculation equations describing the thermophysical properties of primary and secondary loop working fluids, and turbine-generator combined dynamic equations describing the expansion work of main steam and power output.

[0024] Preferably, the discretization of continuous-time model includes: setting the control variables involved in load following operation as input of continuous-time model to zero-order hold at the current discrete time , and implicitly solving the differential-algebraic equations between discrete time steps to realize discretization.

[0025] Preferably, the safety boundary constraints of state variables in load following operation include at least one of the following: change limit of reactor power output, change limit of reactor outlet coolant temperature and steam generator outlet steam temperature, and change limit of main steam pressure.

[0026] The operation constraints of control variables in load following operation include at least one of the following: adjustable range of reactor inlet coolant flow, adjustable range of external input reactivity introduced by control rods, adjustable range of steam generator inlet feedwater flow, and opening range of main steam regulating valve.

[0027] Preferably, the construction of hybrid time domain optimization objective function for load following includes:

[0028] ​​​The tracking phase cost in quadratic form is set according to the expected trajectory, state variable and set value of the control variable combined with the load tracking, for measuring the deviation between the state variable and the expected trajectory or expected value, and the deviation between the control variable and the set value, and the calculation formula is as follows:

[0029] ,

[0030] In the formula, represents the state prediction variable at the discrete time step l in the finite time domain truncated model and the control prediction variable at the discrete time step l, and respectively represent the expected value of the state variable and the control variable at the discrete time step l, represents the Euclidean norm for quantifying the deviation amplitude between the state or control variable and the expected value, and the weight matrix and are used for coordinating the weight of each variable deviation in the tracking phase cost;

[0031] According to the division of the finite time domain and the infinite time domain, the tracking phase cost function is weighted and accumulated in the full time domain to construct the mixed time domain optimization objective function, and the calculation formula is as follows:

[0032] ,

[0033] ,

[0034] ,

[0035] In the formula, is the finite time domain cumulative phase cost, is the infinite time domain approximate phase cost, is the state variable at the discrete time step , and is the control variable at the discrete time step , J is the mixed time domain optimization objective function, and the weight parameters and are used for adjusting the load tracking control performance in the finite time domain near the current discrete time and the infinite time domain after time scale transformation far away.

[0036] Preferably, the optimization problem is represented as follows by integrating the mixed time domain prediction model, the operation constraint, the mixed time domain optimization objective function and the two-point boundary condition:

[0037] ,

[0038] ,

[0039] min, wherein min represents that the target function is minimized to solve, there exists a set of future control sequences in the full time domain, so that the value of the mixed time domain target function reaches the minimum.

[0040] Preferably, the sequence quadratic programming or the interior point method is used to solve the rolling optimization problem to obtain the optimal control sequence of the current discrete time , , is the control sequence index at the discrete time step 0, is the control sequence index at the discrete time step .

[0041] In another aspect, the application also provides a nuclear power plant load tracking system based on mixed time domain model predictive control, comprising a memory and a processor, the memory is used to store a computer program, characterized in that the processor is used to realize the nuclear power plant load tracking method based on mixed time domain model predictive control when the computer program is executed.

[0042] Compared with the prior art, the application has the following beneficial effects:

[0043] The application aims at the limitation that the existing prediction control method is difficult to coordinate the processing of multi-time scale dynamics and strictly guarantee the safety constraints, and provides a nuclear power plant load tracking method and system based on mixed time domain model predictive control. By constructing a mixed time domain prediction model based on finite time domain accurate prediction and infinite time domain equivalent approximation, the complex multi-time scale dynamics under the multi-physical field coupling of the nuclear power plant are uniformly represented. On the premise of strictly guaranteeing the safety and operation constraints of the nuclear power plant, the load tracking error in the full time domain is minimized, so as to realize the rapid and smooth tracking of the grid load instruction. The mixed time domain model predictive control designed in the application can accurately predict the fast and slow dynamic responses of the nuclear power plant at the same time, and implement closed-loop optimal control through online rolling optimization. Not only the safety and stability of the nuclear power plant load tracking operation are guaranteed, but also the operation flexibility of the nuclear power plant participating in the grid peak shaving is effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 The flowchart of the nuclear power plant load tracking method based on mixed time domain model predictive control provided by the embodiment of the application is shown.

[0045] Figure 2 The system structure diagram of the modular high temperature gas cooled reactor nuclear power plant related to the embodiment of the application is shown.

[0046] Figure 3A logic framework diagram of the nuclear power plant load tracking method based on hybrid time domain model predictive control is provided for the embodiments of the present application.

[0047] Figure 4 A change curve of the reactor power output is provided for the implementation.

[0048] Figure 5 A change curve of the reactor outlet hot helium temperature is provided for the embodiments.

[0049] Figure 6 A change curve of the steam generator outlet steam temperature is provided for the implementation.

[0050] Figure 7 A change curve of the main steam pressure before the steam turbine is provided for the embodiments. DETAILED DESCRIPTION

[0051] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments, and it should be noted that the following embodiments are intended to facilitate the understanding of the present application and do not have any limiting effect on the same.

[0052] The present application aims at the limitation that the existing predictive control method is difficult to coordinate the processing of multi-time scale dynamics and strictly guarantee the safety constraints online, and provides a nuclear power plant load tracking method based on hybrid time domain model predictive control, as shown in Figure 1 The implementation steps are as follows:

[0053] S1, the full time domain of the nuclear power plant load tracking operation scene is divided into a finite time domain and an infinite time domain, and a hybrid time domain predictive model containing the finite time domain and the infinite time domain is constructed.

[0054] In the embodiments, as shown in Figure 2 The advanced nuclear power plant is composed of a nuclear island and a conventional island, and the nuclear island is composed of two sets of nuclear steam supply system modules (abbreviated as NSSS modules, containing reactor, steam generator and main helium fan, etc.) of the same design. The main steam formed by the parallel connection of the double NSSS modules is used to drive the steam turbine to generate electricity. The purpose of the present embodiment is to apply control to the nuclear island side, so that the reactor power output under the double NSSS modules can accurately track the expected trajectory set by the grid load instruction of the conventional island side, and the key state variables such as the reactor outlet hot helium temperature, the steam generator outlet steam temperature and the main steam pressure before the steam turbine are maintained near the corresponding set values.

[0055] Therefore, as shown in Figure 3 The reactor, steam generator, control rod module and coolant pump on the nuclear island side of the nuclear power plant, and the typical equipment such as the steam header, steam turbine generator, condenser and feed water pump on the conventional island side. Based on the nuclear power plant block division and the lumped parameter concept, a nuclear power plant mechanism model in the form of a differential algebraic equation system is established, and it is mathematically expressed as ; wherein, denotes continuous time; is a function to uniformly describe the dynamic characteristics of a nuclear power plant under load following operation scenarios, and denote state variables and control variables involved in load following operation scenarios of a nuclear power plant, respectively, both of which will exhibit time-varying properties when the load following operation is presented.

[0056] The dynamic characteristics function should specifically include the following types of equations: mass, energy and momentum conservation equations describing the basic conservation of the system, neutron kinetics equations describing the nuclear fission chain reaction, thermal-hydraulic equations describing the heat transfer and flow processes in the primary and secondary loops, property calculation equations describing the thermophysical properties of the primary and secondary loop working fluids, and combined turbine-generator dynamic equations describing the expansion work of the main steam and the electrical power output.

[0057] Taking a modular high-temperature gas-cooled reactor nuclear power plant as an example, the key state variables and control variables are listed in Tables 1 and 2, respectively.

[0058] Table 1. Key state variables of a modular high-temperature gas-cooled reactor nuclear power plant

[0059]

[0060] Table 2. Control variables of a modular high-temperature gas-cooled reactor nuclear power plant

[0061]

[0062] By setting the nuclear power plant control variables to zero-order hold quantities at a certain time of nuclear power plant load following operation (k = 0, 1, 2, …, used to represent discrete time steps), the differential-algebraic equation system between time steps and is solved implicitly, and thus the discrete-time nuclear power plant mechanism model is obtained:

[0063] ,

[0064] wherein, is the discrete state variable at discrete time step , is the discrete control variable at discrete time step , and is the discrete-time nuclear power plant mechanism model.

[0065] ​Because nuclear power plant operation is a complex process involving the coupling of multiple physical fields such as neutron physics, thermal hydraulics, and energy conversion, it encompasses both rapidly changing dynamics on the order of milliseconds to seconds, such as changes in nuclear fission neutron flux and generator electromagnetic transients, and slowly changing dynamics on the order of minutes to hours, such as changes in the reactor core temperature field and the steam generator level response. Therefore, the characterization of these multiple timescale dynamics should be comprehensively considered when constructing a multi-step model for predictive control. On the one hand, it is necessary to consider the current moment... On the one hand, it is necessary to accurately predict the system behavior within a finite time range at the near end; on the other hand, it is also necessary to approximate the system behavior when the far end time tends to infinity and to control and compensate for it.

[0066] Therefore, this invention divides the entire prediction time domain into a finite time domain and an infinite time domain. Specifically, the time domain division method is used, and the response time is reasonably set according to the rapid dynamic changes (response time of state variables) of the nuclear power plant, with the finite time domain... The length must be sufficient to cover the complete response process of the critical rapid dynamics of the nuclear power plant.

[0067] Specifically, time-domain steps The value of is determined by the dominant fast time constant of the nuclear power plant. For example, it should at least cover the main transient processes (typically on the order of tens of seconds) from the release of nuclear fission energy to the power output response of the generator, so as to ensure that the finite time domain prediction model can fully characterize and optimize these fast dynamics. The sampling period is inherent to the nuclear power plant control system; at the finite time domain endpoint In addition, extending to infinity, all other time dimensions are set to an infinite time domain. Infinite moments satisfy Based on this, a truncated model for accurate prediction in the finite-time domain is constructed:

[0068] ,

[0069] In the formula, and and represent the model prediction variables corresponding to the state variables and control variables at the l-th discrete time step in the finite-time truncation model, respectively.

[0070] Secondly, in order to define it in the infinite time domain The dynamic characteristics of nuclear power plants can be calculated using Sigmoid functions, which have a strictly monotonically increasing property. The original infinite time domain Mapping to a finite-length interval (This is called the transformation domain), and based on the chain rule, the nuclear power plant in the transformation domain... The continuous-time dynamic characteristics of the surface can be calculated as follows:

[0071] ,

[0072] wherein, is the derivative of the Sigmoid type time-scale transformation function at the original continuous time , is the inverse function of the Sigmoid type function.

[0073] Further, based on the integral transformation, the discrete time expression function calculated by the integral of the above continuous time function can be derived and taken as an equivalent approximate model for describing the behavior of the original infinite time domain system:

[0074] ,

[0075] wherein, is the equivalent approximate model obtained after time-scale transformation, and represent the upper and lower bounds of the interval of the transformed domain after time-scale transformation. Since the above equivalent approximate model is integrated on the transformed domain of finite length, it has the computational feasibility, and the integral transformation guarantees the equivalence between the original infinite time domain nuclear power plant dynamic characteristics.

[0076] Taking the hyperbolic tangent function as a specific embodiment of the Sigmoid type function, the time-scale transformation , is the adjustable parameter, and the original infinite time domain is transformed to the interval [0, 1] and can be used to construct an equivalent approximate model for describing the behavior of the infinite time domain system on the domain, The specific form of is as follows, represents the equivalent approximate model for describing the behavior of the infinite time domain system constructed on the domain:

[0077] ;

[0078] According to the continuity at the time , the above truncated model and the equivalent approximate model can be connected, and the state at the end time of the finite time domain is taken as the input of the equivalent approximate model, so as to realize the connection between the finite time domain and the infinite time domain, and obtain a hybrid time domain prediction model.

[0079] S2, according to the safety regulations and control requirements of the nuclear power plant followed in the load tracking operation, the safety boundary constraints of the state variables and the operation constraints of the control variables in the load tracking operation are set as operation constraints.

[0080] The core of this step is to transform the rigid safety procedures that must be followed in the load following operation of a nuclear power plant and the physical limits of each device in the nuclear power plant into executable mathematical constraints in the predictive control optimization problem, to ensure that all the online generated predictive trajectories and control commands are within the absolutely safe and feasible range.

[0081] In the embodiment, first, the safety boundary constraints of the state variables need to be set. In this embodiment, the following key constraints need to be set: the change limit of the reactor power output, the change limit of the reactor outlet hot helium temperature and the steam generator outlet steam temperature, and the change limit of the main steam pressure. The above safety boundary constraints are uniformly expressed in the form of inequality constraints:

[0082]

[0083] In the formula, and respectively represent the safety lower limit and the safety upper limit of the state variable, which together define the interval range that must be strictly followed by the state variable at each prediction time step.

[0084] In addition, the operation constraints of the control variables need to be set, which are determined by the mechanical design, driving ability and control logic of the actuator, and reflect the physical action range that the control system can achieve. In this embodiment, the following key constraints need to be set: the adjustable range of the reactor inlet helium flow, the adjustable range of the external input reactivity introduced by the control rod of the reactor, the adjustable range of the steam generator inlet feedwater flow, and the opening range of the main steam regulating valve. The above operation constraints are uniformly expressed in the form of inequality constraints:

[0085]

[0086] In the formula, and respectively represent the adjustable lower limit and the adjustable upper limit of the control variable, which together define the interval range that must be strictly followed by the control variable at each prediction time step.

[0087] S3, construct a hybrid time domain optimization objective function for load following.

[0088] This step aims to define a mathematical index, i.e. optimization objective function, which can completely measure the performance of full-time domain load following for nuclear power plant operation. This function is used to guide the optimization process while taking into account both short-term accuracy and long-term asymptotic stability.

[0089] In the embodiment, the hybrid time domain optimization objective function is constructed as follows: first, set the following formula to reflect any discrete time step​​​​ The quadratic stage cost of tracking control performance This cost function is used to measure the deviation between key state variables and the desired trajectory or expected value, as well as the deviation between control variables and setpoints.

[0090] ,

[0091] In the formula, This represents the state predictor variable at discrete time step l in a finite-time truncation model. Control predictor variables at discrete time step l The cost function for the tracking phase, and These represent the expected values ​​of the state variable and control variable at discrete time step l, respectively. Their values ​​remain constant within each control cycle and are specifically determined by the setpoint received in S4. The Euclidean norm, representing the magnitude of the deviation between the quantified state or control variable and the expected value, and the weight matrix. and This is used to coordinate the weights of various variable deviations in the cost of the tracking phase. For nuclear power plants with high requirements for load tracking accuracy and real-time performance, the weight matrix components corresponding to the reactor power output term should be appropriately strengthened.

[0092] Based on the defined cost function for the tracking phase, the cost function for the tracking phase is weighted and accumulated over the entire time domain to construct the hybrid time-domain optimization objective function. The calculation formula is as follows:

[0093] ,

[0094] ,

[0095] ,

[0096] In the formula, For the cost of the finite-time accumulation phase, For the infinite time domain approximation stage cost, Discrete time step The state variable at that location, Discrete time step The control variable at point J is the mixed-time domain optimization objective function, and the weight parameters are... and Used to adjust the current discrete time. Load tracking control performance in the finite time domain at the near end and the infinite time domain after time scale transformation at the far end. In this embodiment of the invention, the infinite time domain... The control variables within are held at zero order.

[0097] S4, obtaining real-time state observation values and steady-state state set values through the distributed control module of the nuclear power plant and the upper-layer scheduling module of the nuclear power plant respectively, constructing two-point boundary conditions based on the real-time state observation values and the steady-state state set values, and using the two-point boundary conditions to constrain the predictive control optimization process.

[0098] From this step, the method is transferred from the construction of the prediction model, the operation constraint and the objective function in the offline phase to the execution of the control strategy in the online phase. The core of this step is to set two-point boundary conditions for the predictive control problem by using the state observation values and the scheduling target values that can be obtained in each control period, wherein the two-point boundary conditions include initial conditions and terminal conditions of the predictive control problem, a fixed starting point and an expected ending point, so as to anchor the optimization trajectory of the nuclear power plant load tracking between the real system state and the explicit control target. The two-point boundary conditions set two fixed end points that must be met for the predictive control problem, thereby forcing the optimization problem to start from the actual observation state of the nuclear power plant at the current time for prediction and control calculation, and to finally stabilize the system state to the set value determined by the power grid instruction after long-term operation.

[0099] In the embodiment, the real-time state observation values and the steady-state state set values are obtained through the distributed control system of the nuclear power plant and the upper-layer scheduling optimization system. Specifically, at the current time , the real-time state observation values are collected by the sensor network (such as a neutron detector, a temperature and pressure transmitter, a flowmeter, etc.) that is distributed in the reactor, the primary loop and the secondary loop in the distributed control system, and by means of a standard industrial communication protocol. These data reflect the actual operating state of the nuclear power plant in real time, and are used to construct the initial conditions of the predictive control in the form of the following equation constraint : ;

[0100] Meanwhile, the steady-state state set values are received from the upper-layer scheduling optimization system of the nuclear power plant, and the numerical values thereof represent the steady-state power target and the system state that the nuclear power plant needs to reach and finally maintain in the current power dispatching period, and thus are used to construct the terminal conditions of the predictive control in the form of the following equation constraint: .

[0101] When the traditional predictive control method is used to deal with the load tracking of the nuclear power plant, a key contradiction is faced: in order to ensure long-term stability, an explicit terminal (steady-state) condition needs to be imposed in the optimization problem; but the response process of the slow dynamics such as the core temperature field and the steam generator liquid level in the nuclear power plant is as long as several hours, and the construction of an accurate model with such a long prediction time domain will lead to a huge model size and unfeasible online calculation, so it is usually difficult to effectively impose the terminal condition in practice.

[0102] The present application maps the infinite time domain, which needs to be predicted to infinity, into a finite interval by introducing a time scale transformation. In the transformed domain, the long-term asymptotic behavior (i.e. the final steady state) of the system is compressed and represented at the end of the finite interval (at τ=1). This is equivalent to constructing a describable, equivalent terminal for the optimization problem without significantly increasing the model size and computational load, and fundamentally solving the contradiction between long-time scale dynamic prediction control and real-time calculation feasibility.

[0103] S5, integrating the hybrid time domain prediction model, the operation constraint, the hybrid time domain optimization objective function and the two-point boundary condition, constructing and solving a rolling optimization problem to obtain an optimal control sequence at the current time, implementing a closed loop and updating iterations to complete load tracking of the nuclear power plant.

[0104] This step is the core calculation link of the online execution of the method, and a constraint optimization problem is constructed on the basis of the foregoing steps, and a numerical optimization algorithm is combined to solve the problem to quickly generate an executable control instruction sequence.

[0105] In the embodiment, at the current time The hybrid time domain prediction model, the operation constraint, the hybrid time domain optimization objective function and the two-point boundary condition established in the foregoing steps are integrated to construct a standard nonlinear programming problem, which can be specifically expressed as:

[0106]

[0107]

[0108]

[0109]

[0110]

[0111]

[0112] In the above mathematical expressions, min represents that the objective function immediately following it is minimized, and the essence is to find a set of decision variables (the decision variables of the problem are the control input sequences in the future full time domain ) so that the value of the hybrid time domain objective function reaches the minimum; the symbol s.t. represents "meet the following constraints", which is used to introduce all the equality and inequality constraints that the optimization problem must follow, so that the objective is to minimize the hybrid time domain objective function under the premise of meeting all the equality and inequality constraints, thereby defining the feasible region of the solution. ​​​​​​​

[0113] The above nonlinear programming problem is solved online by using efficient numerical optimization algorithms (such as sequential quadratic programming, interior point method, etc.), which can converge to a local optimal solution satisfying the engineering accuracy requirement within a limited calculation time, and the output is the optimal control sequence at the current time . , is the control sequence index at discrete time step 0, and is the control sequence index at discrete time step . The method uses a rolling optimization mechanism to realize closed-loop feedback, that is, at each control period, the construction and solution of the predictive control problem are repeatedly executed once by combining the feedback information of the nuclear power plant. This rolling optimization design ensures the robustness and adaptability of the nuclear power plant load tracking operation.

[0114] Finally, in each control period of the rolling, the first control quantity is extracted from the optimal control sequence obtained, which is applied to the nuclear power plant actuator, thereby realizing closed-loop control. As the nuclear power plant system transitions to the next control period, according to the real-time feedback of the nuclear power plant reactor power output, primary and secondary loop working medium temperature, main steam pressure and other state information, it is determined whether the nuclear power plant load tracking operation has reached the target steady state. If it has not reached the target steady state, the nonlinear programming problem is solved again based on the updated two-point boundary conditions and a new optimal control sequence is generated. If the target steady state has been reached, it indicates that the nuclear power plant load tracking task is completed, and the iteration is terminated and waits for subsequent grid scheduling instructions.

[0115] On the other hand, the embodiment also provides a nuclear power plant load tracking system based on hybrid time domain model predictive control, comprising a memory and a processor, the memory being used to store a computer program, characterized in that the processor is used to realize the nuclear power plant load tracking method based on hybrid time domain model predictive control when executing the computer program.

[0116] In order to verify the feasibility and effectiveness of the nuclear power plant load tracking method and system based on hybrid time domain model prediction provided by the present application, in the case of wide range load tracking operation of the modular high temperature gas cooled reactor nuclear power plant, the reactor power output curve, primary and secondary loop working medium temperature curve and turbine front main steam pressure curve obtained by controlling and adjusting according to the method are obtained. As shown in Figures 4-7 , the change curves of the key state variables such as the reactor power output of No. 1 and No. 2 reactors, the outlet hot helium temperature of No. 1 and No. 2 reactors, the outlet steam temperature of No. 1 and No. 2 steam generators, and the turbine front main steam pressure are shown respectively. In this case, No. 1 reactor maintains full (100%) power operation, and No. 2 reactor decreases from full (100%) power to half full (50%) power at a specified rate.

[0117] Figures 4-7 It is shown that, although the nuclear power plant load tracking operation in the simulation is affected by external disturbances, the reactor power output of the dual NSSS module and the outlet hot helium temperature can still accurately track the desired trajectory shown by the dashed line, and the steam generator outlet steam temperature and the turbine front main steam pressure of the dual NSSS module are maintained near the respective set values. In summary, the method can accurately predict and coordinate the multi-time scale dynamics of the nuclear power plant, strictly guarantee the operation safety constraints of the nuclear power plant, and on this basis, realize accurate and stable load tracking control.

[0118] It should be noted that the nuclear power plant load tracking method and system based on hybrid time domain model predictive control provided by the above embodiments should be illustrated by the division of the above functional modules when performing nuclear power plant load tracking prediction. The above functions can be completed by different functional modules according to the needs, that is, the internal structure of the terminal or server is divided into different functional modules to complete all or part of the functions described above. In addition, the nuclear power plant load tracking method based on hybrid time domain model predictive control and the nuclear power plant load tracking system based on hybrid time domain model predictive control provided by the above embodiments belong to the same concept, and the specific implementation process is described in detail in the nuclear power plant load tracking method based on hybrid time domain model predictive control. Embodiments, which will not be repeated here.

[0119] The above embodiments have described the technical solutions and advantages of the present application in detail. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the present application. Any modification, supplement and equivalent replacement made within the principle range of the present application should be included in the protection scope of the present application.

Claims

1. A method for load tracking of a nuclear power plant based on hybrid time domain model predictive control, characterized in that, The method comprises the following steps: dividing a full time domain of a nuclear power plant load tracking operation scene into a finite time domain and an infinite time domain, and constructing a hybrid time domain prediction model containing the finite time domain and the infinite time domain; setting, according to safety regulations followed by the load tracking operation and control requirements of the nuclear power plant, safety boundary constraints of state variables and operation constraints of control variables in the load tracking operation as operation constraints; constructing a hybrid time domain optimization objective function for load tracking; obtaining real-time state observation values and steady-state state set values through a distributed control module of the nuclear power plant and an upper-level scheduling module of the nuclear power plant, and constructing two-point boundary conditions based on the real-time state observation values and the steady-state state set values, for constraining a prediction control optimization process; integrating the hybrid time domain prediction model, the operation constraints, the hybrid time domain optimization objective function and the two-point boundary conditions, constructing and solving a rolling optimization problem, obtaining an optimal control sequence at a current time, implementing closed-loop update iteration, and completing nuclear power plant load tracking.

2. The hybrid time domain model predictive control based load tracking method for a nuclear power plant according to claim 1, characterized in that, The method of dividing the full time domain of the nuclear power plant load tracking operation scene into the finite time domain and the infinite time domain comprises: based on dynamic multi-time scale coupling characteristics of the nuclear power plant, identifying fast-varying dynamics and slow-varying dynamics that need to be processed in the prediction model respectively; determining the length of the finite time domain according to a dominant time constant of the fast-varying dynamics, for completely covering a main transient response process from nuclear fission energy release to generator electric energy output; defining a period from the end point of the finite time domain to time infinity as the infinite time domain, for representing and processing long-term asymptotic behavior of the slow-varying dynamics of the nuclear power plant.

3. The hybrid time domain model predictive control based load tracking method for a nuclear power plant according to claim 2, characterized in that, Processing the infinite time domain to realize mathematical description and calculation, specifically comprising: using a Sigmoid type time scale transformation function to map the original infinite time domain to a transformation domain with a finite length, calculating continuous time dynamic characteristics of the nuclear power plant in the transformation domain based on a chain derivation rule, and using an integral transformation to construct an equivalent approximation model for describing dynamic behavior in the infinite time domain based on the continuous time dynamic characteristics.

4. The hybrid time domain model predictive control based load tracking method for a nuclear power plant according to claim 3, characterized in that, The method of constructing the hybrid time domain prediction model containing the finite time domain and the infinite time domain comprises: based on a mechanism of the nuclear power plant, establishing a continuous time model in the form of a differential algebraic equation group for describing evolution of the nuclear power plant under the load tracking operation scene; discretizing the continuous time model to obtain a discrete time model for describing evolution of the nuclear power plant at a sampling period; applying the discrete time model in the finite time domain, taking the state of the nuclear power plant at a current time and a future control sequence as inputs, constructing a truncated model for multi-step forward state prediction, and outputting the state at the end point of the finite time domain; taking the state at the end point of the finite time domain as an input of the equivalent approximation model, realizing connection between the finite time domain and the infinite time domain, and obtaining the hybrid time domain prediction model.

5. The hybrid time domain model predictive control based load tracking method for a nuclear power plant according to claim 4, characterized in that, The continuous-time model is discretized by setting the control variable involved in the load following operation of the input in the continuous-time model to zero order hold At the current discrete time instant Is set to zero order hold The discrete-time step To Implicitly solves the differential algebraic equation group between the discrete-time steps, and realizes the discretization.

6. The hybrid time domain model predictive control based load tracking method for nuclear power plant according to claim 1, characterized in that, The set safety boundary constraints of state variables in the load tracking operation comprise at least one of a change amplitude limit of reactor power output, a change amplitude limit of reactor outlet coolant temperature and steam generator outlet steam temperature, and a change amplitude limit of main steam pressure. The operation constraints of the control variables in the set load tracking operation include at least one of a reactor inlet coolant flow adjustable range, a reactor external input reactivity adjustable range introduced by control rods, a steam generator inlet feedwater flow adjustable range, and a main steam regulating valve opening range.

7. The hybrid time domain model predictive control based load tracking method for nuclear power plant according to claim 1, characterized in that, The mixed time domain optimization objective function for load tracking is constructed, and specifically includes: In combination with the expected trajectory of load tracking, the set values of state variables and control variables, a tracking stage cost in a quadratic function form is set to measure the deviation between the state variables and the expected trajectory or expected value, and the deviation between the control variables and the set values, and the calculation formula is as follows: ， wherein denotes the state variable at discrete time step denotes the control variable at discrete time step the tracking phase cost function, denote the expected value of the state variable and control variable at discrete time step denotes the Euclidean norm of the magnitude of the deviation between the quantized state or control variable and the expected value, the weight matrix is used to coordinate the weight of each variable deviation in the tracking phase cost.​​​​​ According to the division of the finite time domain and the infinite time domain, the tracking stage cost function is weighted and accumulated in the full time domain to construct the mixed time domain optimization objective function, and the calculation formula is as follows: ， ， ， In the formula, For the cost of the finite-time accumulation phase, For the infinite time domain approximation stage cost, Discrete time step The state variable at that location, Discrete time step Control variables at the location, For optimizing the objective function in the mixed time domain, the weight parameters and Used to adjust the current discrete time. Load tracking control performance in the finite time domain of the near end and the infinite time domain after time scale transformation of the far end.

8. The nuclear power plant load tracking method based on mixed time domain model predictive control according to claim 7, wherein a mixed time domain predictive model, operation constraints, a mixed time domain optimization objective function, and two-point boundary conditions are integrated to construct an optimization problem, and the optimization problem is represented as: ， ， min, where min denotes that the objective function is minimized, denotes that there exists a set of control sequences over the future entire time horizon such that the hybrid time-domain objective function attains its minimum value.

9. A nuclear power plant load tracking system based on hybrid time domain model predictive control, comprising a memory and a processor, the memory is used to store a computer program, characterized in that, The processor is configured to implement the nuclear power plant load tracking method based on mixed time domain model predictive control according to any one of claims 1-8 when executing the computer program.

Images