Nuclear power system control method and device

Abstract

The present invention provides a nuclear power system control method and device, the nuclear power system control method comprising: obtaining nuclear power system operating parameters; inputting the nuclear power system operating parameters into a nuclear power system action model, and determining a nuclear power system control strategy output by the nuclear power system action model; wherein the nuclear power system action model is iteratively trained based on nuclear power system sample parameters under the guidance of a nuclear power system evaluation model until a target number of iterations and a target convergence accuracy are reached, and the optimal control strategy is obtained, and then updated to a target weight. The nuclear power system control method and device provided by the present invention determine the target weight of the nuclear power system action model under the guidance of a nuclear power system evaluation model, when the target number of iterations and the target convergence accuracy are met and the optimal control strategy is obtained, so that the training of the nuclear power system action model is more sufficient, the application scenarios of the nuclear power system control method are expanded, the error of the control process is reduced, and the accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of control technology, and in particular to a nuclear power system control method and device. Background Art

[0002] Nuclear energy is a new energy source that is clean, low-carbon, has a stable energy supply and high energy density. With the development of clean energy, it has become a topic of great concern to effectively control nuclear power, connect nuclear power with the power grid, and realize the effective use of nuclear energy. However, the nuclear power system is a discrete-time nonlinear complex system whose parameters are entangled with the system state. Moreover, since the nuclear reactor model has high-order characteristics and requires huge computing power, it brings many difficulties to the power control of the nuclear industry.

[0003] The current control methods for nuclear power systems mainly focus on the linearization process and cannot meet the tracking control requirements of nonlinear processes. The application scenarios are relatively limited, the control process has large errors, and the accuracy is low. Summary of the invention

[0004] The present invention provides a nuclear power system control method and device, which are used to solve the defects that the nuclear power system control method in the prior art mainly focuses on the linearization process, cannot meet the tracking control process of the nonlinear process, has a relatively limited application scenario, has a large error in the control process, and has a low accuracy rate. The invention enables the training of the nuclear power system action model to be more sufficient, expands the application scenario of the nuclear power system control method, reduces the error in the control process, and improves the accuracy rate.

[0005] The present invention provides a nuclear power system control method, comprising: obtaining operating parameters of the nuclear power system; inputting the operating parameters of the nuclear power system into a nuclear power system action model, and determining a nuclear power system control strategy output by the nuclear power system action model; wherein the nuclear power system action model is iteratively trained based on sample parameters of the nuclear power system under the guidance of a nuclear power system evaluation model until a target number of iterations and a target convergence accuracy are reached, and the optimal control strategy is obtained and then updated to a target weight.

[0006] According to a nuclear power system control method provided by the present invention, the training process of the nuclear power system action model includes: obtaining a plurality of sample parameters of the nuclear power system; starting a global loop to initialize the weights of the nuclear power system action model and the weights of the nuclear power system evaluation model; starting a local loop, based on the plurality of sample parameters of the nuclear power system, performing an iterative operation in the direction in which the residual error between adjacent value functions of the nuclear power system evaluation model becomes smaller, until the number of iterations reaches the target number of iterations, obtaining an iterative control law and a local optimal control strategy output by the nuclear power system action model and a local optimal performance index function output by the nuclear power system evaluation model; updating the weights of the nuclear power system action model based on the local optimal control strategy, updating the weights of the nuclear power system evaluation model based on the local optimal performance index function, iterating the entire loop and the local loop until the nuclear power system action model reaches the target convergence accuracy, determining the optimal control strategy, and setting the corresponding weights under the optimal control strategy as the target weights of the nuclear power system action model.

[0007] According to a nuclear power system control method provided by the present invention, the weights of the nuclear power system action model and the weights of the nuclear power system evaluation model are initialized, including: based on a random assignment method, the weights of the nuclear power system action model and the weights of the nuclear power system evaluation model are initialized within a preset weight range (-0.1, 0.1).

[0008] According to a nuclear power system control method provided by the present invention, a method for updating the weights of the nuclear power system evaluation model is:

[0009]

[0010]

[0011] Among them, W Vi is the weight of the nuclear power system evaluation model at the i-th iteration, is the value function of the nuclear power system evaluation model at the i-th iteration; ε k ,u i are the system tracking error and controller controlled variable respectively.

[0012] According to a nuclear power system control method provided by the present invention, the local optimal performance index function is:

[0013]

[0014]

[0015] ε k =f(x k )+g(xk ) k -x d,k ;

[0016] x k =f(x k-1 )+g(x k-1 ) k-1 ;

[0017] x d,k =f(x d,k-1 )+g(x d,k-1 ) d,k-1 ;

[0018] u ε,k =u k -u d,k ;

[0019] in, is the local optimal performance index function, U(ε k ,u ε,k ) is the utility function, which represents the normalized sum of the system tracking error and the controller error, ε k represents the system tracking error, f(x) and g(x) are the discretized system dynamic equation coefficient matrices d and x respectively. k =f(x k-1 )+g(x k-1 ) k-1 represents the dynamic equation of the nuclear power system after discretization, x d,k is the expected output of the system, u ε,k is the controller control error, Q and R are standard matrices.

[0020] According to a nuclear power system control method provided by the present invention, the number of input layer nodes of the nuclear power system action model is 5, the number of hidden layer nodes is 5, and the number of output layer nodes is 1; the number of input layer nodes of the nuclear power system evaluation model is 5, the number of hidden layer nodes is 15, and the number of output layer nodes is 1; the target convergence accuracy is 1.0×10 -5 .

[0021] According to a nuclear power system control method provided by the present invention, the operating parameters of the nuclear power system include: average neutron lifetime, reaction value of a control rod group, nominal nuclear energy, effective total proportion of delayed neutrons in the i-th group, heat transfer coefficient between fuel and coolant, heat capacity of coolant mass flow, reaction coefficient of coolant temperature, reaction coefficient of fuel temperature, coolant heat capacity, fuel heat capacity, equivalent single group delayed neutron precursor nuclear decay constant, nuclear reaction value, average coolant inlet temperature, initial equilibrium inlet temperature of coolant, initial equilibrium outlet temperature of coolant, initial equilibrium temperature of fuel, neutron density relative to initial equilibrium density, delayed neutron density relative to initial equilibrium density, average fuel oil temperature, core outlet coolant temperature, reactivity contributed by control rod movement and at least one of control rod speed.

[0022] The present invention also provides a nuclear power system control device, which includes: an acquisition module for acquiring nuclear power system operating parameters; a determination module for inputting the nuclear power system operating parameters into a nuclear power system action model, and determining a nuclear power system control strategy output by the nuclear power system action model; wherein the nuclear power system action model is iteratively trained based on nuclear power system sample parameters under the guidance of a nuclear power system evaluation model until a target number of iterations and a target convergence accuracy are reached, and the optimal control strategy is obtained and then updated to a target weight.

[0023] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a nuclear power system control method as described above is implemented.

[0024] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the nuclear power system control method as described in any one of the above is implemented.

[0025] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned nuclear power system control methods.

[0026] The nuclear power system control method and device provided by the present invention perform iterative training according to sample parameters of the nuclear power system under the guidance of a nuclear power system evaluation model, determine the target weight of the nuclear power system action model when the target number of iterations and the target convergence accuracy are met and the optimal control strategy is obtained, and use the trained nuclear power system action model to obtain the nuclear power system control strategy, so that the training of the nuclear power system action model is more sufficient, the application scenarios of the nuclear power system control method are expanded, the error of the control process is reduced, and the accuracy is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0028] Figure 1 It is a flow chart of the nuclear power system control method provided by the present invention;

[0029] Figure 2 It is a flowchart of the nuclear power system control method provided by the present invention;

[0030] Figure 3 It is a schematic diagram of the principle of the nuclear power system control method provided by the present invention;

[0031] Figure 4 It is a structural schematic diagram of a nuclear power system control device provided by the present invention;

[0032] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] Combine the following Figure 1-Figure 5 The invention describes the nuclear power system control method and device

[0035] like Figure 1 As shown, the present invention provides a nuclear power system control method, which includes the following steps 110 to 120.

[0036] Step 110: Obtain operating parameters of the nuclear power system.

[0037] It can be understood that the operating parameters of the nuclear power system can be the power model system parameters of nuclear power, and the operating parameters of the nuclear power system can include: average neutron lifetime, reaction value of the control rod group, nominal nuclear energy, effective total proportion of delayed neutrons in the i-th group, heat transfer coefficient between fuel and coolant, heat capacity of coolant mass flow, reaction coefficient of coolant temperature, reaction coefficient of fuel temperature, coolant heat capacity, fuel heat capacity, equivalent single group delayed neutron precursor nuclear decay constant, nuclear reaction value, average coolant inlet temperature, initial equilibrium inlet temperature of coolant, initial equilibrium outlet temperature of coolant, initial equilibrium temperature of fuel, neutron density relative to initial equilibrium density, delayed neutron density relative to initial equilibrium density, average fuel oil temperature, core outlet coolant temperature, reactivity contributed by control rod movement and at least one of the speed of control rod.

[0038] Step 120: Input the nuclear power system operation parameters into the nuclear power system action model, and determine the nuclear power system control strategy output by the nuclear power system action model.

[0039] It can be understood that the nuclear power system action model can be a neural network model. After the nuclear power system operating parameters are input into the nuclear power system action model, the nuclear power system action model can process the nuclear power system operating parameters and output the nuclear power system control strategy. The specific expression of the nuclear power system control strategy can be the control law, that is, the functional relationship between the controlled variable and the system input signal. The nuclear power system control strategy can be used to adjust the nuclear power system tracking error so that the system output response can quickly and efficiently meet the control requirements of the nuclear power plant.

[0040] Among them, the nuclear power system action model is iteratively trained based on the sample parameters of the nuclear power system under the guidance of the nuclear power system evaluation model until the target number of iterations and target convergence accuracy are reached, and the optimal control strategy is obtained and then updated to the target weight.

[0041] It can be understood that the nuclear power system evaluation model can be used to guide the nuclear power system action model. This embodiment adopts an adaptive dynamic programming method. The adaptive dynamic programming method uses a controller and an evaluator structure composed of a neural network to approximate the error value function and the error control strategy through an approximate method. The adaptive dynamic programming method uses a function approximation structure to approximate the optimal performance index function and the optimal control strategy in the dynamic programming equation. The adaptive dynamic programming structure may include a dynamic system, a nuclear power system action model, and a nuclear power system evaluation model. The nuclear power system evaluation model is used to approximate the optimal cost function and provide an evaluation to guide the nuclear power system action model to generate an optimal control strategy. After the optimal control strategy output by the nuclear power system action model acts on the dynamic system, the nuclear power system evaluation model is affected by the rewards or penalties generated at different stages of the dynamic system until the nuclear power system action model updates the control strategy so that the overall cost is optimal, and the overall cost is the sum of rewards and penalties.

[0042] Here, the nuclear power system sample parameters may be relevant parameters collected during the historical operation of the nuclear power plant, and a large number of nuclear power system sample parameters may be used to train the nuclear power system action model and the nuclear power system evaluation model.

[0043] The nuclear power system sample parameters can be used to iteratively train the nuclear power system action model and the nuclear power system evaluation model. By controlling the number of iterative training and the accuracy of iterative training, the nuclear power system action model and the nuclear power system evaluation model can achieve the target convergence accuracy, and the optimal control strategy output by the nuclear power system action model can be obtained accordingly.

[0044] The corresponding weights under the optimal control strategy are set as the target weights of the nuclear power system action model, so that a relatively accurate nuclear power system action model can be obtained. Therefore, through the adaptive dynamic programming method, under the guidance of the nuclear power system evaluation model, a nuclear power system action model that meets the requirements can be quickly trained.

[0045] It is worth mentioning that in the process of training the nuclear power system evaluation model and the nuclear power system action model, dual restrictions are imposed. On the one hand, the number of iterative training is limited, and on the other hand, the accuracy of the model after iterative training is limited. This can make the training of the obtained nuclear power system action model more sufficient and more accurate. The training of the nuclear power system action model is guided by the nuclear power system evaluation model. While satisfying nonlinear control, it also solves the problem of parameter changes with system power in other control strategies, which can expand the application scenarios of nuclear power system control methods, reduce errors in the control process, and improve accuracy.

[0046] The nuclear power system control method provided by the present invention performs iterative training according to sample parameters of the nuclear power system under the guidance of a nuclear power system evaluation model, determines the target weight of the nuclear power system action model when a target number of iterations and a target convergence accuracy are met and an optimal control strategy is obtained, and uses the trained nuclear power system action model to obtain the nuclear power system control strategy, thereby making the training of the nuclear power system action model more sufficient, expanding the application scenarios of the nuclear power system control method, reducing the error of the control process, and improving the accuracy.

[0047] like Figure 2 and Figure 3 As shown, in some embodiments, the training process of the nuclear power system action model includes: obtaining multiple nuclear power system sample parameters; starting a global loop to initialize the weights of the nuclear power system action model and the weights of the nuclear power system evaluation model; starting a local loop, based on multiple nuclear power system sample parameters, iterating in the direction of reducing the residual error between adjacent value functions of the nuclear power system evaluation model until the number of iterations reaches the target number of iterations, and obtaining the iterative control law and the local optimal control strategy output by the nuclear power system action model and the local optimal performance index function output by the nuclear power system evaluation model; updating the weights of the nuclear power system action model based on the local optimal control strategy, updating the weights of the nuclear power system evaluation model based on the local optimal performance index function, iterating all loops and local loops until the nuclear power system action model reaches the target convergence accuracy, determining the optimal control strategy, and setting the corresponding weights under the optimal control strategy as the target weights of the nuclear power system action model.

[0048] It is understandable that the nuclear power system sample parameters of the nuclear power system of the 2500-MW PWR nuclear power plant can be obtained and initialized. At the same time, the nuclear power system can be discretized and related transformations can be performed. A nuclear power system action model and a nuclear power system evaluation model can be constructed based on a neural network, and the weights of the nuclear power system action model and the weights of the nuclear power system evaluation model can be initialized, and the network convergence accuracy of the nuclear power system action model and the network convergence accuracy of the nuclear power system evaluation model can be set.

[0049] Here, the adaptive dynamic programming method (ADP) based on data is constructed without the need for a specific mathematical model. That is, the adaptive dynamic programming algorithm is used to automatically adjust the processing method, processing sequence and processing parameters according to the data characteristics of the processing data during the problem processing and analysis process to match the characteristics of the problem, and adjust and optimize the control strategy method of the nuclear power system of the 2500-MW PWR nuclear power plant. The difference between the actual variables and the expected variables of the system is used as the objective function. According to the output requirements of the nuclear power plant system, the optimal control strategy of the system is quickly matched using the adaptive dynamic programming method.

[0050] According to the requirements, the initial stage of the error tracking optimal control system of the 2500-MW PWR nuclear power plant needs to initialize the parameters and prepare the basic data for the algorithm operation. The sample parameters of the nuclear power system may include: the average neutron lifetime Λ(s), the reaction value of the control rod group G r , nominal nuclear energy P 0a (MW), the effective total proportion of delayed neutrons in the i-th group β, the heat transfer coefficient between the fuel and the coolant O (MW·s / ℃), the heat capacity of the coolant mass flow M (MW·s / ℃), the reaction coefficient of the coolant temperature α c , the reaction coefficient of fuel temperature α f 、Coolant heat capacity μ c (MW·s / ℃), fuel heat capacity μ f (MW·s / ℃), equivalent single group delayed neutron precursor nuclear decay constant λ, nuclear reaction value ρ, average coolant outlet temperature T e (℃), initial equilibrium inlet temperature of coolant T e0 (℃), initial equilibrium outlet temperature of coolant T l0 (℃), initial equilibrium temperature of fuel T f0 (℃), neutron density n relative to the initial equilibrium density r , the delayed neutron density c relative to the initial equilibrium density r , average fuel temperature T f , core outlet coolant temperature T l , the reaction value contributed by the control rod movement ρ r , the speed of the control rod Z r and the full power P0 of the system.

[0051] Examples of values ​​corresponding to sample parameters of nuclear power systems can be shown in Table 1.

[0052] Table 1 Examples of sample parameters of nuclear power system

[0053]

[0054] It should be noted that in the example 2500-MW PWR nuclear power plant system, the system model can be described as:

[0055]

[0056] Among them, in the 2500-MW PWR nuclear power plant system, T e Approximately equal to T e0 , the reactor power of the system is P(t) = P 0a n r , n ris the system full power percentage factor, and the values ​​of related parameters and the relationship between them are as described above.

[0057] At the beginning of training, it is necessary to initialize system parameters, global loops and local loops, and network structure. You can pre-set the number of global loops and local loops, that is, the target number of iterations, and set the target convergence accuracy to 1.0×10 -5 .

[0058] It is worth mentioning that in this embodiment, the initialized system input state variable is: 60 =[0.6,0.6,530.8572,306.5198,-0.0047] T , the expected output is: x 80 =[0.8,0.8,594.0060,310.7468,-0.0018] T .

[0059] The initialized network structure is as follows: the number of input layer nodes of the nuclear power system action model is 5, the number of hidden layer nodes is 5, and the number of output layer nodes is 1; the number of input layer nodes of the nuclear power system evaluation model is 5, the number of hidden layer nodes is 15, and the number of output layer nodes is 1; the target convergence accuracy is 1.0×10 -5 .

[0060] In other embodiments or applications, the number of hidden nodes in the nuclear power system action model and the nuclear power system evaluation model can be adjusted based on experience to obtain the best approximation effect; for the utility function, Q=10I 5×5 , R=0.1I 1×1 , where I is the standard unit matrix; the tracking error is initially randomly set to the difference between the initial state and the expected output trajectory, the control period of each iteration can be 1s, and the number of iterations can be set to 50 times.

[0061] In some embodiments, the weights of the nuclear power system action model and the weights of the nuclear power system evaluation model are initialized, including: based on a random assignment method, the weights of the nuclear power system action model and the weights of the nuclear power system evaluation model are initialized within a preset weight range (-0.1, 0.1).

[0062] It can be understood that after the global loop is started, the weights of the nuclear power system action model and the weights of the nuclear power system evaluation model can be initialized by random assignment.

[0063] In this embodiment, it is necessary to initialize the weights of the nuclear power system action model and the nuclear power system evaluation model. In consideration of the weight convergence speed and stability, within the global loop, the weights W of the two networks are cL (L=1,2,....,15),W aL (L=1,2,....5), obtained by random assignment method within the set range (-0.1, 0.1). Among them, W cL is the weight between the input layer nodes and the hidden layer nodes of the nuclear power system evaluation model, W aL is the weight between the input layer nodes and the hidden layer nodes of the nuclear power system action model.

[0064] The weight initialization of the nuclear power system evaluation model and the nuclear power system action model is within the global loop, and the weight initialization can be re-performed at the beginning of each global loop, thereby better achieving the convergence of the neural network on the basis of ensuring network stability and convergence speed, so as to find the optimal performance indicator function and optimal control strategy of the nuclear power system as soon as possible.

[0065] Start the local loop, and based on the initialized relevant parameters, train the nuclear power system evaluation model and the nuclear power system action model after weight assignment through the adaptive dynamic programming method, iteratively update the network weights in the direction of gradient descent of the network minimization error function until the preset target number of iterations is reached, that is, the number of local loops is reached, and the iterative control law is obtained. The local optimal control strategy and the local optimal performance indicator function are obtained through the nuclear power system evaluation model and the nuclear power system action model after the local loop.

[0066] Based on the initialized nuclear power system sample parameters, n r is the full power percentage factor, c r is the delayed neutron density relative to the initial equilibrium density, T f is the average fuel temperature, T l is the core outlet coolant temperature, ρ r The reaction value contributed by the control rod motion, Z r is the speed of the control rod. According to the discretization definition, the discretized nuclear power system can be described as shown in the following formula:

[0067] x k+1 =f(x k )+g(x k ) k ;

[0068] Among them, f(x), g(x) are system parameter matrices, u=Z r is the control variable.

[0069] The system expected output trajectory is:

[0070] x d,k+1 =f(x d,k )+g(x d,k ) d,k ;

[0071] The tracking error of the system parameters is defined as:

[0072] ε k =x k -x d,k =f(x k )+g(x k ) k -x d,k ;

[0073] The controller error can be defined as u ε,k =u k -u d,k , then the tracking error ε can be rewritten as:

[0074] ε k+1 =f ε (ε k )+g ε,k (ε k ) ε,k ;

[0075] where f ε (ε k )=f(ε k +x d,k )-x d,k+1 , g ε,k (ε k )=g(ε k +x d,k ).

[0076] Therefore, the utility function of the system can be defined as:

[0077]

[0078] The system utility function represents the normalized sum of the system parameter tracking error and the controller control error, where Q and R are standard matrices.

[0079] According to the system utility function, the loss function of the system tracking error can be obtained as:

[0080]

[0081] In some embodiments, the local optimal performance indicator function is:

[0082]

[0083]

[0084] ε k =f(x k )+g(x k ) k -x d,k ;

[0085] x k =f(x k-1 )+g(x k-1 ) k-1 ;

[0086] x d,k =f(x d,k-1 )+g(x d,k-1 ) d,k-1 ;

[0087] u ε,k =u k -u d,k ;

[0088] in, is the local optimal performance index function, U(ε k ,u ε,k ) is the utility function, which represents the normalized sum of the system tracking error and the controller error, ε k represents the system tracking error, f(x) and g(x) are the discretized system dynamic equation coefficient matrices d and x respectively. k =f(x k-1 )+g(x k-1 ) k-1 represents the dynamic equation of the nuclear power system after discretization, x d,k is the expected output of the system, u ε，k is the controller control error, Q and R are standard matrices.

[0089] It can be understood that according to the optimality principle and the Hamilton-Jacobi-Bellman equation (HJB equation), the optimal tracking control strategy of the error system can be obtained as:

[0090]

[0091] Therefore, the standard optimal control rate is:

[0092]

[0093] From the above we can deduce that:

[0094] u d,k =g -1 (xd,k )(x d,k+1 -f(x d,k )).

[0095] In order to obtain the optimal control strategy, the tracking error function can be redefined as:

[0096]

[0097] Among them, according to the Bellman optimization principle, the above formula satisfies the Bellman equation:

[0098]

[0099] Therefore, the optimal control strategy is derived as:

[0100]

[0101] Substituting the optimal control strategy into the Bellman equation, we can obtain the error tracking performance index function:

[0102]

[0103] According to the above, in the 2500-MW PWR nuclear power plant example, the iteration rule based on the adaptive dynamic programming method is:

[0104]

[0105] This formula is the update law of the performance indicator function.

[0106]

[0107] This formula is the update law of the control strategy. It can be proved that when the number of iterations is long enough, the performance index function tends to be optimal and has nothing to do with the initial value.

[0108] The self-learning optimization used in this embodiment includes a global loop and a local loop. The goal of the local loop is to minimize the total cost in each cycle, and the goal of the global loop is to update the iterative performance indicator function and the iterative control law to obtain the optimal solution, that is, the optimal control strategy.

[0109] We can set i=0,1,2,… as global loop variables and convert the performance index function Initialize as follows:

[0110]

[0111] in, is a positive definite function.

[0112] The initial iterative control law is shown as follows:

[0113]

[0114] in,

[0115] Therefore, for the global loop variable i=0, 1, 2, ..., iteration can be performed according to the formula shown above.

[0116] Based on the global loop, a local loop is introduced to obtain In an embodiment, the nuclear power system evaluation model updates weights based on the residual errors of two adjacent performance indicator functions.

[0117] In this embodiment, high-order polynomials can be used to replace neurons in the Actor-Critic neural network to obtain approximate solutions for performance indicator functions and control strategies. The nuclear power system evaluation model can be expressed as:

[0118]

[0119] in is the neural network activation function; is the weight vector, and L is the number of neurons in the hidden layer.

[0120] The action model of the nuclear power system can be expressed as:

[0121]

[0122] in, is the neural network weight, M is the number of neurons in the hidden layer, is the activation function.

[0123] In some embodiments, the weights of the nuclear power system evaluation model are updated by:

[0124]

[0125]

[0126] Among them, W Vi is the weight of the nuclear power system evaluation model at the i-th iteration, is the value function of the nuclear power system evaluation model at the i-th iteration; ε k ,u i are the system tracking error and controller controlled variable respectively.

[0127] In the local cycle, the purpose of the nuclear power system evaluation model is to minimize the residual error of the adjacent value function, thereby obtaining a new objective function, which is defined as:

[0128]

[0129] The weight updating method of the nuclear power system evaluation model is obtained as follows:

[0130]

[0131] The weight update method of the nuclear power system action model is:

[0132]

[0133] The updated iterative control law can replace the current iterative control law, and the local optimal performance indicator function can replace the current performance indicator function. The above steps are repeated until the nuclear power system evaluation model and the nuclear power system action model meet the target convergence accuracy, and the optimal control strategy of the error tracking system of the 2500-MW PWR nuclear power plant is obtained.

[0134] According to the set number of iterations of the local loop, if the local loop has not ended at this time, continue to return to the beginning of the local loop to continue the training of the nuclear power system action model and the nuclear power system evaluation model. If the local loop has ended, continue to judge whether the current global loop has ended. If the global loop has not ended, return to the beginning of the local loop to re-initialize the self-learning-based nuclear power plant error tracking control method, and then use the Actor-Critic network to adjust the system utility function according to the initialization result; if the global loop has ended, the optimal nuclear power system control strategy is obtained.

[0135] The utility function of the system at each moment is calculated through the optimal error tracking control strategy, and the performance index function and control strategy are updated according to the corresponding results. This embodiment continuously searches for the optimal discrete-time nonlinear nuclear power system control strategy and saves the neural network data when the corresponding system utility function is the smallest until the network training is completed.

[0136] The biggest advantage of data-based control algorithms is model-free control. Actual systems have serious nonlinearity, uncertainty, and time-varying factors. When it is impossible to obtain an accurate mathematical model, model-free adaptive optimization control can play its own advantages, overcome the contradiction between theoretical models and practical applications, reduce the requirements for models, and have better comprehensive control effects.

[0137] The nuclear power system control device provided by the present invention is described below. The nuclear power system control device described below and the nuclear power system control method described above can be referenced to each other.

[0138] like Figure 4 As shown, the present invention further provides a nuclear power system control device, which includes: an acquisition module 410 and a determination module 420.

[0139] The acquisition module 410 is used to acquire the operating parameters of the nuclear power system.

[0140] The determination module 420 is used to input the nuclear power system operation parameters into the nuclear power system action model, and determine the nuclear power system control strategy output by the nuclear power system action model.

[0141] Among them, the nuclear power system action model is iteratively trained based on the sample parameters of the nuclear power system under the guidance of the nuclear power system evaluation model until the target number of iterations and target convergence accuracy are reached, and the optimal control strategy is obtained and then updated to the target weight.

[0142] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communication interface 520 and the memory 530 communicate with each other through the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute the nuclear power system control method, which includes: obtaining the operation parameters of the nuclear power system; inputting the operation parameters of the nuclear power system into the nuclear power system action model, and determining the nuclear power system control strategy output by the nuclear power system action model; wherein the nuclear power system action model is obtained by iteratively training based on the nuclear power system sample parameters under the guidance of the nuclear power system evaluation model until the target number of iterations and the target convergence accuracy are reached, and the optimal control strategy is obtained by updating to the target weight.

[0143] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0144] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the nuclear power system control method provided by the above-mentioned methods, and the method includes: obtaining the operating parameters of the nuclear power system; inputting the operating parameters of the nuclear power system into the nuclear power system action model, and determining the nuclear power system control strategy output by the nuclear power system action model; wherein the nuclear power system action model is iteratively trained based on the nuclear power system sample parameters under the guidance of the nuclear power system evaluation model until the target number of iterations and the target convergence accuracy are reached, and the optimal control strategy is obtained, and then updated to the target weight.

[0145] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the nuclear power system control method provided by the above-mentioned methods, the method comprising: obtaining the operating parameters of the nuclear power system; inputting the operating parameters of the nuclear power system into a nuclear power system action model, and determining the nuclear power system control strategy output by the nuclear power system action model; wherein the nuclear power system action model is iteratively trained based on nuclear power system sample parameters under the guidance of a nuclear power system evaluation model until a target number of iterations and a target convergence accuracy are reached, and the optimal control strategy is obtained, and then updated to the target weight.

[0146] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0147] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server or a network device) to execute the method described in each embodiment or a part of the embodiment.

[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A nuclear power system control method, characterized in that: include: Obtaining nuclear power system operating parameters; Inputting the nuclear power system operation parameters into the nuclear power system action model, and determining the nuclear power system control strategy output by the nuclear power system action model; Wherein, the nuclear power system action model is obtained by iteratively training based on sample parameters of the nuclear power system under the guidance of the nuclear power system evaluation model until the target number of iterations and the target convergence accuracy are reached, and the optimal control strategy is obtained by updating the target weight; The training process of the nuclear power system action model includes: Acquiring a plurality of sample parameters of the nuclear power system; Opening a global loop to initialize the weights of the nuclear power system action model and the weights of the nuclear power system evaluation model; Starting a local loop, based on a plurality of the nuclear power system sample parameters, performing an iterative operation in a direction in which the residual error between adjacent value functions of the nuclear power system evaluation model becomes smaller, until the number of iterations reaches the target number of iterations, and obtaining an iterative control law and a local optimal control strategy output by the nuclear power system action model and a local optimal performance indicator function output by the nuclear power system evaluation model; Based on the local optimal control strategy, the weight of the nuclear power system action model is updated, based on the local optimal performance indicator function, the weight of the nuclear power system evaluation model is updated, the global loop and the local loop are iterated until the nuclear power system action model reaches the target convergence accuracy, the optimal control strategy is determined, and the corresponding weight under the optimal control strategy is set as the target weight of the nuclear power system action model; The weight updating method of the nuclear power system evaluation model is: Among them, W Vi is the weight of the nuclear power system evaluation model at the i-th iteration, is the value function of the nuclear power system evaluation model at the i-th iteration; ε k ,u i are the system tracking error and controller controlled variable respectively.

2. The nuclear power system control method according to claim 1, characterized in that: Initializing the weights of the nuclear power system action model and the weights of the nuclear power system evaluation model includes: Based on the random assignment method, the weights of the nuclear power system action model and the weights of the nuclear power system evaluation model are initialized within a preset weight interval (-0.1, 0.1).

3. The nuclear power system control method according to claim 1, characterized in that: The local optimal performance index function is: ε k =f(x k )+g(x k )u k -x d,k ; x k =f(x k-1 )+g(x k-1 )u k-1 ; x d,k =f(x d,k-1 )+g(x d,k-1 )u d,k-1 ; in ε,k =in k -in d,k ; in, is the local optimal performance index function, U(ε k ,u ε,k ) is the utility function, which represents the normalized sum of the system tracking error and the controller error, ε k represents the system tracking error, f(x) and g(x) are the discretized system dynamic equation coefficient matrices d and x respectively. k =f(x k-1 )+g(x k-1 ) k-1 represents the dynamic equation of the nuclear power system after discretization, x d,k is the expected output of the system, u ε,k is the controller control error.

4. The nuclear power system control method according to any one of claims 1 to 3, characterized in that: The number of input layer nodes of the nuclear power system action model is 5, the number of hidden layer nodes is 5, and the number of output layer nodes is 1; the number of input layer nodes of the nuclear power system evaluation model is 5, the number of hidden layer nodes is 15, and the number of output layer nodes is 1; the target convergence accuracy is 1.0×10 -5 .

5. The nuclear power system control method according to any one of claims 1 to 3, characterized in that: The operating parameters of the nuclear power system include: average neutron lifetime, reaction value of the control rod group, nominal nuclear energy, effective total proportion of delayed neutrons in the i-th group, heat transfer coefficient between fuel and coolant, heat capacity of coolant mass flow, reaction coefficient of coolant temperature, reaction coefficient of fuel temperature, coolant heat capacity, fuel heat capacity, equivalent single group delayed neutron precursor nuclear decay constant, nuclear reaction value, average coolant inlet temperature, initial equilibrium inlet temperature of coolant, initial equilibrium outlet temperature of coolant, initial equilibrium temperature of fuel, neutron density relative to initial equilibrium density, delayed neutron density relative to initial equilibrium density, average fuel oil temperature, core outlet coolant temperature, reactivity contributed by control rod movement and at least one of the speed of control rods.

6. A nuclear power system control device, characterized in that: include: An acquisition module, used to obtain operating parameters of the nuclear power system; A determination module, used for inputting the nuclear power system operation parameters into the nuclear power system action model, and determining the nuclear power system control strategy output by the nuclear power system action model; Among them, the nuclear power system action model is iteratively trained based on the nuclear power system sample parameters under the guidance of the nuclear power system evaluation model until the target number of iterations and target convergence accuracy are reached, and the optimal control strategy is obtained and then updated to the target weight.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the nuclear power system control method as described in any one of claims 1 to 5 is implemented.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the nuclear power system control method as described in any one of claims 1 to 5 is implemented.

Images