Nuclear reactor core control method and device and electronic equipment
Through the optimization of core working condition prediction model and genetic algorithm, the core control strategy is automatically generated, which solves the problem of cumbersome core control process and improves control efficiency.
Patent Information
- Application Number
- CN202410836419.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-26
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-06-26
AI Technical Summary
The core control process of nuclear reactors is cumbersome, and staff need to continuously adjust their strategies and calculate the control results, resulting in inefficiency.
Using a genetic algorithm based on the core condition prediction model, the core control parameters are automatically optimized to obtain an ideal control strategy by initializing the adaptability evaluation of the population and the generation of dominant populations.
It realizes the rapid and efficient acquisition of ideal core control strategies, reduces the need for manual adjustment and calculation, and improves the efficiency of core control.
Smart Images

Figure CN119008050A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of nuclear control, and in particular to a nuclear reactor core control method, device and electronic equipment. Background Art
[0002] At present, when nuclear power plants are in operation, in order to ensure the stable operation of the nuclear reactor core, it is necessary to control the core according to specific strategies. Common types of control methods include but are not limited to adjusting the boron concentration and inserting control rods. However, the work of core control is still completed by the staff based on their own experience and the current state of the core. Sometimes, the staff will perform theoretical calculations based on the operating history of the core to obtain a reference core control strategy, but this process requires the staff to constantly adjust the core control strategy and calculate the control results to meet the operating requirements, resulting in a cumbersome search process. Summary of the invention
[0003] The purpose of the embodiments of the present application is to provide a nuclear reactor core control method, device and electronic equipment, so that staff can quickly and efficiently obtain an ideal core control strategy.
[0004] In a first aspect, an embodiment of the present application provides a nuclear reactor core control method, the method comprising:
[0005] In the case where a plurality of first predicted operating condition parameter groups corresponding to a plurality of first control parameter groups are predicted based on a core operating condition prediction model, the plurality of first predicted operating condition parameter groups are used as an initialization population, wherein each of the first control parameter groups includes at least one first control parameter corresponding to at least one time step, each of the first predicted operating condition parameter groups includes at least one first predicted operating condition parameter corresponding to at least one time step, each of the first control parameter groups corresponds to an individual in the initialization population, a factor in the individual is a first control parameter corresponding to each time step in the first control parameter group, a parameter type of the first predicted operating condition parameter includes an axial power offset of the core, and a parameter type of the first control parameter includes a relative power of the core;
[0006] Based on the fitness value of each individual in the initialization population, an iterative process of fitness evaluation and dominant population generation is performed to obtain an optimal individual, wherein the fitness value of each individual is determined by using an fitness evaluation function, and the fitness evaluation function involves the difference between the axial power offset corresponding to each time step in the first prediction operating condition parameter group and the reference axial power offset corresponding to each time step, and the reference axial power offset is determined according to the corresponding relative power and the axial power offset when the core is at full power, and the dominant population generation is used to generate a dominant population according to the result of the fitness evaluation, and the optimal individual is an individual in the dominant population generated when the iterative process satisfies the iteration termination condition;
[0007] The core is controlled using the first control parameter corresponding to the optimal individual.
[0008] In some embodiments, the fitness value is determined by the following steps:
[0009] Determine the reference axial power offset corresponding to each time step;
[0010] Determine the difference between the axial power offset corresponding to each time step in the first predicted operating condition parameter group and the reference axial power offset corresponding to each time step to obtain at least one difference, wherein different differences correspond to different time steps;
[0011] Based on the at least one difference value, the fitness value is determined.
[0012] In some embodiments, determining the fitness value based on the at least one difference value includes:
[0013] determining a ratio between an absolute value of each of the differences and a corresponding reference axial power offset;
[0014] Based on the ratio corresponding to each of the difference values, convert the difference values to obtain a conversion value of the at least one difference value;
[0015] The fitness value is determined based on a converted value of the at least one difference value.
[0016] In some implementations, converting the difference values based on the ratio corresponding to each of the difference values to obtain the converted value of the at least one difference value includes any of the following:
[0017] When the ratio corresponding to the difference is less than a first threshold, converting the difference into a first conversion value;
[0018] When the ratio corresponding to the difference is less than or equal to the second threshold and greater than or equal to the first threshold, converting the difference into a second conversion value;
[0019] When the ratio corresponding to the difference is greater than the second threshold, the difference is converted into a third conversion value.
[0020] In some embodiments, determining the fitness value based on the converted value of the at least one difference value includes:
[0021] determining a maximum absolute value among the absolute values of the at least one difference as a correction amount;
[0022] The converted value of the at least one difference is summed, and the fitness value is determined according to the summed result and the correction amount.
[0023] In some embodiments, generating a dominant population comprises:
[0024] According to the sorting results of all the individuals, all the individuals are divided into a first population and a plurality of second populations, wherein the sorting results are obtained based on the fitness values of the individuals, the individuals in the first population are used for factor retention, factor mutation and factor crossover, and the individuals in each of the second populations are used for factor mutation and factor crossover;
[0025] Using the individuals in the first population and the plurality of second populations to perform factor crossover and factor mutation;
[0026] The dominant population is obtained by combining individuals obtained by factor crossover and factor mutation using individuals in the first population and the multiple second populations with individuals in the first population that have undergone factor retention.
[0027] In some embodiments, the plurality of second populations include a first subpopulation, a second subpopulation, and a third subpopulation, the first subpopulation, the second subpopulation, and the third subpopulation are divided according to the sorting results of the individuals in the initialization population, the individuals used for factor crossover in the first population, the first subpopulation, and the second subpopulation include elite individuals, when the elite individuals perform factor crossover with other individuals to obtain new individuals, the probability that the factors of the elite individuals are retained in the new individuals is a first preset probability, and the first preset probability is greater than 0.5,
[0028] The step of performing factor crossover using individuals in the first population and the plurality of second populations comprises:
[0029] Factor crossover is performed using individuals of the first population, the first sub-population, and the second sub-population, the probability that an individual of the first population is determined to be the elite individual is a second preset probability, the probability that an individual of the first sub-population is determined to be the elite individual is a third preset probability, and the third preset probability is less than the second preset probability;
[0030] Factor crossover is performed using individuals from the first sub-population, the second sub-population, and the third sub-population, the probability that an individual from the second sub-population is determined as the elite individual is a fourth preset probability, the probability that an individual from the third sub-population is determined as the elite individual is 0, and the fourth preset probability is less than the third preset probability.
[0031] In some embodiments, when a plurality of first predicted operating condition parameter groups corresponding to a plurality of first control parameter groups are predicted based on the core operating condition prediction model, before using the plurality of first predicted operating condition parameter groups as the initialization population, the method further includes:
[0032] Using a plurality of second control parameter groups of working conditions to be simulated, performing xenon oscillation simulation processing to obtain simulation data of each of the working conditions to be simulated, wherein the plurality of working conditions to be simulated are generated based on a random strategy, each of the simulation data includes the second control parameter group and a second predicted working condition parameter group, each of the second control parameter groups includes at least one second control parameter corresponding to at least one time step, and each of the second predicted working condition parameter groups includes at least one second predicted working condition parameter corresponding to at least one time step;
[0033] Using the second predicted operating condition parameter groups in each of the simulation data, the second control parameter groups in the simulation data are labeled to generate training samples corresponding to the simulation data;
[0034] Based on the training samples corresponding to the simulation data of the multiple operating conditions to be simulated, the model to be trained is iteratively trained to obtain the core operating condition prediction model.
[0035] In some embodiments, the model to be trained is a convolutional neural network model. Before iteratively training the model to be trained based on training samples corresponding to the simulation data of the multiple operating conditions to be simulated to obtain the core operating condition prediction model, the method further includes:
[0036] The parameters to be added are added to the training sample, and together with the second control parameter group in the training sample, serve as the annotation objects of the second predicted operating condition parameter group in the training sample, wherein the parameters to be added include:
[0037] At least one target control parameter group associated with the second control parameter group in the training sample, the time steps corresponding to some parameters in the target control parameter group are located before the time steps corresponding to each parameter in the second control parameter group.
[0038] In some embodiments, each of the simulation data includes parameters related to a first number of time steps, and each of the training samples includes parameters related to a second number of time steps, the first number being greater than the second number.
[0039] The method of using the second predicted operating condition parameter group in each of the simulation data to label the second control parameter group in the simulation data to generate training samples corresponding to the simulation data includes:
[0040] In the time steps involved in the parameters included in the simulation data, taking any time step as a starting point, selecting a plurality of groups of the second number of time steps;
[0041] Set each group of second control parameters corresponding to the second number of time steps as a sample to be labeled;
[0042] The second predicted operating condition parameter set corresponding to each group of the second number of time steps is set as a labeled sample;
[0043] The corresponding samples to be labeled are labeled using each of the labeled samples to generate training samples corresponding to the simulated data.
[0044] In a second aspect, an embodiment of the present application provides a nuclear reactor core control device, the device comprising:
[0045] An initialization population determination module is used for, when a plurality of first predicted operating condition parameter groups corresponding to a plurality of first control parameter groups are predicted based on a core operating condition prediction model, using the plurality of first predicted operating condition parameter groups as an initialization population, wherein each of the first control parameter groups includes at least one first control parameter corresponding to at least one time step, each of the first predicted operating condition parameter groups includes at least one first predicted operating condition parameter corresponding to at least one time step, each of the first control parameter groups corresponds to an individual in the initialization population, a factor in the individual is a first control parameter corresponding to each time step in the first control parameter group, a parameter type of the first predicted operating condition parameter includes an axial power offset of the core, and a parameter type of the first control parameter includes a relative power of the core;
[0046] an optimal individual acquisition module, for performing an iterative process of fitness evaluation and dominant population generation based on the fitness value of each individual in the initialization population, to obtain an optimal individual, wherein the fitness value of each individual is determined by using an adaptability evaluation function, the adaptability evaluation function involves a difference between an axial power offset corresponding to each time step in the first prediction operating condition parameter group and a reference axial power offset corresponding to each time step, the reference axial power offset is determined according to a corresponding relative power and an axial power offset at full core power, the dominant population generation is used to generate a dominant population according to a result of the adaptability evaluation, and the optimal individual is an individual in the dominant population generated when the iterative process satisfies an iterative termination condition;
[0047] The core control module is used to perform core control using the first control parameter corresponding to the optimal individual.
[0048] In a third aspect, an embodiment of the present application provides an electronic device, including:
[0049] a memory configured to store instructions; and
[0050] The processor is configured to call the instructions from the memory and to implement the nuclear reactor core control method provided in the first aspect of the embodiment of the present application when executing the instructions.
[0051] In an embodiment of the present application, multiple first control parameter groups representing multiple core control strategies are used as the initial population in the genetic algorithm, multiple first predicted operating condition parameters corresponding to the multiple first control parameter groups are predicted by the core operating condition prediction model, and the first predicted operating condition parameters in the first predicted operating condition parameter group are used to perform adaptability evaluation on the corresponding individuals in the initial population and iterative processing of dominant population generation to obtain the optimal individual, which represents the ideal core control strategy obtained by the genetic algorithm iteration. In this way, there is no need to constantly adjust the control strategy and calculate the control results, so that the staff can quickly and efficiently obtain the ideal core control strategy, and use the first control parameter in the first control parameter group corresponding to the optimal individual representing the core control strategy to perform core control. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] Figure 1 It is a schematic flow chart of a nuclear reactor core control method provided in an embodiment of the present application;
[0053] Figure 2 is a schematic structural diagram of a nuclear reactor core control device provided in an embodiment of the present application;
[0054] Figure 3 It is a schematic diagram of the structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION
[0055] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.
[0056] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the terms used in this way are interchangeable where appropriate, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.
[0057] The nuclear reactor core control method and device, and electronic device provided in the embodiments of the present application are described in detail below with reference to the accompanying drawings through specific embodiments and their application scenarios.
[0058] See also Figure 1 , is a flow chart of a nuclear reactor core control method provided in an embodiment of the present application, which is applied to electronic equipment. Figure 1 As shown, the nuclear reactor core control method includes the following steps S100 to S300.
[0059] Step S100: When a plurality of first predicted operating condition parameter groups corresponding to a plurality of first control parameter groups are predicted based on a core operating condition prediction model, the plurality of first predicted operating condition parameter groups are used as an initialization population, wherein each first control parameter group includes at least one first control parameter corresponding to at least one time step, each first predicted operating condition parameter group includes at least one first predicted operating condition parameter corresponding to at least one time step, each first control parameter group corresponds to an individual in the initialization population, the factors in the individual are the first control parameters corresponding to each time step in the first control parameter group, the parameter types of the first predicted operating condition parameters include the axial power offset of the core, and the parameter types of the first control parameters include the relative power of the core.
[0060] The core operating condition prediction model in the embodiment of the present application can predict the operating state of the core by receiving the first control parameter group that characterizes the core control strategy, and the operating state of the core is characterized by the first predicted operating condition parameter group. That is, the core operating condition prediction model is used to predict the operating state of the core after being controlled according to the core control strategy according to the core control strategy.
[0061] In an embodiment of the present application, the first control parameter group characterizing the core control strategy includes at least one first control parameter corresponding to at least one time step, that is, in the first control parameter group, each time step corresponds to a first control parameter. It can be understood by those skilled in the art that a time step characterizes a time interval in the core control strategy, and two adjacent time steps are adjacent time intervals at both ends in the core control strategy. It can be seen that each first control parameter in the first control parameter group characterizes the core control strategy within each time step.
[0062] The first predicted operating condition parameter group characterizing the operating state of the core includes at least one first predicted parameter corresponding to at least one time step, that is, in the first predicted operating condition parameter group, each time step corresponds to a first predicted operating condition parameter. Each first predicted operating condition parameter in the first predicted operating condition parameter group characterizes the operating state of the core at the end of each time step.
[0063] In the embodiment of the present application, the parameter type of the first predicted operating condition parameter includes the axial power offset of the core, and the parameter type of the first control parameter includes the relative power of the core.
[0064] The electronic device can generate multiple different first control parameter groups according to the random strategy, that is, multiple core control strategies are generated. After multiple groups of first predicted operating condition parameter groups corresponding to multiple groups of first control parameter groups are predicted based on the core operating condition prediction model, each first control parameter group is regarded as an individual, and each first control parameter in the first control parameter group is a factor that constitutes the individual and can be crossover or mutated. All individuals constitute an initialization population for subsequent iterative processing based on the genetic algorithm and generation of a dominant population.
[0065] Exemplarily, the parameter type of the first predicted operating condition parameter is the axial power offset (AO), and the parameter type of the first control parameter may include the time step length (T) and the rod position (S) of the temperature control rod in addition to the relative power (P) of the core. The core control strategy includes three time steps, a first predicted operating condition parameter group may be written as [AO1, AO2, AO3], and a first control parameter group may be written as [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)]. The electronic device generates a plurality of [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)] with different specific parameter values according to a random strategy, each [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)] is regarded as an individual, (P1, T1, S1), (P2, T2, S2) and (P3, T3, S3) are factors in the individual, and all [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)] constitute an initialization population for subsequent iterative processing based on a genetic algorithm and generation of a dominant population.
[0066] Step S200: Based on the fitness value of each individual in the initialized population, perform fitness evaluation and iterative processing of dominant population generation to obtain the optimal individual, wherein the fitness value of each individual is determined using a fitness evaluation function, and the fitness evaluation function involves the difference between the axial power offset corresponding to each time step in the first predicted operating condition parameter group and the reference axial power offset corresponding to each time step, and the reference axial power offset is determined according to the corresponding relative power and the axial power offset when the core is at full power. Dominant population generation is used to generate a dominant population based on the result of the fitness evaluation, and the optimal individual is an individual in the dominant population generated when the iterative processing satisfies the iteration termination condition.
[0067] The electronic device uses multiple groups of first predicted operating condition parameters to perform adaptability evaluation on corresponding multiple groups of first control parameters, and performs iterative processing of generating dominant populations according to the adaptability evaluation results until the iteration termination condition is met, so as to obtain the optimal individual in the dominant population when the iteration terminates.
[0068] In an embodiment of the present application, the parameter type of the first predicted operating condition parameter includes the axial power offset of the core, and the parameter type of the first control parameter includes the relative power of the core. When using multiple first predicted operating condition parameter groups for adaptability evaluation, the electronic device needs to determine the fitness value of each individual, and the fitness value is calculated and determined according to the adaptability evaluation function. The adaptability evaluation function involves the difference between the axial power offset corresponding to each time step in the first predicted operating condition parameter group and the reference axial power offset corresponding to each time step. The reference axial power offset of each time step is determined according to the axial power offset when the core is at full power and the corresponding relative power.
[0069] Exemplarily, during the genetic algorithm, the electronic device first performs a fitness evaluation on each individual in the initialized population, i.e., [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)] with different specific parameter values mentioned above, based on the fitness evaluation function.
[0070] For a single [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)], the adaptability assessment needs to utilize [AO1, AO2, AO3] corresponding to the [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)] predicted by the core operating condition prediction model. The adaptability assessment also utilizes the reference axial power offset, which also corresponds to each time step. The reference axial power offset of each time step is determined based on the axial power offset when the core is at full power and the corresponding relative power - the reference axial power offset of each time step is the product of the axial power offset when the core is at full power and the corresponding relative power. If the axial power offset when the core is at full power is expressed as AO0, that is, the reference axial power offset corresponding to [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)] can be expressed as: [(P1×AO0, P2×AO0, P3×AO0]. The adaptive evaluation function involves the difference between the axial power offset corresponding to each time step and the reference axial power offset corresponding to each time step. All differences of an individual can be expressed as: [AO1-P1×AO0, AO2-P2×AO0, AO3-P3×AO0].
[0071] After the adaptability evaluation of each [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)] is completed, the dominant population is generated according to the adaptability evaluation result, that is, an iterative update of all individuals is performed. Then the dominant population obtained after the iteration is subjected to adaptability evaluation and dominant population generation, that is, it is iterated again until the iteration termination condition is met to obtain the [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)] corresponding to the optimal individual. It can be understood by those skilled in the art that in each iteration process, the core operating condition prediction model needs to be used to predict the first predicted operating condition parameter corresponding to each individual as a basis for adaptability evaluation.
[0072] Step S300: Perform core control using the first control parameter corresponding to the optimal individual.
[0073] The first control parameter group corresponding to the optimal individual represents an ideal core control strategy obtained by iterative genetic algorithm, and the electronic equipment can use the first control parameter in the first control parameter group corresponding to the optimal individual to perform core control.
[0074] Exemplarily, the first control parameters corresponding to the optimal individual are [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)]. The staff can use these first control parameters as control strategies to control the relative power of the core, the time step size and the rod position of the temperature control rod.
[0075] Through the above steps S100 to S300, multiple first control parameter groups representing multiple core control strategies are used as the initial population in the genetic algorithm, and multiple first predicted operating condition parameters corresponding to the multiple first control parameter groups are predicted by the core operating condition prediction model. The first predicted operating condition parameters in the first predicted operating condition parameter group are used to perform adaptability evaluation on the corresponding individuals in the initial population and iterative processing of dominant population generation to obtain the optimal individual, which represents the ideal core control strategy obtained by the genetic algorithm iteration. In this way, there is no need to constantly adjust the control strategy and calculate the control results, so that the staff can quickly and efficiently obtain the ideal core control strategy, and use the first control parameter in the first control parameter group corresponding to the optimal individual representing the core control strategy to perform core control.
[0076] In some embodiments, in addition to the relative power of the core, the parameter type of the first control parameter may further include at least one of the following:
[0077] Core control power type parameters, core control rod position type parameters and time step size.
[0078] In addition to the relative power of the core, the parameter type of the first control parameter may include at least one of a core control power type parameter, a core control rod position type parameter, and a time step length. The core control power type parameter includes the relative power of the core, the core burnup, etc., and the core control rod position type parameter includes the temperature control rod position, the power control rod total position, etc. The staff can make adaptive parameter selections at the stage of training the core condition prediction model according to the implementation scenario or control requirements of the control strategy and the network structure type of the core condition prediction model.
[0079] In some embodiments, in addition to the axial power offset of the core, the parameter type of the first predicted operating condition parameter may further include at least one of the following:
[0080] The core enthalpy rise factor, the core hot spot factor and the core boron concentration.
[0081] In addition to the axial power offset of the core, the parameter type of the first predicted operating condition parameter may also include at least one of the enthalpy rise factor of the core, the hot spot factor of the core, and the boron concentration of the core. The staff may select adaptive parameters at the stage of training the core operating condition prediction model according to the implementation scenario or control requirements of the control strategy and the network structure type of the core operating condition prediction model.
[0082] In some implementations, the iteration termination condition may be that the number of iterations reaches a preset threshold, or that the fitness value of an individual in the generated dominant population reaches a preset threshold.
[0083] In some embodiments, the fitness value may be determined by the following steps:
[0084] Determine the reference axial power offset corresponding to each time step;
[0085] Determine the difference between the axial power offset corresponding to each time step in the first predicted operating condition parameter group and the reference axial power offset corresponding to each time step to obtain at least one difference, where different differences correspond to different time steps;
[0086] Based on the at least one difference value, a fitness value is determined.
[0087] Adaptability assessment requires determining the fitness value of each individual. To determine the fitness value of a certain individual, first, the reference axial power offset corresponding to each time step of the individual needs to be determined using the first predicted operating condition parameter group of the individual, and then the difference between the axial power offset corresponding to each time step in the first predicted operating condition parameter group and the reference axial power offset corresponding to each time step is determined. The number of differences is equal to the number of time steps in the first predicted operating condition parameter group, and each difference corresponds to a different time step. Finally, based on all the differences obtained, the fitness value of the individual is determined.
[0088] Exemplarily, all differences of an individual may be expressed as [AO1-P1×AO0, AO2-P2×AO0, AO3-P3×AO0], and the fitness value of the individual is determined based on all the obtained differences.
[0089] In some implementations, determining a fitness value based on at least one difference value may include:
[0090] Determine the absolute value of the difference with the largest absolute value among at least one difference as the fitness value.
[0091] The goal of the selection, iteration and optimization of the core control strategy in the embodiment of the present application is to maintain the first predicted operating parameters at a relatively stable level at each time step, that is, to minimize the gap between the first predicted operating parameters and the reference value. Therefore, the fitness value can be determined as the absolute value of the difference with the largest absolute value among all the differences of the individual, and the absolute value represents the maximum gap between the first predicted operating parameter in the first predicted operating parameter group corresponding to the individual and the reference value. It can be seen that when the fitness value is determined by this fitness value determination method, the smaller the fitness value, the better the control strategy corresponding to the individual, and the optimal individual is the individual with the best fitness value performance in the dominant population when the iteration terminates, that is, the individual with the smallest fitness value.
[0092] Exemplarily, first remove the absolute values of all differences of the individual, and all the absolute values obtained can be expressed as: [|AO1-P1×AO0|, |AO2-P2×AO0|, |AO3-P3×AO0|]. If the largest value among all absolute values is |AO2-P2×AO0|, then the fitness value of the individual is |AO2-P2×AO0|.
[0093] In some implementations, determining a fitness value based on at least one difference value may include:
[0094] determining a ratio between the absolute value of each difference and a corresponding reference axial power offset;
[0095] Based on the ratio corresponding to each difference value, the difference value is converted to obtain a conversion value of at least one difference value;
[0096] Based on the transformed value of the at least one difference value, a fitness value is determined.
[0097] Determining the fitness value of an individual based on at least one difference value of the individual can also be performed in a numerical conversion manner. First, the ratio between the absolute value of each difference value and the corresponding reference axial power offset at the same time step is determined, and based on the ratio of each difference value, each difference value is numerically converted to obtain a conversion value, and then the fitness value of the individual is determined based on all the conversion values.
[0098] Exemplarily, the ratio between the absolute value of each difference and the corresponding reference axial power offset can be expressed as:
[0099]
[0100] Where x is the above ratio, i is the number of the time step, AO i is the axial power offset, P i ×AO0 is the reference axial power offset, P i is the relative power, and AO0 is the axial power offset when the core is at full power.
[0101] In some implementations, based on the ratio corresponding to each difference value, the difference value is converted to obtain at least one conversion value of the difference value, which may include any of the following:
[0102] When the ratio corresponding to the difference is less than the first threshold, converting the difference into a first conversion value;
[0103] When the ratio corresponding to the difference is less than or equal to the second threshold and greater than or equal to the first threshold, converting the difference into a second conversion value;
[0104] When the ratio corresponding to the difference is greater than the second threshold, the difference is converted into a third conversion value.
[0105] Based on the magnitude relationship between the ratio corresponding to each difference and the first threshold and the second threshold, the difference is converted to obtain a conversion value of each difference, and the conversion value is one of a first conversion value, a second conversion value, and a third conversion value. When the ratio corresponding to the difference is less than the first threshold, the difference is converted to the first conversion value; when the ratio corresponding to the difference is less than or equal to the second threshold and greater than or equal to the first threshold, the difference is converted to the second conversion value; when the ratio corresponding to the difference is greater than the second threshold, the difference is converted to the third conversion value.
[0106] Exemplarily, the specific method of determining the conversion value can be expressed by the following formula:
[0107]
[0108] Wherein, x is the above ratio, f(x) is the conversion value, 1 is the first conversion value, 0 is the second conversion value, -1.5 is the third conversion value, 0.5% is the first threshold value, and 1% is the second threshold value.
[0109] It can be seen that, in the case of the conversion value determination method as in formula (2), if the difference between the axial power offset corresponding to a time step and the reference axial power offset is small (x<0.5%), then the conversion value corresponding to the time step is large (1); if the difference between the axial power offset corresponding to a time step and the reference axial power offset is large (0.5%≤x≤1% or x>1%), then the conversion value corresponding to the time step is small (0 or -1.5).
[0110] In some implementations, determining a fitness value based on a transformed value of at least one difference may include:
[0111] determining a maximum absolute value among the absolute values of at least one difference as a correction amount;
[0112] The converted values of at least one difference are summed, and the fitness value is determined according to the summed result and the correction amount.
[0113] The fitness value of each individual is determined based on the sum of the conversion values corresponding to all time steps of each individual. However, if the sum of all conversion values of each individual is simply used as the fitness value of each individual, some individuals will have the same fitness value, and accurate fitness evaluation cannot be performed. Therefore, it is necessary to correct the sum of the conversion values of each individual to increase the difference in fitness values between individuals.
[0114] The largest absolute value among the absolute values of all the above differences in the individual is taken as the correction amount, and the sum of all the conversion values of the individual is corrected by using the correction amount to obtain the correction value of the individual.
[0115] Exemplarily, the specific method of using the correction amount to correct the sum of all individual conversion values can be expressed as the following formula:
[0116]
[0117] Among them, f(x i ) is the conversion value determined according to the above formula (2), x i is the above ratio, i is the number of the time step, n is the number of time steps included in the individual, AO i is the axial power offset, P i ×AO0 is the reference axial power offset, P i is the relative power, AO0 is the axial power offset when the core is at full power, and AO i -P i ×AO0 is the above difference, |AO i -P i ×AO0| is the above absolute value, and k is the preset correction coefficient.
[0118] It can be seen that when the conversion value is determined by formula (2) and the sum of all individual conversion values is corrected by formula (3) to obtain the individual fitness value, the smaller the gap between all axial power offsets corresponding to the individual and the reference axial power offset as a whole, the larger the fitness value, the better the control strategy corresponding to the individual, and the optimal individual is the individual with the best fitness value in the dominant population when the iteration terminates, that is, the individual with the largest fitness value.
[0119] In some embodiments, dominant population generation may include:
[0120] According to the sorting results of all individuals, all individuals are divided into a first population and multiple second populations, wherein the sorting results are obtained based on the fitness values of the individuals, the individuals in the first population are used for factor retention, factor mutation and factor crossover, and the individuals in each second population are used for factor mutation and factor crossover;
[0121] Using the individuals in the first population and a plurality of second populations to perform factor crossover and factor mutation;
[0122] The dominant population is obtained by combining the individuals obtained by factor crossover and factor mutation from the first population and multiple individuals from the second population with the individuals that have undergone factor retention in the first population.
[0123] During the iteration process, each generation of the dominant population is based on the ranking results of all individual fitness values. The ranking results of the fitness values characterize the pros and cons of the control strategies corresponding to each individual. According to the ranking results, all individuals are first divided into populations, divided into the first population and multiple second populations. The control strategies corresponding to the individuals in the first population are better and can be used for factor retention, factor mutation and factor crossover. The control strategies corresponding to the multiple second populations are inferior to the control strategies corresponding to the first population and can be used for factor mutation and factor crossover.
[0124] In the process of generating the dominant population, the first population is used to perform factor crossover and factor mutation with multiple second populations to generate new individuals. The new individuals are combined with the individuals whose factors are retained in the first population (i.e., the individuals that are completely passed on to the next generation of the dominant population) as the dominant population.
[0125] Exemplarily, when the fitness value of an individual is higher and the control strategy represented by the individual is better, the fitness values of all individuals are sorted from large to small, and according to the sorting result, the individuals in the top 5% of the sorting result are taken as the first population, and the remaining individuals are further divided into multiple second individuals according to the sorting result. The total number of individuals in the generated dominant population is consistent with the total number of all individuals for generating the dominant population. In the generated dominant population, the proportion of factor-retained individuals, the proportion of new individuals obtained by factor mutation, and the proportion of new individuals obtained by factor crossover can be adaptively debugged and determined according to the optimization needs.
[0126] In some embodiments, the plurality of second populations may include a first subpopulation, a second subpopulation, and a third subpopulation, the first subpopulation, the second subpopulation, and the third subpopulation are divided according to the sorting results of the individuals in the initialization population, the individuals used for factor crossover in the first population, the first subpopulation, and the second subpopulation include elite individuals, when the elite individuals perform factor crossover with other individuals to obtain new individuals, the probability that the factors of the elite individuals are retained in the new individuals is a first preset probability, the first preset probability is greater than 0.5,
[0127] The factor crossover using individuals from the first population and multiple second populations may include:
[0128] Factor crossover is performed using individuals of the first population, the first subpopulation, and the second subpopulation, the probability that the individuals of the first population are determined to be elite individuals is the second preset probability, the probability that the individuals of the first subpopulation are determined to be elite individuals is the third preset probability, and the third preset probability is less than the second preset probability;
[0129] Factor crossover is performed using individuals from the first subpopulation, the second subpopulation, and the third subpopulation. The probability that an individual from the second subpopulation is determined as an elite individual is a fourth preset probability, and the probability that an individual from the third subpopulation is determined as an elite individual is 0. The fourth preset probability is less than the third preset probability.
[0130] The multiple second populations divided according to the fitness value sorting results include the first sub-population, the second sub-population and the third sub-population. The excellence of the control strategies represented by the individuals in the first sub-population, the second sub-population and the third sub-population decreases in sequence. It can be seen that the excellence of the control strategies represented by the individuals in the third sub-population is the lowest. Therefore, among all the individuals included in the first population, the first sub-population and the second sub-population, the individuals that undergo factor crossover include elite individuals, while the individuals that undergo factor crossover in the third sub-population do not include elite individuals.
[0131] An elite individual refers to an individual with a greater probability of retaining its own factors in the process of generating a new individual by cross-factor between two individuals. This probability is the first preset probability. According to the definition of elite individuals, the first preset probability is greater than 50%.
[0132] For example, the factors of individual A are (a1, a2, a3, a4, a5), and the factors of individual B are (b1, b2, b3, b4, b5). Individual A is an elite individual, and the first preset probability is 80%. That is, when individual A and individual B are crossed to obtain individual C, there is a 0.8 probability that the factors of individual A can be retained in individual C, that is, the possible forms of the factors of individual C are (a1, a2, a3, a4, b5).
[0133] When the individuals of the first population, the first sub-population, and the second sub-population are used for factor crossover, the probability that the individual of the first population with the best fitness performance is determined as the elite individual is the second preset probability, and the probability that the individual of the first sub-population with fitness performance inferior to that of the first population is determined as the elite individual is the third preset probability, and the third preset probability is less than the first preset probability;
[0134] When factor crossover is performed using individuals from the first subpopulation, the second subpopulation, and the third subpopulation, the probability that an individual from the second subpopulation whose fitness performance is inferior to that of the first subpopulation is determined as an elite individual is the fourth preset probability, and the fourth preset probability is less than the third preset probability. The individuals in the third subpopulation that undergo factor crossover do not include elite individuals, so the probability that the third subpopulation is determined as an elite individual is 0.
[0135] There will be no factorial crossover between individuals in the first subpopulation and individuals in the third subpopulation.
[0136] Exemplarily, when factor crossover is performed using individuals of the first population, the first subpopulation, the second subpopulation, and the third subpopulation, the probability that an individual in each population is determined to be an elite individual, and the probability that an individual in each population is determined to be an individual that performs factor crossover with an elite individual can be expressed by the following matrix C:
[0137]
[0138] Matrix C includes 4 columns. The diagonal elements from the 1st column to the 4th column - 0.5, 0.3, 0.2, 0, respectively represent the probability that individuals in the first population, individuals in the first sub-population, individuals in the second sub-population, and individuals in the third sub-population are determined as elite individuals, that is, the second preset probability is 0.5, the third preset probability is 0.3, and the fourth preset probability is 0.2, and individuals in the third sub-population will not be determined as elite individuals; the off-diagonal elements from the 1st column to the 4th column respectively represent the probability that individuals in the first population, individuals in the first sub-population, individuals in the second sub-population, and individuals in the third sub-population are determined as individuals that undergo factor crossover with elite individuals.
[0139] Taking the first row of matrix C as an example, the probability that an individual in the first population is determined as an elite individual is 0.5. When an individual in the first population is selected as an elite individual, the probability that an individual in the first subpopulation and the second subpopulation is determined as an individual that undergoes factor crossover with the elite individual is both 0.5; taking the second row of matrix C as an example, the probability that an individual in the first subpopulation is determined as an elite individual is 0.3. When an individual in the first subpopulation is selected as an elite individual, the probability that an individual in the first population and the second subpopulation is determined as an individual that undergoes factor crossover with the elite individual is 0.6 and 0.4, respectively.
[0140] In some embodiments, when a plurality of first predicted operating condition parameter groups corresponding to a plurality of first control parameter groups are predicted based on the core operating condition prediction model, before using the plurality of first predicted operating condition parameter groups as the initialization population, the method may further include:
[0141] Using a plurality of second control parameter groups of working conditions to be simulated, performing xenon oscillation simulation processing to obtain simulation data of each working condition to be simulated, wherein the plurality of working conditions to be simulated are generated based on a random strategy, each simulation data includes a second control parameter group and a second predicted working condition parameter group, each second control parameter group includes at least one second control parameter corresponding to at least one time step, and each second predicted working condition parameter group includes at least one second predicted working condition parameter corresponding to at least one time step;
[0142] Using the second predicted operating condition parameter group in each simulation data, the second control parameter group in the simulation data is labeled to generate training samples corresponding to the simulation data;
[0143] Based on training samples corresponding to simulation data of multiple operating conditions to be simulated, the model to be trained is iteratively trained to obtain a core operating condition prediction model.
[0144] For the training of the core condition prediction model, the staff can use a computer to calculate and determine the training samples of the core condition prediction model through xenon oscillation simulation processing. First, the electronic device can generate multiple different second control parameter groups according to the random strategy, that is, multiple core control strategies are generated. It can be understood by those skilled in the art that each second control parameter group also includes at least one second control parameter corresponding to at least one time step, and the parameter type included in the second control parameter is the same as the parameter type included in the first control parameter.
[0145] The electronic device uses the plurality of second control parameter groups to perform xenon oscillation simulation processing to obtain simulation data of each second control parameter group, wherein the simulation data includes a second predicted operating condition parameter group corresponding to the second control parameter group. It can be understood by those skilled in the art that each second predicted operating condition parameter group also includes at least one second predicted operating condition parameter corresponding to at least one time step, and the parameter type included in the second control parameter is the same as the parameter type included in the first predicted operating condition parameter.
[0146] The simulation data also includes a second control parameter set for the xenon oscillation simulation process.
[0147] Exemplarily, the parameter type of the second predicted operating condition parameter is the axial power offset, which is the same as the first control parameter and the first predicted operating condition parameter. In addition to the relative power of the core, the parameter type of the second control parameter may also include the time step length and the rod position of the temperature control rod. The core control strategy includes three time steps, a second predicted operating condition parameter group can be written as [AO1, AO2, AO3], and a second control parameter group can be written as [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)]. The electronic device generates a plurality of [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)] with different specific parameter values according to the random strategy, which are used for xenon oscillation simulation processing to obtain simulation data of each [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)]. The simulation data includes [AO1, AO2, AO3] corresponding to [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)], and also includes [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)] itself.
[0148] The second control parameter group in the simulation data is labeled using the second predicted operating condition parameter group in the acquired simulation data to generate training samples for model training, and the training model to be trained is iteratively trained using the training samples corresponding to the simulation data of multiple operating conditions to be simulated to obtain a core operating condition prediction model.
[0149] In some embodiments, the model to be trained is a convolutional neural network model. Before iteratively training the model to be trained based on training samples corresponding to simulation data of multiple operating conditions to be simulated to obtain a core operating condition prediction model, the method may further include:
[0150] The parameters to be added are added to the training sample, and together with the second control parameter group in the training sample, are used as the annotation objects of the second predicted operating condition parameter group in the training sample. The parameters to be added may include:
[0151] At least one target control parameter group associated with the second control parameter group in the training sample, the time steps corresponding to some parameters in the target control parameter group are located before the time steps corresponding to each parameter in the second control parameter group.
[0152] The network structure of the model to be trained, i.e., the core operating condition prediction model, can be a convolutional neural network (CNNs). Since the convolutional neural network has a strong ability to extract features, in order to enhance the accuracy of the model during training, the annotated part in the training sample is suitable for adding adjacent or preceding related features. Therefore, data related to the core control strategy can be added to the training sample, and the data is the parameter to be added. The parameter to be added is selected as at least one target control parameter group associated with the second control parameter group in the training sample, and the time step corresponding to some parameters in the target control parameter group is located before the time step corresponding to each parameter in the second control parameter group.
[0153] Exemplarily, the time step corresponding to some parameters in the target control parameter group is the previous or previous two time steps of the time step corresponding to each parameter in the second control parameter group. For example, the second control parameter group is expressed as [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)], and the target control parameter group of the second control parameter group can be determined as: [(P1, T1, S1), (P1, T1, S1), (P2, T2, S2)], that is, the target control parameter group includes the parameters of the previous time step of (P3, T3, S3) - (P2, T2, S2), the parameters of the previous time step of (P2, T2, S2) - (P1, T1, S1), and (P1, T1, S1) itself (because there is no parameter of the previous time step of (P1, T1, S1), so (P1, T1, S1) itself is added to the target control parameter group).
[0154] There can be multiple target control parameter groups corresponding to each second control parameter group, so the target parameter group in the parameters to be added can also be determined as: [(P1, T1, S1), (P2, T2, S2), (P1, T1, S1)], that is, the target control parameter group includes the parameters of the first two time steps of (P3, T3, S3) - (P1, T1, S1), as well as (P1, T1, S1) and (P2, T2, S2) themselves (because there are no parameters of the first two time steps of (P1, T1, S1) and (P2, T2, S2), so (P1, T1, S1) and (P2, T2, S2) themselves are added to the target control parameter group).
[0155] Those skilled in the art can understand that when the network structure of the core condition prediction model is a network structure other than a convolutional neural network, the types of parameters to be added can also be adaptively adjusted according to the network structure type of the core condition prediction model.
[0156] In some embodiments, each simulation data includes parameters related to a first number of time steps, each training sample includes parameters related to a second number of time steps, and the first number is greater than the second number.
[0157] Using the second predicted operating condition parameter group in each simulation data, labeling the second control parameter group in the simulation data to generate training samples corresponding to the simulation data may include:
[0158] In the time steps involved in the parameters included in the simulation data, taking any time step as a starting point, selecting a plurality of groups of the second number of time steps;
[0159] The second control parameter set corresponding to each group of the second number of time steps is set as a sample to be labeled;
[0160] The second predicted operating condition parameter set corresponding to each group of the second number of time steps is set as a labeled sample;
[0161] The corresponding samples to be labeled are labeled using each labeled sample to generate training samples corresponding to the simulated data.
[0162] In order to expand the number of training samples, when performing xenon oscillation simulation processing, the number of time steps involved in the parameters in the simulation data can be set to be greater than the number of time steps of the required control strategy, that is, greater than the number of time steps involved in the parameters in the training samples. In an embodiment of the present application, the parameters included in the simulation data involve a first number of time steps, and the parameters included in the training samples involve a second number of time steps.
[0163] In this way, multiple groups of second number of time steps can be selected from any time step as the starting point in the simulation data, and the second control parameter set corresponding to each group of second number of time steps is set as a sample to be labeled, and the second predicted operating condition parameter set corresponding to each group of second number of time steps is set as a labeled sample, and finally the corresponding samples to be labeled are labeled using each labeled sample to generate training samples corresponding to the simulation data. In this way, the number of training samples can be greater than the number of simulation data, and the number of training samples is expanded, so that the trained core operating condition prediction model has stronger generalization ability.
[0164] Exemplarily, the first number is 4, the second number is 3, and the second control parameter group included in the simulation data can be expressed as [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3), (P4, T4, S4)]. In [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3), (P4, T4, S4)], taking any time step as the starting point, multiple groups of second control parameters involving 3 time steps are selected, and the selected groups of second control parameters involving 3 time steps are: [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)], and [(P2, T2, S2), (P3, T3, S3), (P4, T4, S4)].
[0165] Correspondingly, the groups of second predicted operating condition parameters involving three time steps selected in the simulation data are: [AO1, AO2, AO3], and [AO2, AO3, AO4].
[0166] Use [AO1, AO2, AO3] to annotate [(P1, T1, S1), (P2, T2, S2), (P3, T3, S3)] to form the first training sample of the simulated data, and use [AO2, AO3, AO4] to annotate [(P2, T2, S2), (P3, T3, S3), (P4, T4, S4)] to form the second training sample of the simulated data. That is, using one simulated data, two different training samples are obtained, thereby expanding the number of training samples.
[0167] See also Figure 2 , is a schematic diagram of the structure of a nuclear reactor core control device provided in an embodiment of the present application. In a second aspect of an embodiment of the present application, a nuclear reactor core control device 10 is provided. The device 10 includes:
[0168] An initialization population determination module 11 is used for, when a plurality of first predicted operating condition parameter groups corresponding to a plurality of first control parameter groups are predicted based on a core operating condition prediction model, using the plurality of first predicted operating condition parameter groups as the initialization population, wherein each first control parameter group includes at least one first control parameter corresponding to at least one time step, each first predicted operating condition parameter group includes at least one first predicted operating condition parameter corresponding to at least one time step, each first control parameter group corresponds to an individual in the initialization population, a factor in the individual is a first control parameter corresponding to each time step in the first control parameter group, a parameter type of the first predicted operating condition parameter includes an axial power offset of the core, and a parameter type of the first control parameter includes a relative power of the core;
[0169] The optimal individual acquisition module 12 is used to perform adaptive evaluation and iterative processing of dominant population generation based on the fitness value of each individual in the initialization population to obtain the optimal individual, wherein the fitness value of each individual is determined by using an adaptive evaluation function, and the adaptive evaluation function involves the difference between the axial power offset corresponding to each time step in the first prediction operating condition parameter group and the reference axial power offset corresponding to each time step, and the reference axial power offset is determined according to the corresponding relative power and the axial power offset when the core is at full power. The dominant population generation is used to generate a dominant population according to the result of the adaptive evaluation, and the optimal individual is an individual in the dominant population generated when the iterative processing satisfies the iteration termination condition;
[0170] The core control module 13 is used to perform core control using the first control parameter corresponding to the optimal individual.
[0171] The nuclear reactor core control device 10 provided in the second aspect of the embodiment of the present application can implement the various processes implemented in the above-mentioned method embodiment and achieve the same beneficial effects. To avoid repetition, it will not be described here.
[0172] See also Figure 3 , is a structural diagram of an electronic device provided in an embodiment of the present application. The third aspect of an embodiment of the present application provides an electronic device 1000, including a processor 1100 and a memory 1200. The memory 1200 stores machine executable instructions that can be executed by the processor 1100. The processor 1100 can execute the machine executable instructions to implement the above-mentioned nuclear reactor core control method.
[0173] A fourth aspect of an embodiment of the present application provides a machine-readable storage medium, on which instructions are stored. When the instructions are executed by a processor, the processor implements the above-mentioned nuclear reactor core control method.
[0174] In some embodiments, the embodiments of the present application also provide a computer program product, including a computer program, which, when executed by a processor, implements the nuclear reactor core control method according to the above-mentioned embodiment.
[0175] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.
[0176] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0177] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0178] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.
[0179] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.
[0180] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0181] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.
[0182] In addition, various embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the present invention, they should also be regarded as the contents disclosed by the present invention.
Claims
1. A nuclear reactor core control method, characterized in that: The method comprises: In the case where a plurality of first predicted operating condition parameter groups corresponding to a plurality of first control parameter groups are predicted based on a core operating condition prediction model, the plurality of first predicted operating condition parameter groups are used as an initialization population, wherein each of the first control parameter groups includes at least one first control parameter corresponding to at least one time step, each of the first predicted operating condition parameter groups includes at least one first predicted operating condition parameter corresponding to at least one time step, each of the first control parameter groups corresponds to an individual in the initialization population, a factor in the individual is a first control parameter corresponding to each time step in the first control parameter group, a parameter type of the first predicted operating condition parameter includes an axial power offset of the core, and a parameter type of the first control parameter includes a relative power of the core; Based on the fitness value of each individual in the initialization population, an iterative process of fitness evaluation and dominant population generation is performed to obtain an optimal individual, wherein the fitness value of each individual is determined by using an fitness evaluation function, and the fitness evaluation function involves the difference between the axial power offset corresponding to each time step in the first prediction operating condition parameter group and the reference axial power offset corresponding to each time step, and the reference axial power offset is determined according to the corresponding relative power and the axial power offset when the core is at full power, and the dominant population generation is used to generate a dominant population according to the result of the fitness evaluation, and the optimal individual is an individual in the dominant population generated when the iterative process satisfies the iteration termination condition; The core is controlled using the first control parameter corresponding to the optimal individual.
2. The method according to claim 1, characterized in that The fitness value is determined by the following steps: Determine the reference axial power offset corresponding to each time step; Determine the difference between the axial power offset corresponding to each time step in the first predicted operating condition parameter group and the reference axial power offset corresponding to each time step to obtain at least one difference, wherein different differences correspond to different time steps; Based on the at least one difference value, the fitness value is determined.
3. The method according to claim 2, characterized in that The determining the fitness value based on the at least one difference value comprises: determining a ratio between an absolute value of each of the differences and a corresponding reference axial power offset; Based on the ratio corresponding to each of the difference values, convert the difference values to obtain a conversion value of the at least one difference value; The fitness value is determined based on a converted value of the at least one difference value.
4. The method according to claim 3, characterized in that The converting the difference values based on the ratio corresponding to each of the difference values to obtain the converted value of the at least one difference value includes any one of the following: When the ratio corresponding to the difference is less than a first threshold, converting the difference into a first conversion value; When the ratio corresponding to the difference is less than or equal to the second threshold and greater than or equal to the first threshold, converting the difference into a second conversion value; When the ratio corresponding to the difference is greater than the second threshold, the difference is converted into a third conversion value.
5. The method according to claim 3, characterized in that: The determining the fitness value based on the conversion value of the at least one difference value comprises: determining a maximum absolute value among the absolute values of the at least one difference as a correction amount; The converted value of the at least one difference is summed, and the fitness value is determined according to the summed result and the correction amount.
6. The method according to claim 1, characterized in that The generation of the dominant population includes: According to the sorting results of all the individuals, all the individuals are divided into a first population and a plurality of second populations, wherein the sorting results are obtained based on the fitness values of the individuals, the individuals in the first population are used for factor retention, factor mutation and factor crossover, and the individuals in each of the second populations are used for factor mutation and factor crossover; Using the individuals in the first population and the plurality of second populations to perform factor crossover and factor mutation; The dominant population is obtained by combining individuals obtained by factor crossover and factor mutation using individuals in the first population and the multiple second populations with individuals in the first population that have undergone factor retention.
7. The method according to claim 6, characterized in that The multiple second populations include a first subpopulation, a second subpopulation, and a third subpopulation. The first subpopulation, the second subpopulation, and the third subpopulation are divided according to the sorting results of the individuals in the initialization population. In the first population, the first subpopulation, and the second subpopulation, the individuals used for factor crossover include elite individuals. When the elite individuals perform factor crossover with other individuals to obtain new individuals, the probability that the factors of the elite individuals are retained in the new individuals is a first preset probability, and the first preset probability is greater than 0.
5. The step of performing factor crossover using individuals in the first population and the plurality of second populations comprises: Factor crossover is performed using individuals of the first population, the first sub-population, and the second sub-population, the probability that an individual of the first population is determined to be the elite individual is a second preset probability, the probability that an individual of the first sub-population is determined to be the elite individual is a third preset probability, and the third preset probability is less than the second preset probability; Factor crossover is performed using individuals from the first sub-population, the second sub-population, and the third sub-population, the probability that an individual from the second sub-population is determined as the elite individual is a fourth preset probability, the probability that an individual from the third sub-population is determined as the elite individual is 0, and the fourth preset probability is less than the third preset probability.
8. The method according to claim 1, characterized in that In the case where a plurality of first predicted operating condition parameter groups corresponding to a plurality of first control parameter groups are predicted based on the core operating condition prediction model, before using the plurality of first predicted operating condition parameter groups as the initialization population, the method further includes: Using a plurality of second control parameter groups of working conditions to be simulated, performing xenon oscillation simulation processing to obtain simulation data of each of the working conditions to be simulated, wherein the plurality of working conditions to be simulated are generated based on a random strategy, each of the simulation data includes the second control parameter group and a second predicted working condition parameter group, each of the second control parameter groups includes at least one second control parameter corresponding to at least one time step, and each of the second predicted working condition parameter groups includes at least one second predicted working condition parameter corresponding to at least one time step; Using the second predicted operating condition parameter groups in each of the simulation data, the second control parameter groups in the simulation data are labeled to generate training samples corresponding to the simulation data; Based on the training samples corresponding to the simulation data of the multiple operating conditions to be simulated, the model to be trained is iteratively trained to obtain the core operating condition prediction model.
9. The method according to claim 8, characterized in that The model to be trained is a convolutional neural network model. Before iteratively training the model to be trained based on training samples corresponding to the simulation data of the multiple operating conditions to be simulated to obtain the core operating condition prediction model, the method further includes: The parameters to be added are added to the training sample, and together with the second control parameter group in the training sample, serve as the annotation objects of the second predicted operating condition parameter group in the training sample, wherein the parameters to be added include: At least one target control parameter group associated with the second control parameter group in the training sample, the time steps corresponding to some parameters in the target control parameter group are located before the time steps corresponding to each parameter in the second control parameter group.
10. The method according to claim 8, characterized in that Each of the simulation data includes parameters related to a first number of time steps, and each of the training samples includes parameters related to a second number of time steps, the first number being greater than the second number, The method of using the second predicted operating condition parameter group in each of the simulation data to label the second control parameter group in the simulation data to generate training samples corresponding to the simulation data includes: In the time steps involved in the parameters included in the simulation data, taking any time step as a starting point, selecting a plurality of groups of the second number of time steps; Set each group of second control parameters corresponding to the second number of time steps as a sample to be labeled; The second predicted operating condition parameter set corresponding to each group of the second number of time steps is set as a labeled sample; The corresponding samples to be labeled are labeled using each of the labeled samples to generate training samples corresponding to the simulated data.
11. A nuclear reactor core control device, characterized in that: The device comprises: An initialization population determination module is used for, when a plurality of first predicted operating condition parameter groups corresponding to a plurality of first control parameter groups are predicted based on a core operating condition prediction model, using the plurality of first predicted operating condition parameter groups as an initialization population, wherein each of the first control parameter groups includes at least one first control parameter corresponding to at least one time step, each of the first predicted operating condition parameter groups includes at least one first predicted operating condition parameter corresponding to at least one time step, each of the first control parameter groups corresponds to an individual in the initialization population, a factor in the individual is a first control parameter corresponding to each time step in the first control parameter group, a parameter type of the first predicted operating condition parameter includes an axial power offset of the core, and a parameter type of the first control parameter includes a relative power of the core; an optimal individual acquisition module, for performing an iterative process of fitness evaluation and dominant population generation based on the fitness value of each individual in the initialization population, to obtain an optimal individual, wherein the fitness value of each individual is determined by using an adaptability evaluation function, the adaptability evaluation function involves a difference between an axial power offset corresponding to each time step in the first prediction operating condition parameter group and a reference axial power offset corresponding to each time step, the reference axial power offset is determined according to a corresponding relative power and an axial power offset at full core power, the dominant population generation is used to generate a dominant population according to a result of the adaptability evaluation, and the optimal individual is an individual in the dominant population generated when the iterative process satisfies an iterative termination condition; The core control module is used to perform core control using the first control parameter corresponding to the optimal individual.
12. An electronic device, characterized in that: include: a memory configured to store instructions; as well as A processor is configured to call the instructions from the memory and implement the nuclear reactor core control method according to any one of claims 1 to 10 when executing the instructions.
Citation Information
Patent Citations
Power tilt prediction method and device, equipment, storage medium and program product
CN115660173A
Oil reservoir well location optimization method, electronic equipment and storage medium
CN115688600A
Wind power prediction method and device, electronic equipment and readable storage medium
CN117277314A
Reactor operation optimization method based on improved multi-population genetic algorithm
US20230274848A1
Method and system for determining nuclear reactor core control blade positioning
US5790616A
Cited By
Nuclear reactor xenon oscillation prediction method based on data assimilation and physical constraint
CN120596779A
A method for predicting xenon oscillations in nuclear reactors based on data assimilation and physical constraints
CN120596779B
Control method for axial power distribution of pressurized water reactor, and apparatus
WO2026001304A1
