Nuclear reactor core digital twin optimization method based on data fusion

Abstract

A data fusion-based digital twin optimization method for nuclear reactor cores integrates low-resolution sensor monitoring data from the reactor core with a nuclear thermal coupling numerical simulation model (NMC) to be optimized, using ensemble Kalman filtering to perform real-time correction of key parameters. This results in a high-resolution, high-precision digital twin image of the reactor core, enabling centimeter-level monitoring of the reactor core during the online phase. This invention fully considers the complexity and heterogeneous nature of data in nuclear reactor core digital twin technology. By fusing low-resolution, multi-source, heterogeneous in-core sensor monitoring data with a high-resolution NMC model and optimizing the model's parameters, the real-time low-resolution sensor monitoring data is integrated into the results of the high-resolution numerical simulation model. Through real-time updates and continuous optimization of the numerical simulation model parameters as the reactor operates, a high-resolution digital twin image of the reactor core is achieved, reducing the cumulative error of the digital twin.

Description

Technical Field

[0001] This invention relates to a technology in the field of nuclear reactor control, specifically a digital twin optimization method for nuclear reactor core based on data fusion. Background Technology

[0002] Current nuclear reactor core condition monitoring relies on low spatial resolution neutron flux and temperature sensors. High-resolution nuclear thermal coupling numerical simulation models derived from core design and safety analysis cannot predict actual fluctuations in reactor operating conditions, thus hindering their application in core condition monitoring. Existing nuclear reactor critical parameter inversion techniques fail to integrate measured data from actual reactor operation with digital twin system model data. This prevents the digital twin system from updating and optimizing based on the actual reactor condition, consequently hindering effective monitoring of the reactor core's operating status. Summary of the Invention

[0003] To address the aforementioned shortcomings of existing technologies, this invention proposes a data fusion-based optimization method for digital twins of nuclear reactor cores. This method fully considers the complexity and heterogeneous nature of data in nuclear reactor core digital twin technology. It achieves data fusion and parameter correction optimization by combining low-resolution, heterogeneous in-core sensor monitoring data with a high-resolution nuclear thermal coupling numerical simulation model. Real-time low-resolution sensor monitoring data is integrated into the results of the high-resolution numerical simulation model. Through real-time updates and continuous optimization of the numerical simulation model parameters as the reactor operates, a high-resolution digital twin image of the nuclear reactor core is achieved, reducing the cumulative error of the digital twin.

[0004] This invention is achieved through the following technical solution:

[0005] This invention relates to a data fusion-based digital twin optimization method for nuclear reactor cores. The method involves fusing low-resolution sensor monitoring data from the nuclear reactor core with the nuclear thermal coupling numerical simulation model to be optimized using ensemble Kalman filtering. This process allows for real-time correction of key parameters in the nuclear thermal coupling numerical simulation model, resulting in a high-resolution, high-precision digital twin image of the nuclear reactor core. This enables centimeter-level monitoring of the real-time operating nuclear reactor core during the online phase.

[0006] The key parameters of the nuclear thermal coupling numerical simulation model to be optimized are obtained in the following way: based on the monitoring of the nuclear thermal state of the reactor core by in-core sensors, sensitivity analysis is performed on the parameters in the nuclear thermal coupling numerical simulation model of the reactor core, and parameters that have an important impact on the calculation results of the nuclear reactor core monitoring variables and the macroscopic nuclear thermal performance are screened out. Finally, a set of sensitive model parameters is formed, which are the key parameters of the nuclear thermal coupling numerical simulation model to be optimized.

[0007] The low-resolution sensors inside the nuclear reactor core include: a neutron flux density detector and a coolant temperature sensor.

[0008] The sensor monitoring data is obtained by reading the signals from the sensors arranged inside the reactor to obtain the Z signals of each sensor at time t. t The correspondence between sensor signals and sensor monitoring data can be linearly represented by the observation matrix H, and the actual noise present in sensor measurements can be represented by the covariance matrix R of the measurement noise.

[0009] The data fusion method based on ensemble Kalman filtering described herein employs, but is not limited to, the Kalman filtering method described in Chinese Patent Document No. CN108491974A. The key parameters of the nuclear thermal coupling numerical simulation model to be optimized and the state prediction data calculated by the nuclear thermal coupling numerical simulation model are used as the state variable space and input into the ensemble Kalman filter for iterative calculation to obtain the key parameters of the optimized nuclear thermal coupling numerical simulation model.

[0010] The state variable space is obtained as follows: For the key parameters of the nuclear thermal coupling numerical simulation model to be optimized, the relative variance of their perturbation values ​​(e.g., 5%) is determined. Based on the Gaussian distribution, the perturbed key parameters are sampled (using methods such as Latin hypercube sampling) to generate n sample sets of key parameters. These samples, together with the nuclear reactor core state variables at time t-1 calculated by the nuclear thermal coupling numerical simulation model, constitute n state variable vectors X. t-1，i (1≤i≤n) forms the state variable space; the above-mentioned nuclear reactor core state variables are the state variables that the sensors can monitor.

[0011] The data fusion method based on ensemble Kalman filtering is specifically as follows: For the n state variables X in the state variable space at time t-1... t-1，i The calculated values ​​X of n state variables in the state variable space at time t are obtained by using a nuclear thermal coupling numerical simulation model. t，i (1≤i≤n); Calculate the optimal prediction of the state variables at time t. in: The optimal prediction of the state variables at time t includes the optimal prediction of the key parameters of the model, K. e It is the Kalman gain and has K e =PH T (HPH T +R) -1 P is the optimal prediction at time t. The error covariance matrix and have X t The average of the values ​​of the n state vectors at time t is calculated as follows: R is the covariance matrix of the measurement noise; when the optimal predicted state variable space satisfies the iterative convergence condition of the ensemble Kalman filter, it is considered that one data fusion has been completed, and the optimal prediction of the key parameters of the nuclear thermal coupling numerical simulation model is obtained; otherwise, The above calculation process is repeated as the state variable at time t-1 until the iterative convergence condition is met.

[0012] The aforementioned high-resolution nuclear reactor digital twin mirror image, under the condition that the numerical simulation results and sensor measurement data are in good agreement, constructs a digital twin mirror image of the nuclear reactor core to achieve the goal of reducing the cumulative error of the digital twin mirror image and improving its accuracy. Specifically, it involves replacing the corresponding parameters in the original model with the key parameters of the optimized nuclear thermal coupling numerical simulation model to obtain the optimized model, forming an updated digital twin mirror image of the nuclear reactor core. The numerical simulation calculation is then performed again using the updated digital twin mirror image. If it shows a high degree of consistency with the actual operating conditions, the optimization process is considered complete. Furthermore, as the nuclear reactor operates, data fusion of the nuclear thermal coupling numerical simulation model is continuously performed to optimize key parameters and update the digital twin mirror image.

[0013] The aforementioned digital twin image of the nuclear reactor core accurately tracks the core's operating status in real time. Specific parameters include: fuel rod power, three-dimensional distribution of coolant temperature, neutron / photon flux length, high-resolution burnup data, core power distribution, component outlet temperature, critical boron concentration, axial power offset, and deviation from the bubble-nucleation boiling ratio.

[0014] The aforementioned digital twin image of the nuclear reactor core operates with the nuclear reactor, continuously performing data fusion and model optimization. The time interval between each optimization is 20s-30s, which can achieve 2-3 data assimilations within one minute to reduce cumulative errors and improve simulation accuracy.

[0015] This invention relates to a digital twin image of a nuclear reactor core obtained by the above-mentioned method, comprising: a sensor information collection and processing unit, a nuclear thermal coupling numerical analysis unit, and a data fusion unit, wherein: the sensor information collection and processing unit is used to acquire the digital analog signals of the nuclear reactor core sensors and process them to obtain core nuclear thermal state data; the nuclear thermal coupling numerical analysis unit calculates the core state at the next moment through numerical simulation based on the sensor monitoring data and the high-resolution nuclear thermal distribution parameters of the data fusion unit; the data fusion unit fuses the core nuclear thermal state data from the sensors and the high-resolution nuclear thermal coupling numerical simulation results to obtain the key parameters of the optimized nuclear thermal coupling numerical simulation model. Technical effect

[0016] This invention fuses low-resolution sensor monitoring data with high-resolution nuclear thermal coupling numerical simulation model data, corrects key parameters of the nuclear thermal coupling numerical simulation model, and forms a high-resolution, high-precision digital twin image of the nuclear reactor core, which has higher adaptability to nuclear reactor core systems. Compared with existing technologies, this invention has high real-time simulation accuracy and can predict fuel assembly power with an error within 8%. The data synchronization time interval is short, enabling 2-3 data synchronizations within one minute, reducing the cumulative error of the digital twin image of the nuclear reactor core. Attached Figure Description

[0017] Figure 1 This is a system flowchart of the present invention.

[0018] Figure 2 This is a flowchart of the data fusion method of the present invention.

[0019] Figure 3 This is a rendering of the invention;

[0020] In the figure: 1. Arrangement of fuel assemblies in the reactor core involved in this embodiment; 2. Monitoring rod with sensors; in traditional monitoring, the sensors in the monitoring rod feed back monitoring data with a resolution of component level, i.e., decimeter level, to the monitoring system; 3. Enlarged view of a part of the fuel assembly; 4. Single fuel rod; 5. Calculation accuracy of the nuclear thermal coupling numerical simulation image, fuel rod level, i.e., centimeter level. Detailed Implementation

[0021] like Figure 1 As shown in this embodiment, a data fusion-based digital twin optimization method for nuclear reactor cores is applied to the construction of a digital twin image of the core of a 1400 MWe-class third-generation advanced pressurized water reactor and the core condition monitoring and maintenance process. In this embodiment, the nuclear reactor is equipped with nuclear thermal sensors such as neutron flux density detectors and coolant temperature sensors. After the nuclear reactor begins operation, based on the core layout, the initial values ​​of each system in the nuclear reactor, and the monitoring data from the nuclear thermal sensors, a nuclear thermal coupling numerical simulation model is used to perform core-oriented nuclear thermal coupling simulation calculations. Subsequently, based on the sensor monitoring data, sensitivity analysis is performed on the parameters in the nuclear thermal coupling numerical simulation model of the nuclear reactor core, and parameters that have a significant impact on the calculation results of nuclear reactor core monitoring variables and nuclear thermal macroscopic performance are selected, forming a key model parameter set. Based on the results of the above sensitivity analysis, a state variable space composed of the key model parameter set and nuclear reactor core state variables, and a monitoring variable space composed of sensor monitoring data are constructed. The data from the state variable space and the monitoring variable space are then fed into an ensemble Kalman filter for data fusion processing.

[0022] like Figure 2 As shown, the data fusion process includes:

[0023] S1, based on the sensitivity analysis results of the parameters of the nuclear thermal coupling numerical simulation model, a 5% relative variance of the perturbation values ​​is determined for the key model parameters. Latin hypercube sampling is performed on the perturbation parameters based on a Gaussian distribution to generate n sample sets of the key model parameters. These samples, together with the reactor core state variables at time t-1, form n state variable vectors X. t-1，i (1≤i≤n);

[0024] The state variables include: fuel rod power, three-dimensional distribution of coolant temperature, medium / photon flux length, high-resolution burnup data, core power distribution, assembly outlet temperature, critical boron concentration, axial power offset, and deviation from nucleation boiling ratio.

[0025] S2, the state vector X at time t-1 t-1，i The n calculated values ​​X of the state vector at time t were obtained by using a nuclear thermal coupling numerical simulation model. t，i (1≤i≤n);

[0026] S3, read the signals from the sensors located inside the reactor to obtain the sensor signal Z at time t. t Based on the sensor characteristics, the observation matrix H is used to linearly represent the correspondence between the sensor signal and the sensor monitoring data. The actual noise present in the sensor measurement can be represented by the covariance matrix R of the measurement noise.

[0027] S4, calculate the average value of the n state vectors at time t. Used for subsequent data fusion steps;

[0028] S5, calculate the error of the calculated value of the state vector at time t, that is, the covariance matrix of the state vector at time t.

[0029] S6, Calculate the Kalman gain K e =PH T (HPH T +R) -1 ;

[0030] S7, obtain the predicted value of the state vector at time t. in The optimal prediction of the state vector at time t;

[0031] S8. Repeat the calculations from S1 to S7. When the calculation results satisfy the iterative convergence condition of the ensemble Kalman filter, it is considered that one data fusion has been completed and the optimal prediction of the state variables and key model parameters has been obtained.

[0032] S9. Replace the corresponding parameters in the proto-nuclear thermal coupling numerical simulation model with the model parameters in the above optimal prediction to obtain the optimized model, forming a high-resolution and high-precision digital twin image of the nuclear reactor core.

[0033] Finally, as the nuclear reactor operates, the calculation process from S1 to S8 is repeated to continuously optimize the key parameters of the nuclear thermal coupling numerical simulation model, update the digital twin image, and ultimately construct a digital twin image of the nuclear reactor core with small cumulative error and high accuracy.

[0034] Through specific practical applications, such as Figure 3 As shown, this method fuses sensor monitoring data with a nuclear thermal coupling numerical simulation model. The digital twin image of the nuclear reactor core using this method can improve the monitoring accuracy of the nuclear reactor core from the decimeter level to the centimeter level.

[0035] Through specific application examples, in the normal operating conditions and expected operating conditions of the nuclear reactor involved in this embodiment, by comparing the actual operating state of the nuclear reactor core with the digital twin mirror data, it was calculated that the predicted power difference for components with relative power > 0.9 is < 5%; and the predicted power difference for components with relative power < 0.9 is < 8%. Through actual testing, this method can achieve 2-3 digital twin mirror updates within 1 minute.

[0036] The digital twin image of the nuclear reactor core, after the above steps, has high predictive accuracy for the operation of the nuclear reactor core involved in this embodiment. It features high real-time simulation accuracy, short data synchronization time interval, and small cumulative error of the digital twin image. After the construction and optimization of the core digital twin image are completed, it can be put into use for monitoring the actual operating status of the nuclear reactor core, core operation and maintenance, and fault monitoring.

[0037] This method addresses the low-resolution and heterogeneous nature of sensor monitoring data within nuclear reactor cores. It integrates low-resolution sensor monitoring data with high-resolution nuclear thermal coupling numerical simulation models in real time using a given ensemble Kalman filter method. This optimizes the key parameters of the numerical simulation model, resulting in an optimized nuclear thermal coupling numerical simulation model, and ultimately forming a high-resolution, high-precision digital twin image of the nuclear reactor core.

[0038] The above-described specific implementations can be partially adjusted by those skilled in the art in different ways without departing from the principles and purpose of the present invention. The scope of protection of the present invention is defined by the claims and is not limited to the above-described specific implementations. All implementation schemes within the scope of the claims are bound by the present invention.

Claims

1. A method for optimizing a nuclear reactor core using a digital twin based on data fusion, characterized in that, The data is fused with ensemble Kalman filtering based on low-resolution sensor monitoring data inside the nuclear reactor core and the nuclear thermal coupling numerical simulation model to be optimized. The key parameters of the nuclear thermal coupling numerical simulation model are corrected in real time to form a high-resolution and high-precision digital twin image of the nuclear reactor core. Centimeter-level monitoring of the real-time operating nuclear reactor core is then performed in the online phase. The high-resolution, high-precision digital twin image of the nuclear reactor core is obtained by replacing the corresponding parameters in the original model with the key parameters of the optimized nuclear thermal coupling numerical simulation model, thus forming an updated digital twin image of the nuclear reactor core. The numerical simulation calculation is then performed again using the updated digital twin image. If it shows a high degree of consistency with the actual operating conditions, the optimization process is considered complete. Furthermore, as the nuclear reactor operates, data fusion is continuously performed on the nuclear thermal coupling numerical simulation model to optimize key parameters and update the model. The data fusion specifically includes: S1, based on the sensitivity analysis results of the nuclear thermal coupling numerical simulation model parameters, for key model parameters, a 5% relative variance of the perturbation values ​​is determined. Latin hypercube sampling is then performed on the perturbation parameters based on a Gaussian distribution to generate the key model parameters. A sample set, and with The state variables of the nuclear reactor core at any given time are combined to form the state variables of the reactor core. A state vector ); S2, for State vector at time step Calculations were performed using a nuclear thermal coupling numerical simulation model to obtain the results for... The state vector at time step Calculated values ); S3, reads signals from sensors located inside the reactor, and obtains... Time sensor signal Based on sensor characteristics, use the observation matrix The linear representation represents the correspondence between sensor signals and sensor monitoring data. The actual noise present in sensor measurements is represented by the covariance matrix of the measurement noise. express; S4, find time The average value of each state vector is calculated. This is used in subsequent data fusion steps; S5, Calculation The error in calculating the value of the state vector at time step, i.e. Covariance matrix of the state vector at time step ; S6, Calculate Kalman gain ; S7, obtained Predicted value of the state vector at time step ,in for Optimal prediction of the state vector at time step; S8. Repeat the calculations from S1 to S7. When the calculation results satisfy the iterative convergence condition of the ensemble Kalman filter, it is considered that one data fusion has been completed and the optimal prediction of the state variables and key model parameters has been obtained. S9. Replace the corresponding parameters in the proto-nuclear thermal coupling numerical simulation model with the model parameters in the above optimal prediction to obtain the optimized model, forming a high-resolution and high-precision digital twin image of the nuclear reactor core. Finally, as the nuclear reactor operates, the calculation process from S1 to S8 is repeated to continuously optimize the key parameters of the nuclear thermal coupling numerical simulation model, update the digital twin image, and ultimately construct a digital twin image of the nuclear reactor core with small cumulative error and high accuracy.

2. The method for optimizing a nuclear reactor core based on data fusion according to claim 1, characterized in that, The key parameters of the nuclear thermal coupling numerical simulation model to be optimized are obtained in the following way: based on the monitoring of the nuclear thermal state of the reactor core by in-core sensors, sensitivity analysis is performed on the parameters in the nuclear thermal coupling numerical simulation model of the reactor core, and parameters that have an important impact on the calculation results of the nuclear reactor core monitoring variables and the macroscopic nuclear thermal performance are screened out. Finally, a set of sensitive model parameters is formed, which are the key parameters of the nuclear thermal coupling numerical simulation model to be optimized.

3. The method for optimizing a nuclear reactor core based on data fusion according to claim 1, characterized in that, The sensor monitoring data is obtained by reading the signals from sensors located within the reactor. Signals from each sensor at any time The correspondence between sensor signals and sensor monitoring data is expressed using an observation matrix. Linear representation: The actual noise present in sensor measurements is expressed through the covariance matrix of the measurement noise. express.

4. The method for optimizing a nuclear reactor core based on data fusion according to claim 1, characterized in that, For the key parameters of the nuclear thermal coupling numerical simulation model to be optimized, the relative variance of their perturbation values ​​is determined. Based on a Gaussian distribution, the perturbed key parameters are sampled to generate the key parameters. A sample set, and calculated using a nuclear thermal coupling numerical simulation model. The state variables of the nuclear reactor core at any given time constitute the total state of the reactor core. A state vector These form the state variable space; the aforementioned nuclear reactor core state variables are the state variables that the sensors can monitor.

5. The method for optimizing a nuclear reactor core based on data fusion according to claim 1, characterized in that, The data fusion method based on ensemble Kalman filtering specifically involves: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] In the state variable space at time 1 State variables Calculations were performed using a nuclear thermal coupling numerical simulation model, and the results were obtained. Time-state variable space Calculated values ​​of each state variable );calculate Optimal prediction of state variables at time step ,in: for The optimal prediction of the state variables at time step, including the optimal prediction of key model parameters. For Kalman gain and have , for Time-optimal prediction The error covariance matrix and have , time The average of the values ​​of each state vector is calculated, i.e. , To measure the covariance matrix of the noise; when the optimal predicted state variable space satisfies the iterative convergence condition of the ensemble Kalman filter, it is considered that one data fusion has been completed, and the optimal prediction of the key parameters of the nuclear thermal coupling numerical simulation model is obtained; otherwise, As The state variable at time t is calculated repeatedly until the iterative convergence condition is met.

6. The method for optimizing a nuclear reactor core based on data fusion according to claim 1, characterized in that, The aforementioned digital twin image of the nuclear reactor core accurately tracks the core's operating status in real time. Specific parameters include: fuel rod power, three-dimensional distribution of coolant temperature, neutron / photon flux length, high-resolution burnup data, core power distribution, component outlet temperature, critical boron concentration, axial power offset, and deviation from the bubble-nucleation boiling ratio.

7. A digital twin image of a nuclear reactor core obtained by the data fusion-based optimization method for nuclear reactor core digital twins according to any one of claims 1-6, characterized in that, include: The system comprises a sensor information collection and processing unit, a nuclear thermal coupling numerical analysis unit, and a data fusion unit. Specifically: the sensor information collection and processing unit acquires analog signals from the reactor core sensors and processes them to obtain core nuclear thermal state data; the nuclear thermal coupling numerical analysis unit calculates the core state at the next moment through numerical simulation based on the sensor monitoring data and the high-resolution nuclear thermal distribution parameters from the data fusion unit; and the data fusion unit fuses the core nuclear thermal state data from the sensors and the high-resolution nuclear thermal coupling numerical simulation results to obtain the key parameters of the optimized nuclear thermal coupling numerical simulation model.

Images