Method, device and medium for estimating SOC of aging emergency power supply of nuclear power plant

Abstract

The present application discloses a method, device and medium for estimating the SOC of an aging emergency power supply of a nuclear power plant. According to the historical operation data of the aging emergency power supply, an open circuit voltage machine learning model is obtained; according to the open circuit voltage machine learning model and the open circuit voltage factory model of the aging emergency power supply, a mathematical model of a second-order RC drift model including control parameters, discrete state equations and observation equations is constructed; the observation equation is used to observe and calculate the drift error of the system state quantity in the discrete state equation under the input of the control parameter; a volumetric Kalman filter is constructed according to the mathematical model of the second-order RC drift model to obtain the filter gain and the drift observation error; according to the discrete state equation, the filter gain, the drift observation error, the preset first estimation algorithm or the preset dynamic a posteriori correction algorithm and the preset second estimation algorithm, a more robust and accurate SOC estimation result can be obtained without calibration through any offline test.

Description

Technical Field

[0001] The present application relates to the field of nuclear power technology, and in particular to a method, device and medium for estimating the SOC of an aging emergency power supply of a nuclear power plant. Background Art

[0002] In nuclear power plants, emergency diesel generators and emergency batteries are usually installed. Among them, batteries are also called emergency power supplies. When a nuclear power plant fails and needs to be shut down, they can quickly provide short-term power supply to the main coolant pump, main feed water pump and other instrumentation and control systems to prevent the failure from turning into an accident. Therefore, emergency power supplies have a great impact on ensuring nuclear safety. With the rapid development of my country's nuclear power installed capacity, more and more emergency batteries are used in nuclear power plants.

[0003] At present, nuclear power plants generally use the method of regularly charging emergency power supplies to ensure that their power is always available. However, as the service life increases, the performance of the battery will change, and the extensive management mode of regular and scheduled charging is difficult to adapt. From the perspective of safety application, the emergency power supply balances supply and demand, stores excess energy, and releases energy when necessary. The performance is important. It is crucial to estimate the state of charge (SOC) of the aging emergency power supply. The SOC of the emergency power supply in a nuclear power plant directly determines the length of time the system can provide backup power, which is crucial to ensure the balance of supply and demand of the system, improve energy utilization efficiency, and stable operation.

[0004] At present, model-based SOC estimation methods have been widely used in power management. Such methods mainly rely on accurate and reliable power models. However, the accuracy of power models will gradually decrease during the aging process and cannot be corrected, which greatly restricts the accuracy of SOC estimation.

[0005] Therefore, a method to more accurately estimate the SOC of the aging emergency power supply of nuclear power plants is urgently needed. Summary of the invention

[0006] The present application aims to solve one of the technical problems in the related art to a certain extent. To this end, the present application provides a method, device and medium for estimating the SOC of an aging emergency power supply of a nuclear power plant.

[0007] As a first aspect of the present application, a method for estimating the SOC of an aging emergency power supply of a nuclear power plant is provided, wherein the method comprises:

[0008] According to the historical operation data of the aging emergency power supply, an open circuit voltage machine learning model is obtained; wherein the historical operation data includes historical open circuit voltage data, historical state of charge SOC data and historical state of health SOH data, the input of the open circuit voltage machine learning model is SOC data and SOH data, and the output of the open circuit voltage machine learning model is open circuit voltage data;

[0009] A mathematical model of a second-order RC drift model is constructed according to the open-circuit voltage machine learning model and the factory model of the open-circuit voltage of the aging emergency power supply; wherein the second-order RC drift model includes a loop consisting of the open-circuit voltage of the aging emergency power supply, the factory ohmic internal resistance of the aging emergency power supply, a first RC loop, a second RC loop, and the terminal voltage of the aging emergency power supply, the RC loop includes a factory resistor and a factory capacitor connected in parallel, and the mathematical model of the second-order RC drift model includes control parameters, discrete state equations, and observation equations, and the observation equations are used to observe and calculate the drift error of the system state quantity in the discrete state equation under the input of the control parameters;

[0010] Constructing a volumetric Kalman filter according to the mathematical model of the second-order RC drift model, obtaining a filter gain, a drift observation error of the open circuit voltage machine learning model, and a drift observation error of the open circuit voltage factory model;

[0011] When the drift observation error of the open circuit voltage machine learning model is not greater than the drift observation error of the open circuit voltage factory model, the SOC estimation result of the aging emergency power supply is determined according to the discrete state equation, the filter gain, the drift observation error of the open circuit voltage machine learning model and the preset first estimation algorithm;

[0012] When the drift observation error of the open circuit voltage machine learning model is greater than the drift observation error of the open circuit voltage factory model, the drift observation error of the open circuit voltage factory model is corrected according to a preset dynamic a posteriori correction algorithm, and the SOC estimation result of the aging emergency power supply is determined according to the discrete state equation, the filter gain, the corrected drift observation error of the open circuit voltage factory model and the preset second estimation algorithm.

[0013] Optionally, constructing a mathematical model of a second-order RC drift model according to the open circuit voltage machine learning model and the open circuit voltage factory model of the aged emergency power supply includes:

[0014] A mapping relationship between the terminal voltage and the excitation current of the aged emergency power supply is established according to the open circuit voltage machine learning model, the open circuit voltage factory model, the load voltage of the first RC loop, the load voltage of the second RC loop, and the factory ohmic internal resistance of the aged emergency power supply;

[0015] According to the power capacity and coulomb efficiency of the aged emergency power supply, a dynamic equation of SOC data with respect to the excitation current is established;

[0016] According to the factory resistance and factory capacitance of the RC loop, a Kirchhoff current relationship equation of the load voltage of the RC loop with respect to the excitation current is established;

[0017] According to the mapping relationship between the terminal voltage and the excitation current of the aging emergency power supply, the dynamic equation of the SOC data with respect to the excitation current, and the Kirchhoff current relationship equation of the load voltage of the RC loop with respect to the excitation current, a state space of SOC estimation is defined; wherein the state space of SOC estimation includes the control parameter, the state quantity of SOC estimation, and the observation quantity;

[0018] The discrete state equation and observation equation are constructed according to the state quantity and observation quantity of the SOC estimation.

[0019] Optionally, in the step of constructing a mathematical model of a second-order RC drift model according to the open circuit voltage machine learning model and the open circuit voltage factory model of the aged emergency power supply, the step includes:

[0020] The mapping relationship between the terminal voltage and the excitation current of the aging emergency power supply is established by the following formulas (1) and (2):

[0021] (1);

[0022] (2);

[0023] The dynamic equation of SOC data with respect to excitation current is established by the following formula (3):

[0024] (3);

[0025] The Kirchhoff current relationship equation of the load voltage of the RC loop with respect to the excitation current is established by the following formulas (4) and (5):

[0026] (4);

[0027] (5);

[0028] The state space of SOC estimation is defined by the following formula (6):

[0029] (6);

[0030] The discrete state equation and observation equation are constructed by the following formulas (7), (8), and (9):

[0031] (7);

[0032] (8);

[0033] (9);

[0034] In the above formulas (1) to (9), Indicates the terminal voltage of the aging emergency power supply, represents the excitation current, including the open circuit voltage machine learning model and the open circuit voltage factory model, Indicates the open circuit voltage of the factory model, and represents a constant, Indicates the aging of the emergency power supply SOC Second power, represents the load voltage of the first RC loop, represents the load voltage of the second RC loop, Indicates the factory ohmic internal resistance of the aged emergency power supply. , Respectively indicate that the aging emergency power supply , SOC data at the moment, Indicates the power capacity of the aged emergency power supply, represents the coulombic efficiency of the aged emergency power supply, Indicates the factory resistance of the first RC circuit, Indicates the factory resistance of the second RC circuit, Indicates the factory capacitance of the first RC circuit, Indicates the factory capacitance of the second RC circuit, represents the state quantity of SOC estimation, represents the observed value of SOC estimation, represents the control parameter, Indicates discrete moments, Indicates The system state quantity in the discrete state equation at discrete moments is: Indicates The system observable quantity in the observation equation at discrete moments is Indicates The control parameters at discrete moments are Indicates The state white noise with zero mean and independent at discrete moments is Indicates The observation white noise with zero mean and independent at discrete moments is represents the state transfer matrix, represents the control matrix.

[0035] Optionally, constructing a volumetric Kalman filter according to the mathematical model of the second-order RC drift model to obtain a filter gain, a drift observation error of the open circuit voltage machine learning model, and a drift observation error of the open circuit voltage factory model includes:

[0036] Constructing a volumetric Kalman filter to update the system state quantity in the discrete state equation according to the control parameter;

[0037] According to the observation equation and the updated discrete state equation, determining the estimated volume point of the system state quantity in the updated discrete state equation, the propagation volume point of the open circuit voltage machine learning model, and the propagation volume point of the open circuit voltage factory model;

[0038] Determining a measurement estimate of the open circuit voltage machine learning model and a measurement estimate of the open circuit voltage factory model according to the propagation volume points of the open circuit voltage machine learning model and the propagation volume points of the open circuit voltage factory model;

[0039] Determining the filter gain according to the estimated volume point of the system state quantity in the updated discrete state equation and the measured estimated value of the open circuit voltage machine learning model;

[0040] Based on the measured estimated value of the open circuit voltage machine learning model and the measured estimated value of the open circuit voltage factory model, the drift observation error of the open circuit voltage machine learning model and the drift observation error of the open circuit voltage factory model are determined.

[0041] Optionally, in the step of constructing a volumetric Kalman filter according to the mathematical model of the second-order RC drift model, and obtaining the drift observation error of the open circuit voltage machine learning model and the drift observation error of the open circuit voltage factory model, the following steps are included:

[0042] The system state quantity in the discrete state equation is updated by the following formulas (10) and (11):

[0043] (10);

[0044] (11);

[0045] The estimated volume point and propagation volume point are determined by the following formulas (12) and (13):

[0046] (12);

[0047] (13);

[0048] The measurement estimate is determined by the following formulas (14) and (15):

[0049] (14);

[0050] (15);

[0051] The filter gain is determined by the following formulas (16), (17), and (18):

[0052] (16);

[0053] ,in, (17);

[0054] (18);

[0055] The drift observation error is determined by the following formula (19):

[0056] (19);

[0057] In the above formulas (10) to (19), Indicates discrete moments, Indicates The system state quantity in the discrete state equation at discrete moments is: Indicates The system observable quantity in the observation equation at discrete moments is Indicates The control parameters at discrete moments are Indicates The state quantity of the system at a discrete time The prior estimate of Indicates the first The state quantity of the system at a discrete time The posterior estimate of and represents the control parameter, [] indicates QR decomposition method, Represents the updated system state quantity in the discrete state equation The variance matrix of the corresponding state white noise, the state white noise has zero mean and is independent, Represents the updated system state quantity in the discrete state equation The corresponding Volume points, Represents the updated system state quantity in the discrete state equation No. Estimated volume points, represents the first propagation volume points, Indicates the open circuit voltage factory model propagation volume points, The value is 1, 2, ... , for[ ] List, for The unit array, is the state quantity The dimension of Indicates that the open circuit voltage machine learning model is The open circuit voltage output value at a discrete moment, Indicates the open circuit voltage factory model for the first The open circuit voltage output value at a discrete moment, represents the measured estimate of the open circuit voltage machine learning model, represents the measured estimated value of the open circuit voltage factory model, Represents the observed quantity The weighted center matrix of Indicates the state The weighted center matrix of represents the filter gain, express The corresponding variance matrix is represents independent observation white noise with zero mean, express The square root matrix of Represents the square root matrix of the observation covariance matrix, which is related to the system observations The corresponding process calculation matrix, Represents the square root matrix of the state observation covariance matrix, which is related to the state quantity And the observed quantity The corresponding process calculation matrix, represents the drift observation error of the open circuit voltage factory model, represents the drift observation error of the open circuit voltage machine learning model, Indicates the terminal voltage of the aging emergency power supply.

[0058] Optionally, the drift observation error of the open circuit voltage factory model is corrected by the following formula:

[0059] (20);

[0060] (twenty one);

[0061] In the above formulas (20) and (21), represents the drift observation error of the open circuit voltage machine learning model, represents the drift observation error of the open circuit voltage factory model, represents the drift observation error of the open circuit voltage factory model after correction, represents the modified expected gain.

[0062] Optionally, the preset first estimation algorithm includes the following formula (22):

[0063] = + (twenty two);

[0064] The preset first estimation algorithm includes the following formula (23):

[0065] = + (twenty three);

[0066] In the above formulas (22) and (23), Indicates The state quantity of the system at a discrete time The a posteriori estimate of is the SOC estimation result of the aging emergency power supply, Indicates The state quantity of the system at a discrete time The prior estimate of represents the filter gain, represents the drift observation error of the open circuit voltage machine learning model, It represents the drift observation error of the open circuit voltage factory model after correction.

[0067] Optionally, the type of the open circuit voltage machine learning model includes any one of vector regression, artificial neural network, and Gaussian process regression.

[0068] As a second aspect of the present application, an electronic device is provided, wherein the electronic device includes:

[0069] one or more processors;

[0070] A memory having one or more computer programs stored thereon, wherein when the one or more computer programs are executed by the one or more processors, the one or more processors implement the method for estimating the SOC of the aging emergency power supply of a nuclear power plant as described in the first aspect of the present application.

[0071] As a third aspect of the present application, a computer-readable medium is provided, on which a computer program is stored, wherein when the computer program is executed by a processor, the method for estimating the SOC of an aging emergency power supply of a nuclear power plant described in the first aspect of the present application is implemented.

[0072] In the method for estimating the SOC of the aging emergency power supply of a nuclear power plant provided in the embodiment of the present application, by using the historical operation data of the aging emergency power supply to train a low-cost open-circuit voltage machine learning model, the characteristics of the black box model can be used to establish a direct causal relationship reflecting the relevant factors based on the input-output relationship, so as to achieve more flexible and efficient capture of the aging characteristics of the open-circuit voltage mechanism deviation, and to achieve a rough adjustment of the second-order RC drift model of the aging emergency power supply in advance; the trained open-circuit voltage machine learning model is embedded in the open-circuit voltage factory model when the power supply leaves the factory, and a mathematical model of the second-order RC drift model is constructed, and the mathematical model is used to characterize the aging characteristics of the open-circuit voltage mechanism deviation. Then, by constructing a volumetric Kalman filter to respectively construct the drift observation errors of the two types of models, the aging characteristics of the initial value deviation of the model parameters are characterized, and the aging difference between the current actual state and the factory state of the aging emergency power supply is highlighted. Finally, by comparing the drift observation errors of the two types of models and using the preset dynamic a posteriori correction algorithm to correct the drift observation error of the open circuit voltage factory model with smaller drift observation error, the influence of the open circuit voltage mechanism deviation and the influence of the initial value deviation of the model parameters are highlighted, and the second-order RC drift model of the aging emergency power supply is fine-tuned to ensure the convergence and stability of the filter. It is possible to obtain more robust and accurate SOC estimation results without any offline test calibration when only the open circuit voltage factory model of the aging emergency power supply is used, and even if the prior knowledge of SOC and SOH deviates at the same time, it is possible to obtain more robust and accurate SOC estimation results without any offline test calibration, providing accurate and effective data support for the working scenarios where the aging emergency power supply of nuclear power plants needs uninterrupted preparation for a long period of time, and providing strong guidance for the management and safety maintenance of aging emergency power supplies. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The present application will be further described below in conjunction with the accompanying drawings:

[0074] Figure 1 It is a flow chart of an implementation method of the method for estimating the SOC of an aging emergency power supply of a nuclear power plant provided in an embodiment of the present application;

[0075] Figure 2 is a structural schematic diagram of a second-order RC drift model provided in an embodiment of the present application;

[0076] Figure 3 It is a flow chart of an implementation method of constructing a mathematical model of a second-order RC drift model provided in an embodiment of the present application;

[0077] Figure 4 This is a flow chart of an implementation method of obtaining filter gain and drift observation error provided in an embodiment of the present application;

[0078] Figure 5 It is a schematic diagram of the SOC estimation result and time of an aged emergency power supply with a SOH of 44.59% provided in an embodiment of the present application;

[0079] Figure 6 It is a schematic diagram showing the correlation between the mean absolute error of the SOC estimation result of the aged emergency power supply with a SOH of 44.59% and time provided in an embodiment of the present application;

[0080] Figure 7 It is a module diagram of an implementation of an electronic device provided in an embodiment of the present application;

[0081] Figure 8 It is a schematic diagram of a computer-readable medium provided in an embodiment of the present application.

[0082] Description of Reference Numerals

[0083] 101: Processor 102: Memory

[0084] 103: I / O interface 104: bus DETAILED DESCRIPTION

[0085] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments based on the embodiments are intended to be used to explain the present application and cannot be understood as limiting the present application.

[0086] References to "one embodiment" or "an example" or "an example" in this specification mean that a particular feature, structure, or characteristic described in conjunction with the embodiment itself may be included in at least one embodiment disclosed in the present application. The appearance of the phrase "in one embodiment" in various places in the specification does not necessarily refer to the same embodiment.

[0087] At present, model-based SOC estimation methods have been widely used in power management. Such methods usually require the establishment of an accurate and reliable power model, and then use observers such as Kalman filters to estimate SOC. As a power model, the equivalent circuit model has a relatively simple structure and strong mechanism, which can meet the online requirements of emergency power supply under different loads. However, the accuracy of the equivalent circuit model is also the biggest constraint on the accuracy of SOC estimation, especially for the aging emergency power supply of nuclear power plants. This is not conducive to the SOC estimation of the power supply and the corresponding emergency capability and reliability.

[0088] After analyzing the equivalent circuit model, the inventor found that the equivalent circuit model of the emergency power supply is mainly composed of model parameters and open-circuit voltage mapping curves, wherein the model parameters refer to the parameters of the RC equivalent devices, and the open-circuit voltage mapping curve refers to the mapping relationship between the open-circuit voltage of the power supply and the SOC. Correspondingly, during the aging process of the power supply, the model drift error is mainly composed of model parameter deviation and open-circuit voltage deviation. For the model parameter deviation, it can be divided into parameter initial value deviation and parameter update deviation. The parameter update deviation is determined by the performance of online parameter estimation methods such as least squares during the use of the power supply. The parameter initial value deviation refers to the deviation between the initial model parameters obtained in the factory test of the power supply and the actual state of the aging power supply. This deviation continues to increase with the aging of the emergency power supply and is one of the main reasons for the serious drift of the aging power supply model. For the open-circuit voltage deviation, it can be divided into regression deviation and mechanistic deviation. Regression deviation refers to the calculation error when fitting the open-circuit voltage mapping curve in the factory test of the power supply. The mechanistic deviation refers to the deviation between the open-circuit voltage mapping curve and the actual state of the power supply under the influence of factors such as non-consistency, temperature, and aging. This is another leading cause of serious drift in the aging power supply model.

[0089] The inventor further proposed that when the emergency power supply of a nuclear power plant is obviously aged during long-term use, due to its long service life and inconvenient online characteristics of disassembly and debugging, the equivalent circuit model calculated when the emergency power supply leaves the factory will drift more and more during the aging process of the power supply, and it is impossible to timely correct the drift caused by the initial value deviation of the model parameters and the mechanistic deviation of the open circuit voltage through offline disassembly and debugging of the model. At the same time, the robustness problem caused by the inherent prior knowledge deviation in the SOC estimation problem is more severe in this case. Therefore, the model-based SOC estimation method will inevitably lose a certain estimation accuracy when applied to the aging emergency power supply of a nuclear power plant, and it is more likely to have extreme phenomenon of estimation divergence, which cannot guarantee the reliable management and safe application of the emergency power supply.

[0090] The inventor further proposed that by using the historical operation data of the aging emergency power supply to train a low-cost open-circuit voltage machine learning model, the aging characteristics of the open-circuit voltage mechanistic deviation can be captured; the trained open-circuit voltage machine learning model is embedded in the open-circuit voltage factory model when the power supply leaves the factory, and a mathematical model of the second-order RC drift model is constructed to characterize the aging characteristics of the open-circuit voltage mechanistic deviation. By constructing a volumetric Kalman filter, the drift observation errors of the two types of models are respectively constructed to characterize the aging characteristics of the initial value deviation of the model parameters. Finally, by comparing the drift observation errors of the two types of models, a model with a smaller drift observation error is used to estimate the SOC of the aging emergency power supply. A more robust and accurate SOC estimation result can be obtained, providing accurate and effective data support for the working scenarios where the aging emergency power supply of nuclear power plants requires uninterrupted preparation for a long period of time.

[0091] As a first aspect of an embodiment of the present application, a method for estimating the SOC of an aging emergency power supply of a nuclear power plant is provided, such as Figure 1 As shown, the method may include the following steps:

[0092] Step S110, obtaining an open circuit voltage machine learning model according to the historical operation data of the aging emergency power supply; wherein the historical operation data includes historical open circuit voltage data, historical state of charge SOC data and historical state of health (SOH) data, the input of the open circuit voltage machine learning model is SOC data and SOH data, and the output of the open circuit voltage machine learning model is open circuit voltage data;

[0093] Step S120, constructing a mathematical model of a second-order RC drift model according to the open-circuit voltage machine learning model and the factory model of the open-circuit voltage of the aged emergency power supply; wherein the second-order RC drift model includes a loop consisting of the open-circuit voltage of the aged emergency power supply, the factory ohmic internal resistance of the aged emergency power supply, a first RC loop, a second RC loop, and the terminal voltage of the aged emergency power supply, the RC loop includes a factory resistor and a factory capacitor connected in parallel, and the mathematical model of the second-order RC drift model includes a control parameter, a discrete state equation, and an observation equation, and the observation equation is used to observe and calculate the drift error of the system state quantity in the discrete state equation under the input of the control parameter;

[0094] Step S130, constructing a volumetric Kalman filter according to the mathematical model of the second-order RC drift model, obtaining a filter gain, a drift observation error of the open circuit voltage machine learning model, and a drift observation error of the open circuit voltage factory model;

[0095] Step S140, when the drift observation error of the open circuit voltage machine learning model is not greater than the drift observation error of the open circuit voltage factory model, determining the SOC estimation result of the aging emergency power supply according to the discrete state equation, the filter gain, the drift observation error of the open circuit voltage machine learning model and a preset first estimation algorithm;

[0096] Step S150, when the drift observation error of the open circuit voltage machine learning model is greater than the drift observation error of the open circuit voltage factory model, the drift observation error of the open circuit voltage factory model is corrected according to a preset dynamic a posteriori correction algorithm, and the SOC estimation result of the aging emergency power supply is determined according to the discrete state equation, the filter gain, the corrected drift observation error of the open circuit voltage factory model and the preset second estimation algorithm.

[0097] Among them, historical operation data refers to the operation data of the aging emergency power supply during the aging process. SOC refers to the ratio of the remaining capacity of the battery after it has been used for a period of time or has been shelved for a long time to its capacity in a fully charged state. The SOC of a fully discharged battery is 0, and the SOC of a fully charged battery is 1. SOH is also known as battery capacity, health or performance status, that is, the percentage of the fully charged capacity of the battery relative to the rated capacity. The SOH of a new battery is 100%, and the SOH is 0% when it is completely scrapped.

[0098] Among them, in the step of obtaining the open circuit voltage machine learning model based on the historical operation data of the aging emergency power supply, in order to improve the model training efficiency and accuracy, the three different types of historical operation data, namely the historical open circuit voltage data, the historical SOC data and the historical SOH data, can be normalized and preprocessed first. Then, the processed historical SOC data and the processed historical SOH data are used as model inputs, and the processed historical open circuit voltage data is used as model output to train the initial model and obtain the trained open circuit voltage machine learning model.

[0099] Among them, the embodiment of the present application does not specifically limit the type of open circuit voltage machine learning model. For example, it can include any one of vector regression, artificial neural network, and Gaussian process regression.

[0100] By utilizing the historical operating data of the aging emergency power supply, a low-cost open-circuit voltage machine learning model is obtained. With the help of the characteristics of the black box model, a direct causal relationship between relevant factors can be established based on the input-output relationship, so as to capture the aging characteristics of the open-circuit voltage mechanism deviation more flexibly and efficiently, and pre-adjust the second-order RC drift model of the aging emergency power supply.

[0101] like Figure 2The second-order RC drift model provided in the embodiment of the present application is shown in FIG. The second-order RC drift model includes the open-circuit voltage of the aging emergency power supply , aging emergency power supply factory ohmic internal resistance , the terminal voltage of the first RC circuit, the second RC circuit and the aging emergency power supply The first RC loop consists of a parallel factory resistor And factory capacitor , the second RC circuit includes the factory resistor in parallel And factory capacitor .besides, represents the excitation current in the loop, represents the load voltage of the first RC loop, Represents the load voltage of the second RC circuit.

[0102] The inventors further proposed to embed the trained open-circuit voltage machine learning model into the open-circuit voltage factory model of the power supply when it leaves the factory, and to characterize the aging characteristics of the open-circuit voltage mechanistic deviation by constructing the control parameters, discrete state equations and observation equations of the second-order RC drift model and using the observation equation to observe and calculate the drift error of the system state quantity in the discrete state equation under the input of the control parameters.

[0103] Then, the drift observation errors of the two types of models are constructed by constructing a volumetric Kalman filter, the filter gain of the volumetric Kalman filter is calculated, and the aging characteristics of the initial value deviation of the model parameters are characterized, highlighting the aging difference between the current actual state and the factory state of the aging emergency power supply.

[0104] It can be understood that in the embodiment of the present application, by comparing the drift observation error of the open circuit voltage machine learning model with the drift observation error of the open circuit voltage factory model, it is determined whether to execute step S140 or step S150 based on the size relationship between the drift observation error of the open circuit voltage machine learning model and the drift observation error of the open circuit voltage factory model.

[0105] Since the drift observation error of the open circuit voltage machine learning model is the result of characterizing the aging characteristics of the open circuit voltage mechanism deviation, in theory it should not be greater than the drift observation error of the open circuit voltage factory model, that is, the open circuit voltage output by the open circuit voltage machine learning model should be closer to the actual open circuit voltage of the aged emergency power supply. Therefore, when the drift observation error of the open circuit voltage machine learning model is not greater than the drift observation error of the open circuit voltage factory model, the SOC of the aged emergency power supply can be estimated directly using the discrete state equation, filter gain, drift observation error of the open circuit voltage machine learning model, and the preset first estimation algorithm, while highlighting the impact of the open circuit voltage mechanism deviation and the impact of the initial value deviation of the model parameters.

[0106] When the drift observation error of the open circuit voltage machine learning model is greater than the drift observation error of the open circuit voltage factory model, it means that compared with the open circuit voltage machine learning model, the open circuit voltage output by the open circuit voltage factory model may be closer to the actual open circuit voltage of the aging emergency power supply. However, the influence of the open circuit voltage mechanism deviation needs to be considered at this time. The drift observation error of the open circuit voltage factory model is corrected by using a preset dynamic a posteriori correction algorithm to highlight the influence of the open circuit voltage mechanism deviation. Finally, the discrete state equation, filter gain, the drift observation error of the corrected open circuit voltage factory model and the preset second estimation algorithm are used to estimate the SOC of the aging emergency power supply, highlighting the influence of the open circuit voltage mechanism deviation and the influence of the initial value deviation of the model parameters.

[0107] In the method for estimating the SOC of the aging emergency power supply of a nuclear power plant provided in the embodiment of the present application, by using the historical operation data of the aging emergency power supply to train a low-cost open-circuit voltage machine learning model, the characteristics of the black box model can be used to establish a direct causal relationship reflecting the relevant factors based on the input-output relationship, so as to achieve more flexible and efficient capture of the aging characteristics of the open-circuit voltage mechanism deviation, and to achieve a rough adjustment of the second-order RC drift model of the aging emergency power supply in advance; the trained open-circuit voltage machine learning model is embedded in the open-circuit voltage factory model when the power supply leaves the factory, and a mathematical model of the second-order RC drift model is constructed, and the mathematical model is used to characterize the aging characteristics of the open-circuit voltage mechanism deviation. Then, by constructing a volumetric Kalman filter to respectively construct the drift observation errors of the two types of models, the aging characteristics of the initial value deviation of the model parameters are characterized, and the aging difference between the current actual state and the factory state of the aging emergency power supply is highlighted. Finally, by comparing the drift observation errors of the two types of models and using the preset dynamic a posteriori correction algorithm to correct the drift observation error of the open circuit voltage factory model with smaller drift observation error, the influence of the open circuit voltage mechanism deviation and the influence of the initial value deviation of the model parameters are highlighted, and the second-order RC drift model of the aging emergency power supply is fine-tuned to ensure the convergence and stability of the filter. It is possible to obtain more robust and accurate SOC estimation results without any offline test calibration when only the open circuit voltage factory model of the aging emergency power supply is used, and even if the prior knowledge of SOC and SOH deviates at the same time, it is possible to obtain more robust and accurate SOC estimation results without any offline test calibration, providing accurate and effective data support for the working scenarios where the aging emergency power supply of nuclear power plants needs uninterrupted preparation for a long period of time, and providing strong guidance for the management and safety maintenance of aging emergency power supplies.

[0108] Accordingly, in some embodiments, the mathematical model of the second-order RC drift model (i.e., involved in step S120) is constructed based on the open circuit voltage machine learning model and the open circuit voltage factory model of the aging emergency power supply, such as Figure 3 As shown, the following steps S210 to S250 may be included:

[0109] Step S210, establishing a mapping relationship between the terminal voltage and the excitation current of the aged emergency power supply according to the open circuit voltage machine learning model, the open circuit voltage factory model, the load voltage of the first RC loop, the load voltage of the second RC loop, and the factory ohmic internal resistance of the aged emergency power supply;

[0110] Step S220, establishing a dynamic equation of SOC data with respect to excitation current according to the power capacity and coulomb efficiency of the aged emergency power supply;

[0111] Step S230, establishing a Kirchhoff current relationship equation of the load voltage of the RC loop with respect to the excitation current according to the factory resistance and factory capacitance of the RC loop;

[0112] Step S240, defining a state space of SOC estimation according to a mapping relationship between the terminal voltage and the excitation current of the aging emergency power supply, a dynamic equation of the SOC data with respect to the excitation current, and a Kirchhoff current relationship equation of the load voltage of the RC loop with respect to the excitation current; wherein the state space of SOC estimation includes the control parameters, the state quantity of SOC estimation, and the observation quantity;

[0113] Step S250, constructing the discrete state equation and observation equation according to the state quantity and observation quantity of the SOC estimation.

[0114] Among them, the control parameters, the state variables of SOC estimation and the observed variables can be expressed as , .

[0115] However, it should be noted that the discrete state equations mentioned above are all used to estimate the state quantity when performing SOC estimation. Representation and state quantity Correspondingly, in fact, in addition to SOC estimation, it is also possible to consider updating the parameters of the open circuit voltage factory model. In step S240, a state quantity related to updating the parameters of the open circuit voltage factory model can also be constructed. In step S250, the state quantity can also be constructed. Representation and state quantity The related discrete state equations, that is, the mathematical model of the second-order RC drift model can also include control parameters, state quantities Corresponding discrete state equations and state quantities The corresponding discrete state equation and observation equation. And similarly, the control parameters and state quantities can be used to calculate the The corresponding discrete state equation and observation equation are used to construct a volumetric Kalman filter, obtain the filter gain and drift observation error, and update the parameters of the open circuit voltage factory model.

[0116] Accordingly, in some embodiments, the step of constructing a mathematical model of a second-order RC drift model according to the open circuit voltage machine learning model and the open circuit voltage factory model of the aged emergency power supply (i.e., involved in step S120) may include:

[0117] The mapping relationship between the terminal voltage and the excitation current of the aging emergency power supply is established by the following formulas (1) and (2):

[0118] (1);

[0119] (2);

[0120] The dynamic equation of SOC data with respect to excitation current is established by the following formula (3):

[0121] (3);

[0122] The Kirchhoff current relationship equation of the load voltage of the RC loop with respect to the excitation current is established by the following formulas (4) and (5):

[0123] (4);

[0124] (5);

[0125] The state space of SOC estimation is defined by the following formula (6):

[0126] (6);

[0127] The discrete state equation and observation equation are constructed by the following formulas (7), (8), and (9):

[0128] (7);

[0129] (8);

[0130] (9);

[0131] In the above formulas (1) to (9), Indicates the terminal voltage of the aging emergency power supply, represents the excitation current, including the open circuit voltage machine learning model and the open circuit voltage factory model, Indicates the open circuit voltage of the factory model, and represents a constant, Indicates the aging of the emergency power supply SOC Second power, represents the load voltage of the first RC loop, represents the load voltage of the second RC loop, Indicates the factory ohmic internal resistance of the aged emergency power supply. , Respectively indicate that the aging emergency power supply , SOC data at the moment, Indicates the power capacity of the aged emergency power supply, represents the coulombic efficiency of the aged emergency power supply, Indicates the factory resistance of the first RC circuit, Indicates the factory resistance of the second RC circuit, Indicates the factory capacitance of the first RC circuit, Indicates the factory capacitance of the second RC circuit, represents the state quantity of SOC estimation, represents the observed value of SOC estimation, represents the control parameter, Indicates discrete moments, Indicates The system state quantity in the discrete state equation at discrete moments is: Indicates The system observable quantity in the observation equation at discrete moments is Indicates The control parameters at discrete moments are Indicates The state white noise with zero mean and independent at discrete moments is Indicates The observation white noise with zero mean and independent at discrete moments is represents the state transfer matrix, represents the control matrix.

[0132] As described above, it is also possible to construct state quantities related to updating the parameters of the open circuit voltage factory model. And the state quantity Corresponding discrete state equation. Accordingly, in the step of defining the state space of SOC estimation by the above formula (6), the state quantity of model parameter update can also be defined : In the steps of constructing the discrete state equation and observation equation through the above formulas (7), (8), and (9), the state quantity The corresponding discrete state equation is: ,in, Indicates The system state quantity in the discrete state equation at discrete moments is: Indicates The state white noise is zero mean and independent at discrete moments.

[0133] In some embodiments, the volumetric Kalman filter is constructed according to the mathematical model of the second-order RC drift model to obtain the filter gain, the drift observation error of the open circuit voltage machine learning model, and the drift observation error of the open circuit voltage factory model (i.e., involved in step S130), such as Figure 4 As shown, the following steps S310 to S350 may be included:

[0134] Step S310, constructing a cubature Kalman filter, and updating the system state quantity in the discrete state equation according to the control parameter;

[0135] Step S320, determining the estimated volume points of the system state quantity in the updated discrete state equation, the propagation volume points of the open circuit voltage machine learning model, and the propagation volume points of the open circuit voltage factory model according to the observation equation and the updated discrete state equation;

[0136] Step S330, determining a measurement estimate of the open circuit voltage machine learning model and a measurement estimate of the open circuit voltage factory model according to the propagation volume point of the open circuit voltage machine learning model and the propagation volume point of the open circuit voltage factory model;

[0137] Step S340, determining the filter gain according to the estimated volume point of the system state quantity in the updated discrete state equation and the measured estimated value of the open circuit voltage machine learning model;

[0138] Step S350, determining the drift observation error of the open circuit voltage machine learning model and the drift observation error of the open circuit voltage factory model based on the measured estimated value of the open circuit voltage machine learning model and the measured estimated value of the open circuit voltage factory model.

[0139] Accordingly, in some embodiments, the step of constructing a volumetric Kalman filter according to the mathematical model of the second-order RC drift model, and obtaining the drift observation error of the open circuit voltage machine learning model and the drift observation error of the open circuit voltage factory model (i.e., involved in step S130) may include:

[0140] The system state quantity in the discrete state equation is updated by the following formulas (10) and (11):

[0141] (10);

[0142] (11);

[0143] The estimated volume point and propagation volume point are determined by the following formulas (12) and (13):

[0144] (12);

[0145] (13);

[0146] The measurement estimate is determined by the following formulas (14) and (15):

[0147] (14);

[0148] (15);

[0149] The filter gain is determined by the following formulas (16), (17), and (18):

[0150] (16);

[0151] ,in, (17);

[0152] (18);

[0153] The drift observation error is determined by the following formula (19):

[0154] (19);

[0155] In the above formulas (10) to (19), Indicates discrete moments, Indicates The system state quantity in the discrete state equation at discrete moments is: Indicates The system observable quantity in the observation equation at discrete moments is Indicates The control parameters at discrete moments are Indicates The state quantity of the system at a discrete time The prior estimate of Indicates the first The state quantity of the system at a discrete time The posterior estimate of and represents the control parameter, [] indicates QR decomposition method, Represents the updated system state quantity in the discrete state equation The variance matrix of the corresponding state white noise, the state white noise has zero mean and is independent, Represents the updated system state quantity in the discrete state equation The corresponding Volume points, Represents the updated system state quantity in the discrete state equation No. Estimated volume points, represents the first propagation volume points, Indicates the open circuit voltage factory model propagation volume points, The value is 1, 2, ... , for[ ] List, for The unit array, is the state quantity The dimension of Indicates that the open circuit voltage machine learning model is The open circuit voltage output value at a discrete moment, Indicates the open circuit voltage factory model for the first The open circuit voltage output value at a discrete moment, represents the measured estimate of the open circuit voltage machine learning model, represents the measured estimated value of the open circuit voltage factory model, Represents the observed quantity The weighted center matrix of Indicates the state The weighted center matrix of represents the filter gain, express The corresponding variance matrix is represents independent observation white noise with zero mean, express The square root matrix of Represents the square root matrix of the observation covariance matrix, which is related to the system observations The corresponding process calculation matrix, Represents the square root matrix of the state observation covariance matrix, which is related to the state quantity And the observed quantity The corresponding process calculation matrix, represents the drift observation error of the open circuit voltage factory model, represents the drift observation error of the open circuit voltage machine learning model, Indicates the terminal voltage of the aging emergency power supply.

[0156] In some embodiments, the drift observation error of the open circuit voltage factory model is corrected by the following formula:

[0157] (20);

[0158] (twenty one);

[0159] In the above formulas (20) and (21), represents the drift observation error of the open circuit voltage machine learning model, represents the drift observation error of the open circuit voltage factory model, represents the drift observation error of the open circuit voltage factory model after correction, represents the modified expected gain.

[0160] In some embodiments, the preset first estimation algorithm includes the following formula (22):

[0161] = + (twenty two);

[0162] The preset first estimation algorithm includes the following formula (23):

[0163] = + (twenty three);

[0164] In the above formulas (22) and (23), Indicates The state quantity of the system at a discrete time The a posteriori estimate of is the SOC estimation result of the aging emergency power supply, Indicates The state quantity of the system at a discrete time The prior estimate of represents the filter gain, represents the drift observation error of the open circuit voltage machine learning model, It represents the drift observation error of the open circuit voltage factory model after correction.

[0165] In some embodiments, the type of the open circuit voltage machine learning model includes any one of vector regression, artificial neural network, and Gaussian process regression.

[0166] Furthermore, the inventors also selected three groups of R9 aged emergency power supply modules with SOH of 58.29%, 51.22% and 44.59% respectively and their historical operating data during the aging process, and actually implemented the method for estimating the SOC of the aged emergency power supply of a nuclear power plant provided in this application to verify the method.

[0167] The nominal capacity of the R9 aging emergency power module is 2.1Ah and the initial SOC is 100%.

[0168] Firstly, a total of 79 sets of data with the SOH of the power supply ranging from 100% to 68.93% were collected from the historical operation data of three groups of R9 aging emergency power supply modules during the aging process, including historical open circuit voltage data, historical SOC data, and historical SOH data, and three types of data sets were established.

[0169] Furthermore, the three types of data sets are normalized and preprocessed. The processed historical SOC data and the processed historical SOH data are used as model inputs, and the processed historical open circuit voltage data are used as model outputs. The initial model is trained to obtain three types of open circuit voltage machine learning models, including vector regression, artificial neural network, and Gaussian process regression. The output values ​​of the three types of open circuit voltage machine learning models are recorded as , , .

[0170] Furthermore, the open circuit voltage machine learning model is embedded into the open circuit voltage factory model of the power supply when it leaves the factory, and the mathematical model of the second-order RC drift model is constructed. As shown in Table 1 below, the parameters of the second-order RC drift model and the open circuit voltage factory model are The parameters of the second-order RC drift model include the factory ohmic internal resistance of the aging emergency power supply , Factory resistance of the first RC circuit And factory capacitor , Factory resistance of the second RC circuit And factory capacitor . Open circuit voltage factory model The parameters include 10 constants to .

[0171] Table 1

[0172]

[0173] Furthermore, a volumetric Kalman filter is constructed based on the mathematical model of the second-order RC drift model to obtain the filter gain, the drift observation error of the open-circuit voltage machine learning model, and the drift observation error of the open-circuit voltage factory model. The SOC estimation result is determined based on the discrete state equation, the filter gain, the drift observation error of the open-circuit voltage machine learning model or the drift observation error of the open-circuit voltage factory model, the preset first estimation algorithm or the preset dynamic a posteriori correction algorithm, and the preset second estimation algorithm.

[0174] Furthermore, the determined SOC estimation result is verified. The result calculated by the coulomb counting method is used as the SOC reference value, and in order to verify the robustness of the method provided in this application for estimating the SOC of the aging emergency power supply of a nuclear power plant, the initial value of the SOC is set to 90% of the true value, and the initial value of the actual capacity of the power supply is set to 110% of the true value. The root mean square error (RMSE) and the mean absolute error (MAE) are used as performance indicators of the SOC estimation result, and the performance result corresponding to the open circuit voltage factory model is recorded as The performance results of directly performing SOC estimation after training the three types of open circuit voltage machine learning models are recorded as , , After the three types of open circuit voltage machine learning models are trained and corrected using the preset dynamic a posteriori correction algorithm, the performance results of SOC estimation are recorded as , , Will , , , As a control, , , The statistical results are shown in Table 2 below.

[0175] Table 2

[0176]

[0177] As can be seen from Table 2 above, when estimating the SOC of the aging emergency power supply of a nuclear power plant, when the prior knowledge of SOC and SOH deviates at the same time, the results of the traditional model method are There is a serious divergence, which is completely unable to provide correct data support and guidance for the management and safety maintenance of nuclear power plant emergency power supply. However, the method for estimating the SOC of the aging emergency power supply of nuclear power plants provided in this application has significantly reduced the error of SOC estimation even if the three types of open circuit voltage machine learning models are trained and then directly used for SOC estimation, especially the two types of open circuit voltage machine learning models (artificial neural network and Gaussian process regression) , ), and the RMSE of its SOC estimation is about 5%, which confirms that machine learning can effectively capture the aging characteristics of the mechanistic deviation of the open-circuit voltage and effectively coarse-tune the second-order RC drift model of the aging emergency power supply.

[0178] Furthermore, after training the three types of open-circuit voltage machine learning models, there is still a certain residual error when SOC estimation is performed directly. This is because the aging characteristics caused by the initial value deviation of the model parameters are contained. For this part of the residual error, it is possible to consider fine-tuning the second-order RC drift model of the aging emergency power supply. The method for estimating the SOC of the aging emergency power supply of a nuclear power plant provided in the present application, after training the three types of open-circuit voltage machine learning models, uses a preset dynamic a posteriori correction algorithm to correct it before performing SOC estimation. For the three R9 aging emergency power supply modules with different aging degrees, the error of the SOC estimation results is reduced to about 2%, which confirms that the method for estimating the SOC of the aging emergency power supply of a nuclear power plant provided in the present application has significantly improved the convergence and stability of SOC estimation. In particular, for the three groups of experiments corresponding to vector regression, the errors of SOC estimation were all greater than 10% when dynamic a posteriori correction was not performed, which is still not applicable for the management and safety maintenance of emergency power supplies in nuclear power plants. After dynamic a posteriori correction, the RMSE of SOC estimation were 2.23%, 1.88%, and 1.97%, respectively. The estimation accuracy was significantly improved by an order of magnitude, which confirmed the effectiveness of the dynamic a posteriori correction method.

[0179] Specifically, take the test results of the R9 aging emergency power supply module with a SOH of 44.59% in Table 2 as an example. Figure 5 As shown, it is a schematic diagram of the SOC estimation result and time, the horizontal axis is time, the unit is seconds (s), and the vertical axis is SOC, the unit is a constant between 0 and 1. It can be seen that Compared to For example, the drift of the SOC estimation result is significantly reduced and is obviously close to the reference value. Figure 6 As shown, it is a schematic diagram of the relationship between the mean absolute error of SOC estimation and time. The horizontal axis is time, the unit is seconds (s), and the vertical axis is the mean absolute error of SOC estimation, the unit is a constant with a value between 0 and 0.5. It can be seen that Compared to The mean absolute error of SOC estimation has been significantly reduced, and even the error is frequently zero.

[0180] Therefore, the method for estimating the SOC of an aging emergency power supply of a nuclear power plant provided in the present application can obtain a more robust and accurate SOC estimation result by coarse adjustment and fine adjustment of the second-order RC drift model without any offline test calibration, using only the open-circuit voltage factory model of the aging emergency power supply, and even if deviations occur in the prior knowledge of SOC and SOH at the same time. It provides accurate and effective data support for the working scenarios in which the aging emergency power supply of a nuclear power plant needs uninterrupted preparation for a long period of time, and provides strong guidance for the management and safety maintenance of the aging emergency power supply.

[0181] As a second aspect of the embodiments of the present application, an electronic device is provided, wherein: Figure 7 As shown, the electronic device includes:

[0182] One or more processors 101;

[0183] The memory 102 stores one or more computer programs. When the one or more computer programs are executed by the one or more processors 101, the one or more processors 101 implement the method for estimating the SOC of the aging emergency power supply of a nuclear power plant provided in the first aspect of the embodiment of the present application.

[0184] The electronic device may further include one or more I / O interfaces 103 connected between the processor 101 and the memory 102 and configured to implement information interaction between the processor 101 and the memory 102 .

[0185] Among them, the processor 101 is a device with data processing capabilities, including but not limited to a central processing unit (CPU), etc.; the memory 102 is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory (FLASH); the I / O interface (read-write interface) is connected between the processor and the memory, and can realize information exchange between the processor and the memory, including but not limited to a data bus (Bus), etc.

[0186] In some embodiments, the processor 101 , the memory 102 , and the I / O interface 103 are connected to each other via a bus 104 , and further connected to other components of the computing device.

[0187] As a third aspect of the embodiment of the present application, Figure 8As shown, a computer-readable medium is provided, on which a computer program is stored, wherein when the computer program is executed by a processor, the method for estimating the SOC of the aging emergency power supply of a nuclear power plant provided in the first aspect of the embodiment of the present application is implemented.

[0188] It will be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. Accordingly, the computer program can be stored in a non-volatile computer-readable storage medium, and the computer program can implement the method of any of the above-mentioned embodiments when executed. Among them, any reference to memory, storage, database or other media used in the embodiments provided in the embodiments of the present application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0189] The above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Those skilled in the art should understand that the present application includes but is not limited to the contents described in the drawings and the above specific embodiments. Any modification that does not deviate from the functional and structural principles of the present application will be included in the scope of the claims.

Claims

1. A method for estimating the SOC of an aging emergency power supply of a nuclear power plant, characterized in that: The method comprises: According to the historical operation data of the aging emergency power supply, an open circuit voltage machine learning model is obtained; wherein the historical operation data includes historical open circuit voltage data, historical state of charge SOC data and historical state of health SOH data, the input of the open circuit voltage machine learning model is SOC data and SOH data, and the output of the open circuit voltage machine learning model is open circuit voltage data; A mathematical model of a second-order RC drift model is constructed according to the open-circuit voltage machine learning model and the factory model of the open-circuit voltage of the aging emergency power supply; wherein the second-order RC drift model includes a loop consisting of the open-circuit voltage of the aging emergency power supply, the factory ohmic internal resistance of the aging emergency power supply, a first RC loop, a second RC loop, and the terminal voltage of the aging emergency power supply, the RC loop includes a factory resistor and a factory capacitor connected in parallel, and the mathematical model of the second-order RC drift model includes control parameters, discrete state equations, and observation equations, and the observation equations are used to observe and calculate the drift error of the system state quantity in the discrete state equation under the input of the control parameters; Constructing a volumetric Kalman filter according to the mathematical model of the second-order RC drift model, obtaining a filter gain, a drift observation error of the open circuit voltage machine learning model, and a drift observation error of the open circuit voltage factory model; When the drift observation error of the open circuit voltage machine learning model is not greater than the drift observation error of the open circuit voltage factory model, the SOC estimation result of the aging emergency power supply is determined according to the discrete state equation, the filter gain, the drift observation error of the open circuit voltage machine learning model and the preset first estimation algorithm; When the drift observation error of the open circuit voltage machine learning model is greater than the drift observation error of the open circuit voltage factory model, the drift observation error of the open circuit voltage factory model is corrected according to a preset dynamic a posteriori correction algorithm, and the SOC estimation result of the aging emergency power supply is determined according to the discrete state equation, the filter gain, the corrected drift observation error of the open circuit voltage factory model and the preset second estimation algorithm.

2. The method according to claim 1, characterized in that: The mathematical model of the second-order RC drift model is constructed according to the open circuit voltage machine learning model and the open circuit voltage factory model of the aging emergency power supply, including: A mapping relationship between the terminal voltage and the excitation current of the aged emergency power supply is established according to the open circuit voltage machine learning model, the open circuit voltage factory model, the load voltage of the first RC loop, the load voltage of the second RC loop, and the factory ohmic internal resistance of the aged emergency power supply; According to the power capacity and coulomb efficiency of the aged emergency power supply, a dynamic equation of SOC data with respect to the excitation current is established; According to the factory resistance and factory capacitance of the RC loop, a Kirchhoff current relationship equation of the load voltage of the RC loop with respect to the excitation current is established; According to the mapping relationship between the terminal voltage and the excitation current of the aging emergency power supply, the dynamic equation of the SOC data with respect to the excitation current, and the Kirchhoff current relationship equation of the load voltage of the RC loop with respect to the excitation current, a state space of SOC estimation is defined; wherein the state space of SOC estimation includes the control parameter, the state quantity of SOC estimation, and the observation quantity; The discrete state equation and observation equation are constructed according to the state quantity and observation quantity of the SOC estimation.

3. The method according to claim 2, characterized in that In the step of constructing a mathematical model of a second-order RC drift model according to the open circuit voltage machine learning model and the open circuit voltage factory model of the aging emergency power supply, the step includes: The mapping relationship between the terminal voltage and the excitation current of the aging emergency power supply is established by the following formulas (1) and (2): （1）； （2）； The dynamic equation of SOC data with respect to excitation current is established by the following formula (3): （3）； The Kirchhoff current relationship equation of the load voltage of the RC loop with respect to the excitation current is established by the following formulas (4) and (5): （4）； （5）； The state space of SOC estimation is defined by the following formula (6): （6）； The discrete state equation and observation equation are constructed by the following formulas (7), (8), and (9): （7）； （8）； （9）； In the above formulas (1) to (9), Indicates the terminal voltage of the aging emergency power supply, represents the excitation current, including the open circuit voltage machine learning model and the open circuit voltage factory model, Indicates the open circuit voltage of the factory model, and represents a constant, Indicates the aging of the emergency power supply SOC Second power, represents the load voltage of the first RC loop, represents the load voltage of the second RC loop, Indicates the factory ohmic internal resistance of the aged emergency power supply. , Respectively indicate that the aging emergency power supply , SOC data at the moment, Indicates the power capacity of the aged emergency power supply, represents the coulombic efficiency of the aged emergency power supply, Indicates the factory resistance of the first RC circuit, Indicates the factory resistance of the second RC circuit, Indicates the factory capacitance of the first RC circuit, Indicates the factory capacitance of the second RC circuit, represents the state quantity of SOC estimation, represents the observed value of SOC estimation, represents the control parameter, Indicates discrete moments, Indicates The system state quantity in the discrete state equation at discrete moments is: Indicates The system observable quantity in the observation equation at discrete moments is Indicates The control parameters at discrete moments are Indicates The state white noise with zero mean and independent at discrete moments is Indicates The observation white noise with zero mean and independent at discrete moments is represents the state transfer matrix, represents the control matrix.

4. The method according to claim 1, characterized in that The method of constructing a volumetric Kalman filter according to the mathematical model of the second-order RC drift model, obtaining a filter gain, a drift observation error of the open circuit voltage machine learning model, and a drift observation error of the open circuit voltage factory model, includes: Constructing a volumetric Kalman filter to update the system state quantity in the discrete state equation according to the control parameter; According to the observation equation and the updated discrete state equation, determining the estimated volume point of the system state quantity in the updated discrete state equation, the propagation volume point of the open circuit voltage machine learning model, and the propagation volume point of the open circuit voltage factory model; Determining a measurement estimate of the open circuit voltage machine learning model and a measurement estimate of the open circuit voltage factory model according to the propagation volume points of the open circuit voltage machine learning model and the propagation volume points of the open circuit voltage factory model; Determining the filter gain according to the estimated volume point of the system state quantity in the updated discrete state equation and the measured estimated value of the open circuit voltage machine learning model; Based on the measured estimated value of the open circuit voltage machine learning model and the measured estimated value of the open circuit voltage factory model, the drift observation error of the open circuit voltage machine learning model and the drift observation error of the open circuit voltage factory model are determined.

5. The method according to claim 4, characterized in that In the step of constructing a volumetric Kalman filter according to the mathematical model of the second-order RC drift model, and obtaining the drift observation error of the open circuit voltage machine learning model and the drift observation error of the open circuit voltage factory model, it includes: The system state quantity in the discrete state equation is updated by the following formulas (10) and (11): （10）； （11）； The estimated volume point and propagation volume point are determined by the following formulas (12) and (13): （12）； （13）； The measurement estimate is determined by the following formulas (14) and (15): （14）； （15）； The filter gain is determined by the following formulas (16), (17), and (18): （16）； ,in, (17); （18）； The drift observation error is determined by the following formula (19): （19）； In the above formulas (10) to (19), Indicates discrete moments, Indicates The system state quantity in the discrete state equation at discrete moments is: Indicates The system observable quantity in the observation equation at discrete moments is Indicates The control parameters at discrete moments are Indicates The state quantity of the system at a discrete time The prior estimate of Indicates the first The state quantity of the system at a discrete time The posterior estimate of and represents the control parameter, [] indicates QR decomposition method, Represents the updated system state quantity in the discrete state equation The variance matrix of the corresponding state white noise, the state white noise has zero mean and is independent, Represents the updated system state quantity in the discrete state equation The corresponding Volume points, Represents the updated system state quantity in the discrete state equation No. Estimated volume points, represents the first propagation volume points, Indicates the open circuit voltage factory model propagation volume points, The value is 1, 2, ... , for[ ] List, for The unit array, is the state quantity The dimension of Indicates that the open circuit voltage machine learning model is The open circuit voltage output value at a discrete moment, Indicates the open circuit voltage factory model for the first The open circuit voltage output value at a discrete moment, represents the measured estimate of the open circuit voltage machine learning model, represents the measured estimated value of the open circuit voltage factory model, Represents the observed quantity The weighted center matrix of Indicates the state The weighted center matrix of represents the filter gain, express The corresponding variance matrix is represents independent observation white noise with zero mean, express The square root matrix of Represents the square root matrix of the observation covariance matrix, which is related to the system observations The corresponding process calculation matrix, Represents the square root matrix of the state observation covariance matrix, which is related to the state quantity And the observed quantity The corresponding process calculation matrix, represents the drift observation error of the open circuit voltage factory model, represents the drift observation error of the open circuit voltage machine learning model, Indicates the terminal voltage of the aging emergency power supply.

6. The method according to any one of claims 1 to 5, characterized in that The drift observation error of the open circuit voltage factory model is corrected by the following formula: （20）； （21）； In the above formulas (20) and (21), represents the drift observation error of the open circuit voltage machine learning model, represents the drift observation error of the open circuit voltage factory model, represents the drift observation error of the open circuit voltage factory model after correction, represents the modified expected gain.

7. The method according to any one of claims 1 to 5, characterized in that The preset first estimation algorithm includes the following formula (22): = + （22）； The preset first estimation algorithm includes the following formula (23): = + （23）； In the above formulas (22) and (23), Indicates The state quantity of the system at a discrete time The a posteriori estimate of is the SOC estimation result of the aging emergency power supply, Indicates The state quantity of the system at a discrete time The prior estimate of represents the filter gain, represents the drift observation error of the open circuit voltage machine learning model, It represents the drift observation error of the open circuit voltage factory model after correction.

8. The method according to any one of claims 1 to 5, characterized in that The type of the open circuit voltage machine learning model includes any one of vector regression, artificial neural network, and Gaussian process regression.

9. An electronic device, characterized in that: The electronic device comprises: one or more processors; A memory having one or more computer programs stored thereon, wherein when the one or more computer programs are executed by the one or more processors, the one or more processors implement the method for estimating the SOC of the aging emergency power supply of a nuclear power plant according to any one of claims 1-8.

10. A computer readable medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for estimating the SOC of an aged emergency power supply of a nuclear power plant according to any one of claims 1 to 8 is implemented.

Images