Nuclear power plant fault diagnosis method based on interpretable deep transfer learning

Abstract

The invention relates to the technical field of nuclear power plant safety, and discloses a nuclear power plant fault diagnosis method based on interpretable deep transfer learning, which comprises the following steps: acquiring and preprocessing operation data or simulation data under multiple power levels, and dividing a source domain and a target domain; constructing and preliminarily training a multivariable interpretable gating cycle unit model, and screening features; training a source domain fault diagnosis model based on an adaptive focus loss function; training by adopting a self-adaptive focus mixed maximum mean value difference loss function, and migrating the source domain model to a target domain; and finally, integrating the source domain model and the target domain model to form a diagnosis system, and inputting to-be-tested data clusters and allocating the to-be-tested data clusters to the corresponding models to obtain a fault diagnosis result. According to the method, fault diagnosis tasks under different power levels can be effectively processed, operation condition data under the power levels are integrated and transmitted into a single model for training diagnosis, the accuracy and interpretability of the model are improved, and nuclear power plant operators can be further assisted in decision making.

Description

Technical Field

[0001] The present invention relates to the field of nuclear power plant safety technology, and in particular to a nuclear power plant fault diagnosis method based on explainable deep transfer learning. Background Art

[0002] Technological progress has promoted the development of industrial automation, but also made the system more complex, which in turn put forward higher requirements for the safety and reliability of the system. Although the safety and reliability of the system can be enhanced to a certain extent by improving the quality, reliability and robustness of system components, it is still difficult to completely avoid the occurrence of failures. Therefore, fault diagnosis technology has become a key means to ensure the safety and reliability of industrial systems, especially complex nuclear power plant systems. The goal of fault diagnosis technology is to monitor the normal operation of the system and its components, detect the type of fault early, determine the location and severity of the fault, so as to prevent the fault from further damaging the safe operation of the nuclear power plant.

[0003] Fault diagnosis methods can generally be divided into three categories: mathematical model-based methods, expert system-based methods, and data-driven methods. Mathematical model-based methods require the establishment of an accurate mathematical model for the diagnosis object, but this method is usually difficult to achieve for complex systems such as nuclear power plants. Expert system methods are difficult to implement a complete expert rule base due to the difficulty of building a knowledge base and the numerous and difficult to exhaust rules, which greatly limits their practical application. In contrast, data-driven methods train diagnostic models by using historical operating data of the diagnosis object. These data can be obtained through increasingly popular digital instruments and control systems, reducing reliance on artificial assumptions and improving the accuracy and robustness of the model when processing large amounts of diverse data. With these advantages, data-driven methods have high flexibility and wide applicability, and have become the mainstream in the field of fault diagnosis in recent years.

[0004] Although intelligent fault diagnosis of nuclear power plants based on data-driven methods has achieved many successes, there are still two challenges in expanding the application of fault diagnosis. First, due to the influence of many factors in reality (such as changes in power demand, equipment maintenance and overhaul, fuel management, etc.), nuclear power plants usually operate under different power levels. Therefore, for faults generated at different power levels, it is difficult to achieve accurate fault diagnosis by relying solely on a single data-driven model. Secondly, although current data-driven methods, especially neural network methods, have achieved remarkable results in diagnostic performance in multivariate data processing, they are still insufficient in terms of interpretability. Due to the complex and opaque hidden layer structure of the neural network, it is difficult to clearly identify and quantify the specific contribution of each input variable in the model training process.

[0005] Therefore, how to overcome the problem that the probability distribution of sensor monitoring data may be significantly different when nuclear power plants operate at different power levels, resulting in low accuracy of fault diagnosis results when the operating condition data at each power level are integrated and passed into a single model for training and diagnosis, and improve the interpretability of the model, is a technical problem that needs to be solved at present. Summary of the invention

[0006] To this end, the present invention provides a nuclear power plant fault diagnosis method based on interpretable deep transfer learning, which integrates the operating condition data at various power levels and transmits them into a single model for training and diagnosis. The fault diagnosis results have high accuracy and the interpretability of the model is improved as much as possible.

[0007] To achieve this object, the present invention provides a nuclear power plant fault diagnosis method based on interpretable deep transfer learning, and the nuclear power plant fault diagnosis method comprises the following steps:

[0008] Step S1, obtaining real normal operation data, real fault operation data of a nuclear power plant at multiple power levels, and / or simulated normal operation data and simulated fault operation data at multiple power levels generated by a nuclear power plant simulator, dividing and preprocessing the real normal operation data, the real fault operation data and / or the simulated normal operation data, the simulated fault operation data according to their corresponding power levels to generate source domain data and target domain data;

[0009] Step S2: Simplify and construct a model structure of a multivariate encoding gated recurrent unit model using the source domain data and the target domain data;

[0010] Step S3, using the source domain data as input of the simplified coding gated recurrent unit model, training it through an adaptive focal loss function, and generating a source domain nuclear power plant fault diagnosis model of the source domain;

[0011] Step S4, migrating the source domain nuclear power plant fault diagnosis model to the target domain, and using the target domain data as input to train it through an adaptive focal hybrid maximum mean difference loss function to generate a target domain nuclear power plant fault diagnosis model;

[0012] Step S5, combining the source domain nuclear power plant fault diagnosis model and the target domain nuclear power plant fault diagnosis model in parallel to form a nuclear power plant multi-operating condition fault diagnosis system for detecting whether a nuclear power plant fails at multiple power levels;

[0013] The source domain is one or more power levels of a nuclear power plant, and the target domain is one or more power levels of a nuclear power plant excluding the source domain.

[0014] Furthermore, in step S1, the real normal operation data, the real fault operation data and / or the simulated normal operation data, the simulated fault operation data are divided and preprocessed according to their corresponding power levels to generate source domain data and target domain data, including:

[0015] Acquiring the real normal operation data and the real fault operation data and / or the simulated normal operation data and the simulated fault operation data of the nuclear power plant operating at multiple power levels;

[0016] The real normal operation data, the real fault operation data and / or the simulated normal operation data, and the simulated fault operation data under the maximum power level among multiple power levels are divided into the source domain data set, and the real normal operation data, the real fault operation data and / or the simulated normal operation data, and the simulated fault operation data under the remaining power levels are divided into multiple target domain data sets respectively;

[0017] The maximum value and the minimum value in the source domain data set are extracted, and the source domain data set and the target domain data set are normalized based on the maximum value and the minimum value to generate the source domain data and the target domain data.

[0018] Further, in step S2, the multivariable encoding gated recurrent unit model includes an independent variable encoding gated recurrent unit and a hybrid attention mechanism;

[0019] The independent variable encoding gated recurrent unit encodes the hidden states of the source domain data and the target domain data;

[0020] The hybrid attention mechanism is connected to the independent variable encoding gated recurrent unit to integrate the hidden state and output the global variable contribution;

[0021] The independent variable encoding gated recurrent unit obtains an optimal contribution threshold according to the global variable contribution, and simplifies the model structure according to the optimal contribution threshold.

[0022] Furthermore, the step S2 comprises the following steps:

[0023] Step S21, dividing the input source domain data and the target domain data according to the time step and performing matrix dot multiplication with the corresponding weights to obtain a first output matrix;

[0024] Step S22, performing matrix multiplication on the hidden state and the corresponding weight matrix to obtain a second output matrix;

[0025] Step S23, accumulating the hidden state through a cycle of time steps to generate an overall sequence for representing the overall hidden state;

[0026] Step S24, integrating and feature extracting the hidden state through the hybrid attention mechanism, inputting the update gate, reset gate, candidate hidden state and hidden state data of the independent variable encoding gated recurrent unit to simplify the model structure.

[0027] Further, the hybrid attention mechanism includes a variable-by-variable time attention sub-mechanism and a feature attention sub-mechanism, and step S24 includes:

[0028] The overall hidden state is used to generate the influence weight of each time step in the overall sequence and the contribution of the hidden state of each time step through the variable-by-variable time attention sub-mechanism;

[0029] The contribution is dynamically adjusted through the feature attention mechanism.

[0030] Furthermore, the contribution is dynamically adjusted through the feature attention mechanism, specifically:

[0031] After concatenating the contribution with the hidden state of the current time step, the attention weight of the feature is calculated, the confidence is calculated according to the attention weight, and the confidence is used to adjust the hybrid attention mechanism.

[0032] Furthermore, the contribution of the global variable is calculated by quantifying the attention weight of a single variable in the hybrid attention layer of the hybrid attention mechanism;

[0033] Eliminate feature variables that are lower than the initial contribution threshold, use the source domain data again to train the hybrid attention mechanism, verify whether the diagnostic accuracy of the hybrid attention mechanism is consistent with that before the feature variables are eliminated, and if consistent, increase the initial contribution threshold to train and verify the hybrid attention mechanism again until the diagnostic accuracy is inconsistent with that before the feature variables are eliminated, and determine the current initial contribution threshold as the optimal contribution threshold;

[0034] Alternatively, a genetic algorithm is used, the contribution threshold is set as the individual and the chromosome, the diagnostic accuracy is set as the fitness, and the optimal contribution threshold is obtained.

[0035] Furthermore, step S3 includes the following steps:

[0036] Step S31, the adaptive focal loss function calculates the average accuracy of fault classification for multiple training data belonging to the same batch of samples, and calculates the individual accuracy of fault classification for a single training data of a single sample;

[0037] Step S32, calculating the degree of concern of various faults using the difference between the individual accuracy and the average accuracy;

[0038] The fault classification includes difficult-to-classify faults and easy-to-classify faults.

[0039] Further, in step S4, the adaptive focus hybrid maximum mean difference loss function includes an adaptive focus loss strategy and a maximum mean difference strategy.

[0040] Furthermore, the nuclear power plant multi-operating condition fault diagnosis system comprises the source domain nuclear power plant fault diagnosis model, the target domain nuclear power plant fault diagnosis model and a data distributor;

[0041] The data classifier uses KNN clustering to classify, and classifies the input data into the power level level closest to it by calculating the distance between the input data and the data sets of different power levels;

[0042] After preprocessing, the input data is identified by the data distributor to which the power level belongs, and the source domain nuclear power plant fault diagnosis model and the target domain nuclear power plant fault diagnosis model of the corresponding power level are input for fault diagnosis.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. The present invention improves the adaptability and fault diagnosis accuracy of the model under different working conditions through the division and transfer learning of multi-power level data.

[0045] 2. The present invention constructs a multivariate interpretable gated recurrent unit model, enhances the ability to quantify the contribution of input variables, and improves the interpretability of the model.

[0046] 3. The present invention adopts adaptive focal loss function and hybrid loss function to optimize the migration performance of the model between the source domain and the target domain, and improves the effectiveness of fault diagnosis.

[0047] 4. The present invention retains key variables through feature screening, simplifies the model structure, and ensures the accuracy of diagnostic performance.

[0048] 5. The multi-operating condition fault diagnosis system finally constructed by the present invention can efficiently diagnose nuclear power plant faults at various power levels, thereby improving the reliability and practicality of the diagnosis results.

[0049] 6. The method designed in the present invention is a universal multi-operating condition fault diagnosis method under different power levels and is suitable for fault diagnosis of similar nuclear power plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art description. In the drawings:

[0051] Figure 1 A schematic flow chart of a nuclear power plant fault diagnosis method based on interpretable deep transfer learning provided by an embodiment of the present invention;

[0052] Figure 2 A model operation logic flow chart of a nuclear power plant fault diagnosis method based on interpretable deep transfer learning provided by an embodiment of the present invention, taking a single sample as an example;

[0053] Figure 3 A histogram of the global contribution of each variable in the nuclear power plant fault diagnosis method based on interpretable deep transfer learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0055] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0056] like Figures 1 to 3 As shown, the present invention provides a nuclear power plant fault diagnosis method based on interpretable deep transfer learning, which integrates the operating condition data at various power levels and transmits them into a single model for training and diagnosis. The fault diagnosis results have high accuracy and the interpretability of the model is improved as much as possible.

[0057] like Figures 1 to 3 As shown, this embodiment provides a nuclear power plant fault diagnosis method based on interpretable deep transfer learning, and the nuclear power plant fault diagnosis method includes the following steps:

[0058] Step S1, obtaining real normal operation data, real fault operation data of a nuclear power plant at multiple power levels, and / or simulated normal operation data and simulated fault operation data at multiple power levels generated by a nuclear power plant simulator, dividing and preprocessing the real normal operation data, the real fault operation data and / or the simulated normal operation data, the simulated fault operation data according to their corresponding power levels to generate source domain data and target domain data;

[0059] It should be noted that since the simulated normal operation data and simulated fault operation data generated by the nuclear power plant simulator are used to supplement the real normal operation data and the real fault operation data respectively, when the real data is sufficient, there is no need for simulated data. At the same time, when the acquisition of real data is more difficult and the accuracy is low, only simulated data can be used.

[0060] Step S2: Simplify and construct a model structure of a multivariate encoding gated recurrent unit model using the source domain data and the target domain data;

[0061] Step S3, using the source domain data as input of the simplified coding gated recurrent unit model, training it through an adaptive focal loss function, and generating a source domain nuclear power plant fault diagnosis model of the source domain;

[0062] Step S4, migrating the source domain nuclear power plant fault diagnosis model to the target domain, and using the target domain data as input to train it through an adaptive focal hybrid maximum mean difference loss function to generate a target domain nuclear power plant fault diagnosis model;

[0063] Step S5, combining the source domain nuclear power plant fault diagnosis model and the target domain nuclear power plant fault diagnosis model in parallel to form a nuclear power plant multi-operating condition fault diagnosis system for detecting whether a nuclear power plant fails at multiple power levels;

[0064] The source domain is one or more power levels of a nuclear power plant, and the target domain is one or more power levels of a nuclear power plant excluding the source domain.

[0065] Furthermore, in step S1, the real normal operation data, the real fault operation data and / or the simulated normal operation data, the simulated fault operation data are divided and preprocessed according to their corresponding power levels to generate source domain data and target domain data, including:

[0066] Acquiring the real normal operation data and the real fault operation data and / or the simulated normal operation data and the simulated fault operation data of the nuclear power plant operating at multiple power levels;

[0067] The real normal operation data, the real fault operation data and / or the simulated normal operation data, and the simulated fault operation data under the maximum power level among multiple power levels are divided into the source domain data set, and the real normal operation data, the real fault operation data and / or the simulated normal operation data, and the simulated fault operation data under the remaining power levels are divided into multiple target domain data sets respectively;

[0068] The maximum value and the minimum value in the source domain data set are extracted, and the source domain data set and the target domain data set are normalized based on the maximum value and the minimum value to generate the source domain data and the target domain data.

[0069] In a preferred embodiment, real or simulated data of normal operating conditions and multiple fault conditions of a nuclear power plant under 100% power level, 90% power level, and 80% power level are obtained; the 100% power level is used as the source domain data set, and the remaining power levels are used as target domain data sets; the maximum and minimum values ​​of each variable in the source domain data set are extracted, and the source domain data set and each target domain data set are normalized based on this standard.

[0070] like Figure 1 As shown, step S1 includes the following specific steps:

[0071] S101: Acquire real normal operating data and real fault data of multiple power levels of a nuclear power plant, and / or multi-power level and multi-condition operating data generated by a simulator, divide the source domain data and the target domain data, and perform data preprocessing.

[0072] In the model training and verification process of the source domain nuclear power plant fault diagnosis model, ideally, actual nuclear power plant operating data should be used. However, due to the extremely low frequency of nuclear power plant emergencies, the relevant data resources are extremely limited, and it is difficult to obtain enough actual data to support the effective training and verification of the model. This embodiment is designed using a computer transient analyzer (PCTRAN) to simulate various operating scenarios and accident conditions of nuclear power plants. PCTRAN is a professional software tool that simulates various operating scenarios and accident conditions of nuclear power plants. It was developed by Micro-Simulation Technology (MST) Inc. and allows real-time simulation of nuclear reactor transients and accidents on personal computers.

[0073] In the disclosed embodiment, PCTRAN was used to simulate the operating data of the nuclear power plant at 100%, 90% and 80% power levels. At each power level, the stable operation was first simulated for 60 seconds, and then 16 different faults were inserted respectively, and the stable operation was continued without inserting a fault for 500 seconds. Through this process, a total of 17 groups of operating data under different working conditions (16 groups of fault working condition data and one group of normal operating condition data) were finally obtained at each power level. The 100% power level data was set as the source domain data, and the 90% and 80% power levels were set as the target domain data, which was convenient for the subsequent transfer learning model training.

[0074] The data obtained through PCTRAN contains 81 characteristic variables (such as temperature, pressure, liquid level, flow, control lever position, valve opening, etc.), and the distribution dimensions of each data are not consistent. In order to ensure that different features are comparable during the model training process and avoid unnecessary deviations of the model due to the large value range of some features, it is necessary to normalize the data. The maximum and minimum values ​​of each characteristic variable of the 100% power level data are extracted as x in data normalization. imax With x imin The simulation data at 100%, 90% and 80% power levels are normalized in turn. The normalization operation is specifically as follows:

[0075]

[0076] In the formula, X on the right side of the equation is i is the output, normalized simulation data, and X on the right side of the equation i is the input simulation data, X imin is the minimum value of the characteristic variable and power level to which the input simulation data belongs, X imax is the maximum value of the characteristic variable and power level to which the input simulation data belongs.

[0077] For the fault diagnosis model used later, since it is essentially a time series model, and the fault data of nuclear power plants presents continuous time series characteristics, including important time-dependent information such as parameter trend changes, it is necessary to reasonably segment the data of each power level. Specifically, the original data will be segmented into a time window of 5 seconds to generate a series of time series samples so that the model can better capture the dynamic change characteristics in the data.

[0078] S102: Use the preprocessed data to construct a multivariate interpretable gated recurrent unit model for training to extract the contribution of each variable, extract key variables based on the contribution without significantly affecting the model performance, and simplify the model structure.

[0079] Further, in step S2, the multivariable encoding gated recurrent unit model includes an independent variable encoding gated recurrent unit and a hybrid attention mechanism;

[0080] The independent variable encoding gated recurrent unit encodes the hidden states of the source domain data and the target domain data;

[0081] The hybrid attention mechanism is connected to the independent variable encoding gated recurrent unit to integrate the hidden state and output the global variable contribution;

[0082] The independent variable encoding gated recurrent unit obtains an optimal contribution threshold according to the global variable contribution, and simplifies the model structure according to the optimal contribution threshold.

[0083] Furthermore, the step S2 comprises the following steps:

[0084] Step S21, dividing the input source domain data and the target domain data according to the time step and performing matrix dot multiplication with the corresponding weights to obtain a first output matrix;

[0085] Step S22, performing matrix multiplication on the hidden state and the corresponding weight matrix to obtain a second output matrix;

[0086] Step S23, accumulating the hidden state through a cycle of time steps to generate an overall sequence for representing the overall hidden state;

[0087] Step S24, integrating and feature extracting the hidden state through the hybrid attention mechanism, inputting the update gate, reset gate, candidate hidden state and hidden state data of the independent variable encoding gated recurrent unit to simplify the model structure.

[0088] The multivariable interpretable gated recurrent unit model is a new deep learning model with interpretability, which consists of two parts: an independent variable encoding gated recurrent unit and a hybrid attention mechanism.

[0089] In a preferred embodiment, in step S2, the independent variable encoding gated recurrent unit specifically encodes the hidden state of a single variable by adjusting the operational logic of the basic unit of the gated recurrent unit, thereby facilitating the quantitative extraction and analysis of the contribution of each input variable in the subsequent stage; the hybrid attention mechanism is connected to the independent variable encoding gated recurrent unit to integrate the hidden states of each input variable, thereby outputting the evaluation results of the model on the input data diagnosis and the global variable contribution in the quantitative training process.

[0090] In a preferred embodiment, the independent variable encoding gated recurrent unit includes an update gate z t , reset gate t , candidate hidden state c t and the hidden state h t ; The independent variable encoding gated recurrent unit redesigns the linear transformation of the input vector x and the hidden state h based on the original gated recurrent unit. The hidden state and gate matrix in the independent variable encoding gated recurrent unit are both represented by ~ to distinguish them from the operations in the gated recurrent unit. First, define the input variable X, where X∈R T×N , T represents the time step, the input vector x at the tth time step t, where x t ∈R 1×N , N represents the number of variables, and the corresponding weight Where d represents the number of neurons in the layer. During the operation, x t Each variable in Each column is element-wise multiplied (dot product) to obtain the output matrix o x , o x ∈R n×d For the hidden state The linear transformation of The corresponding weight matrix In this operation, and Perform a standard matrix multiplication operation to obtain the output matrix o h , o h ∈R n×d Through the cyclic accumulation of time steps, the overall hidden state The construction method preserves the independent characteristics of each input variable at all time steps, forming a complete hidden state representation. can be viewed as consisting of independent hidden states at each time step The sequence composed of in The hidden state output by the independent variable encoding gated recurrent unit will be further integrated and feature extracted by the subsequent hybrid attention mechanism. The specific operation process of the independent variable encoding gated recurrent unit is as follows:

[0091]

[0092] In the formula, the definitions of each variable are as described above, and the superscript ~ indicates that the independent variable encodes the gated recurrent unit. represents the update gate of the independent variable encoding gated recurrent unit, Reset door, Candidate hidden states, Hidden state, where the subscript t represents the tth time step, and the subscript t-1 represents the t-1th time step, and represents the corresponding weight matrix, and represents the dot multiplication along the N-axis.

[0093] Output Matrix In the formula, x represents the corresponding input variable, and w represents the weight of the corresponding input variable.

[0094] Further, the hybrid attention mechanism includes a variable-by-variable time attention sub-mechanism and a feature attention sub-mechanism, and step S24 includes:

[0095] The overall hidden state is used to generate the influence weight of each time step in the overall sequence and the contribution of the hidden state of each time step through the variable-by-variable time attention sub-mechanism;

[0096] The contribution is dynamically adjusted through the feature attention mechanism.

[0097] Furthermore, the contribution is dynamically adjusted through the feature attention mechanism, specifically:

[0098] After concatenating the contribution with the hidden state of the current time step, the attention weight of the feature is calculated, the confidence is calculated according to the attention weight, and the confidence is used to adjust the hybrid attention mechanism.

[0099] In a preferred embodiment, the hybrid attention mechanism first applies a variable-by-variable time attention mechanism to the hidden state generated by the independent variable encoding gated recurrent unit to dynamically allocate the influence weights of different time steps on the overall sequence representation: the variable-by-variable time attention mechanism calculates the relative contribution of the hidden state of each time step in the sequence through a weighted sum operation; based on the global feature representation generated by the variable-by-variable time attention mechanism, the model further applies a feature attention mechanism to dynamically adjust the relative contribution of each input feature in the sequence: the feature attention mechanism concatenates the temporal attention output with the hidden state of the current time step through a linear transformation, and calculates the attention weight of each feature; these two steps together constitute a probabilistic hybrid model, which is specifically:

[0100]

[0101] in, is the temporal attention weight, is the feature attention weight, Confidence output for model diagnosis.

[0102] Furthermore, the contribution of the global variable is calculated by quantifying the attention weight of a single variable in the hybrid attention layer of the hybrid attention mechanism;

[0103] Eliminate feature variables that are lower than the initial contribution threshold, use the source domain data again to train the hybrid attention mechanism, verify whether the diagnostic accuracy of the hybrid attention mechanism is consistent with that before the feature variables are eliminated, and if consistent, increase the initial contribution threshold to train and verify the hybrid attention mechanism again until the diagnostic accuracy is inconsistent with that before the feature variables are eliminated, and determine the current initial contribution threshold as the optimal contribution threshold;

[0104] It should be noted that the consistent preference refers to that the difference between the diagnostic accuracy of the hybrid attention mechanism after training and the diagnostic accuracy before feature variable elimination is within a set range.

[0105] Alternatively, a genetic algorithm is used, the contribution threshold is set as the individual and the chromosome, the diagnostic accuracy is set as the fitness, and the optimal contribution threshold is obtained.

[0106] In a preferred embodiment, the step S2 extracts the contribution of each feature and the feature screening, and the model quantifies the contribution of the global variable by the attention weight of a single variable in the mixed attention layer to extract the specific contribution of the variable in the model training. The calculation formula of the global variable contribution is n as follows:

[0107]

[0108] in Indicates normalization. Where M is the number of samples in the training set, and T is the number of training times.

[0109] After obtaining the contribution of the global variable, a lower contribution threshold can be set to eliminate feature variables below the threshold, and the source domain data can be used again for model training to verify whether the model's diagnostic accuracy is equivalent to that before feature screening. If the diagnostic accuracy has not decreased, the contribution threshold can be appropriately increased for screening and training again until the optimal contribution threshold is found. Alternatively, a genetic algorithm can be used to set the contribution threshold to individuals and chromosomes, and the diagnostic accuracy to fitness to obtain the optimal contribution threshold.

[0110] Furthermore, step S3 includes the following steps:

[0111] Step S31, the adaptive focal loss function calculates the average accuracy of fault classification for multiple training data belonging to the same batch of samples, and calculates the individual accuracy of fault classification for a single training data of a single sample;

[0112] Step S32, calculating the degree of concern of various faults using the difference between the individual accuracy and the average accuracy;

[0113] The fault classification includes difficult-to-classify faults and easy-to-classify faults.

[0114] In a preferred embodiment, the adaptive focus loss function in step S3 is improved on the original focus loss function. The adaptive focus loss function analyzes the fault classification in the training data and divides the fault classification into difficult-to-classify faults and easy-to-classify faults according to the average accuracy of the batch samples; then, the difference between the accuracy of each fault classification and the average accuracy of the batch is used to measure the degree of attention required for each fault in subsequent training:

[0115]

[0116] It means that η c represents the accuracy of the previous batch of fault classification c in model training, Indicates the initial value of the adjustable focus parameter.

[0117] Further, in step S4, the adaptive focus hybrid maximum mean difference loss function includes an adaptive focus loss strategy and a maximum mean difference strategy.

[0118] In a preferred embodiment, the adaptive focus maximum mean difference hybrid loss function in step S4 combines the adaptive focus loss and maximum mean difference strategies:

[0119]

[0120] in is the target domain classification label, Predict the output label for the target domain, X S is the source domain data, X T The target domain data.

[0121] Furthermore, the nuclear power plant multi-operating condition fault diagnosis system comprises the source domain nuclear power plant fault diagnosis model, the target domain nuclear power plant fault diagnosis model and a data distributor;

[0122] The data classifier uses KNN clustering to classify, and classifies the input data into the power level level closest to it by calculating the distance between the input data and the data sets of different power levels;

[0123] After preprocessing, the input data is identified by the data distributor to which the power level belongs, and the source domain nuclear power plant fault diagnosis model and the target domain nuclear power plant fault diagnosis model of the corresponding power level are input for fault diagnosis.

[0124] In a preferred embodiment, the nuclear power plant multi-condition fault diagnosis system described in step S5 includes: a source domain model, a target domain model and a data distributor; the classifier uses KNN clustering to perform classification, and classifies the input data into the nearest power level by calculating the distance between the input data and different power level data sets, so as to determine the operating condition to which it belongs; after the input data is preprocessed, the data distributor identifies the operating power level to which it belongs, and inputs the corresponding power level model for fault diagnosis.

[0125] The multivariate interpretable gated recurrent unit model is constructed and initially trained.

[0126] In the training model parameter settings, the number of input parameters is set to 29, the number of neurons is set to 64, the number of output parameters is set to 17, the batch size is set to 64, the optimizer selects the RMSprop algorithm, the learning rate is uniformly set to 0.001, the number of training iterations is set to 300 times, the loss function uses the adaptive focus loss function, and the weights of each layer of the model are randomly initialized. The data after 100% horizontal power screening is used as the input of the model for training again, and the data is randomly divided into training set and test set in a ratio of 7 to 3 to verify the performance of the model.

[0127] Take the training of a single time series training sample for a single training as an example, Figure 2As shown in the figure, the independent variable encoding gated recurrent unit can be roughly regarded as a group of 81 parallel gated recurrent units. Each unit undertakes an independent task, that is, processing the dynamic changes of a specific input variable among the 81 input variables in the entire time series, and then generating a hidden state unique to the variable. After extracting the hidden state, the 81 gated recurrent units are aggregated into an overall hidden state, and its dimensional shape should be 5 time windows, 81 variables, and 64 neurons, ensuring that the feature extraction process of each input variable is independent. In the hybrid attention mechanism, the variable-by-variable temporal attention sub-mechanism is applied to the hidden state generated by the independent variable encoding gated recurrent unit, and the weight of the influence of each time step on the overall sequence representation is dynamically allocated. Through the weighted sum operation, the temporal attention sub-mechanism calculates the relative importance of different time steps in the sequence, thereby generating a feature representation of the time series, and the data dimension becomes: 81 variables and 64 neurons. The model further applies the feature attention sub-mechanism to dynamically adjust the importance of the input features. Specifically, the output of the temporal attention is cascaded with the hidden state of the current time step to form a feature state representation. The feature attention sub-mechanism calculates the attention weight of each feature, which reflects the relative contribution of each input feature to the final prediction result. Then, these feature weights are weighted and summed through linear transformation to generate the final diagnostic output of the model. The relative contribution of the features of all sample data is plotted, and a re-quantification process is provided to obtain an overall evaluation of the global contribution of each variable.

[0128] Then, an adaptive focal loss function is used to diagnose the difference between the category and the true category. Before the model reaches the set number of iterations, the stochastic gradient descent algorithm is used to iteratively optimize the weights of each target category in the initial model along the negative gradient direction of the loss function until the model converges.

[0129] The adaptive focus loss function used in this embodiment is expressed as follows:

[0130]

[0131] In the formula, the definition of each variable is the definition of the conventional loss function.

[0132] In the example disclosed in the present invention, the global contribution of each variable extracted is as follows: Figure 3 As shown, it can be seen that most variables contribute little. To this end, a contribution value threshold can be set to perform feature screening for variables below the set threshold. Briefly select 0.01, 0.0125 and 0.015 as contribution thresholds, screen and eliminate variables below these thresholds, and retrain the model for the screened features to verify the accuracy of the model at different thresholds. Train again and verify the model accuracy with the test set. The accuracy of the model retraining after contribution screening is shown in the following table:

[0133] Contribution threshold The number of variables after screening Retraining Accuracy 0 81 98.82％ 0.01 43 98.94％ 0.0125 29 98.75％ 0.015 14 86.94％

[0134] Therefore, it can be preliminarily concluded that the 29 characteristic variables screened out using a contribution threshold of 0.0125 have met the needs of fault diagnosis. For this reason, the preprocessed data at each power level are processed again and screened to contain only the samples of these 29 data.

[0135] S103: Construct a simplified multivariate interpretable gated recurrent unit model with source domain data as input, and use an adaptive focal loss function to train the source domain fault diagnosis model.

[0136] Construct simplified training model parameters, set the number of input parameters to 29, the number of neurons to 64, the output parameters to 17, the batch size to 64, select the RMSprop algorithm for the optimizer, set the learning rate to 0.001, set the number of training iterations to 300, select the adaptive focus loss function for the loss function, and randomly initialize the weights of each layer of the model. Use the data after 100% horizontal power screening as the input of the model for training again, and randomly divide the data into training set and test set in a ratio of 7 to 3 to verify the performance of the model.

[0137] The model with the highest accuracy based on the training set during the training process is saved as the diagnosis model for 100% power level faults and the pre-trained model for training the remaining faults.

[0138] S104: Migrate the trained source domain model to the target domain as a pre-trained model, and further train the target domain data through an adaptive focal hybrid maximum mean difference loss function.

[0139] Based on the pre-trained model obtained in S103, a fault diagnosis model for 90% and 80% power levels is constructed, retaining the model structure (number of input parameters, number of neurons and number of output parameters) and the weights of each layer.

[0140] RMSprop is used as the optimization algorithm, and the learning rate of the independent variable encoding gated recurrent unit layer is set to 0.0001, and the learning rate of the hybrid attention mechanism layer is set to 0.001. The reason for this setting is that the independent variable encoding gated recurrent unit is used to extract low-level features of the sequence. At different power levels, the basic patterns of these low-level features are relatively similar, so only a smaller learning rate is needed for fine-tuning to retain the existing knowledge of the source domain model; the hybrid attention mechanism layer is used to summarize the relationship between input features and re-evaluate the importance of each feature. Using a larger learning rate helps the model adjust high-level features faster and adapt to new power levels. The loss function uses the adaptive focal maximum mean difference hybrid loss function:

[0141]

[0142] in is the target domain classification label, Predict the output label for the target domain, X S is the source domain data, X T is the target domain data, and λ is set to λ=2 / (1+e -10p )-1, so that the loss function is L in the early stage of training MMD Dominant in the late training by L AF , p represents the current training times / total training times. This setting helps the model balance the needs of reducing distribution differences and improving classification performance at different stages of training.

[0143] Since the model has learned a lot of useful features in the source domain, only a small-scale fine-tuning is needed to complete the training in the target domain. At the same time, too high an iteration number may cause overfitting. Therefore, the iteration number is set to 50.

[0144] The data after 90% and 80% power screening were used as the input of the model for transfer learning, and the data were randomly divided into training set and test set in a ratio of 7 to 3 to verify the model performance.

[0145] The model with the highest accuracy based on the training set during the training process is saved as the diagnosis model for 90% and 80% power level faults.

[0146] S105: The above source domain model and target domain model are combined in parallel to form a final nuclear power plant multi-operating condition fault diagnosis system, and the data to be tested are clustered and input into the fault diagnosis system to obtain a fault diagnosis result.

[0147] The fault diagnosis models of 100%, 90% and 80% power levels obtained by training S103 and S104 are connected in parallel to form a multi-condition fault diagnosis system. The test set obtained by random segmentation of S103 and S104 is used to verify the performance of the diagnosis system. The power level of the sample data is identified by the KNN clustering classifier and sent to the corresponding model for identification and diagnosis.

[0148] This embodiment aims at the fault diagnosis problem of nuclear power plants at different power levels, and discloses a nuclear power plant fault diagnosis method based on interpretable deep transfer learning. Based on the multivariable interpretable gated recurrent unit model, by introducing the adaptive focal loss function and the maximum mean difference mixed loss strategy, the method used in this embodiment can not only effectively handle the fault diagnosis tasks at different power levels, but also significantly improve the accuracy and interpretability of the model. In addition, by introducing the hybrid attention mechanism, this embodiment enables the model to provide good interpretability while improving the accuracy of fault diagnosis, which helps to deeply understand the impact of each input variable on the model decision under different operating conditions. Visualization and variable importance analysis further verify the effectiveness and feasibility of the model, and provide a new solution for the intelligence and reliability improvement of nuclear power plant fault diagnosis systems in the future.

[0149] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0150] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

Claims

1. A nuclear power plant fault diagnosis method based on interpretable deep transfer learning, characterized in that: include: Step S1, obtaining real normal operation data, real fault operation data of a nuclear power plant at multiple power levels, and / or simulated normal operation data and simulated fault operation data at multiple power levels generated by a nuclear power plant simulator, dividing and preprocessing the real normal operation data, the real fault operation data and / or the simulated normal operation data, the simulated fault operation data according to their corresponding power levels to generate source domain data and target domain data; Step S2: Simplify and construct a model structure of a multivariate encoding gated recurrent unit model using the source domain data and the target domain data; Step S3, using the source domain data as input of the simplified coding gated recurrent unit model, training it through an adaptive focal loss function, and generating a source domain nuclear power plant fault diagnosis model of the source domain; Step S4, migrating the source domain nuclear power plant fault diagnosis model to the target domain, and using the target domain data as input to train it through an adaptive focal hybrid maximum mean difference loss function to generate a target domain nuclear power plant fault diagnosis model; Step S5, combining the source domain nuclear power plant fault diagnosis model and the target domain nuclear power plant fault diagnosis model in parallel to form a nuclear power plant multi-operating condition fault diagnosis system for detecting whether a nuclear power plant fails at multiple power levels; The source domain is one or more power levels of a nuclear power plant, and the target domain is one or more power levels of a nuclear power plant excluding the source domain.

2. The method for nuclear power plant fault diagnosis based on interpretable deep transfer learning according to claim 1, characterized in that: In the step S1, the real normal operation data, the real fault operation data and / or the simulated normal operation data, the simulated fault operation data are divided and preprocessed according to their corresponding power levels to generate source domain data and target domain data, including: Acquiring the real normal operation data and the real fault operation data and / or the simulated normal operation data and the simulated fault operation data of the nuclear power plant operating at multiple power levels; The real normal operation data, the real fault operation data and / or the simulated normal operation data, and the simulated fault operation data under the maximum power level among multiple power levels are divided into the source domain data set, and the real normal operation data, the real fault operation data and / or the simulated normal operation data, and the simulated fault operation data under the remaining power levels are divided into multiple target domain data sets respectively; The maximum value and the minimum value in the source domain data set are extracted, and the source domain data set and the target domain data set are normalized based on the maximum value and the minimum value to generate the source domain data and the target domain data.

3. The method for nuclear power plant fault diagnosis based on interpretable deep transfer learning according to claim 1, characterized in that: In the step S2, the multivariate encoding gated recurrent unit model includes an independent variable encoding gated recurrent unit and a hybrid attention mechanism; The independent variable encoding gated recurrent unit encodes the hidden states of the source domain data and the target domain data; The hybrid attention mechanism is connected to the independent variable encoding gated recurrent unit to integrate the hidden state and output the global variable contribution; The independent variable encoding gated recurrent unit obtains an optimal contribution threshold according to the global variable contribution, and simplifies the model structure according to the optimal contribution threshold.

4. The method for nuclear power plant fault diagnosis based on interpretable deep transfer learning according to claim 3 is characterized in that: The step S2 comprises the following steps: Step S21, dividing the input source domain data and the target domain data according to the time step and performing matrix dot multiplication with the corresponding weights to obtain a first output matrix; Step S22, performing matrix multiplication on the hidden state and the corresponding weight matrix to obtain a second output matrix; Step S23, accumulating the hidden state through a cycle of time steps to generate an overall sequence for representing the overall hidden state; Step S24, integrating and feature extracting the hidden state through the hybrid attention mechanism, inputting the update gate, reset gate, candidate hidden state and hidden state data of the independent variable encoding gated recurrent unit to simplify the model structure.

5. The method for nuclear power plant fault diagnosis based on interpretable deep transfer learning according to claim 3 is characterized in that: The hybrid attention mechanism includes a variable-by-variable time attention sub-mechanism and a feature attention sub-mechanism, and step S24 includes: The overall hidden state is used to generate the influence weight of each time step in the overall sequence and the contribution of the hidden state of each time step through the variable-by-variable time attention sub-mechanism; The contribution is dynamically adjusted through the feature attention mechanism.

6. The method for nuclear power plant fault diagnosis based on interpretable deep transfer learning according to claim 5, characterized in that: The dynamic adjustment of contribution through the feature attention mechanism is as follows: After concatenating the contribution with the hidden state of the current time step, the attention weight of the feature is calculated, the confidence is calculated according to the attention weight, and the confidence is used to adjust the hybrid attention mechanism.

7. The method for nuclear power plant fault diagnosis based on interpretable deep transfer learning according to claim 3, characterized in that: The contribution of the global variable is calculated by quantifying the attention weight of a single variable in the hybrid attention layer of the hybrid attention mechanism; Eliminate feature variables that are lower than the initial contribution threshold, use the source domain data again to train the hybrid attention mechanism, verify whether the diagnostic accuracy of the hybrid attention mechanism is consistent with that before the feature variables are eliminated, and if consistent, increase the initial contribution threshold to train and verify the hybrid attention mechanism again until the diagnostic accuracy is inconsistent with that before the feature variables are eliminated, and determine the current initial contribution threshold as the optimal contribution threshold; Alternatively, a genetic algorithm is used, the contribution threshold is set as the individual and the chromosome, the diagnostic accuracy is set as the fitness, and the optimal contribution threshold is obtained.

8. The method for nuclear power plant fault diagnosis based on interpretable deep transfer learning according to claim 1, characterized in that: The step S3 comprises the following steps: Step S31, the adaptive focal loss function calculates the average accuracy of fault classification for multiple training data belonging to the same batch of samples, and calculates the individual accuracy of fault classification for a single training data of a single sample; Step S32, calculating the degree of concern of various faults using the difference between the individual accuracy and the average accuracy; The fault classification includes difficult-to-classify faults and easy-to-classify faults.

9. The method for nuclear power plant fault diagnosis based on interpretable deep transfer learning according to any one of claims 1 to 8, characterized in that: In step S4, the adaptive focus hybrid maximum mean difference loss function includes an adaptive focus loss strategy and a maximum mean difference strategy.

10. The method for nuclear power plant fault diagnosis based on interpretable deep transfer learning according to any one of claims 1 to 8, characterized in that: The multi-operating-condition fault diagnosis system of a nuclear power plant comprises the source domain nuclear power plant fault diagnosis model, the target domain nuclear power plant fault diagnosis model and a data distributor; The data classifier uses KNN clustering to classify, and classifies the input data into the power level level closest to it by calculating the distance between the input data and the data sets of different power levels; After preprocessing, the input data is identified by the data distributor to which the power level belongs, and the source domain nuclear power plant fault diagnosis model and the target domain nuclear power plant fault diagnosis model of the corresponding power level are input for fault diagnosis.

Images