Fault early warning method for intermediate range detector of out-of-pile nuclear measurement system

Abstract

The invention relates to the technical field of fault early warning of an out-of-pile nuclear measurement system intermediate range detector, and discloses an out-of-pile nuclear measurement system intermediate range detector fault early warning method, which comprises the following steps of: acquiring channel current data through an out-of-pile nuclear measurement system intermediate range detector, and filling the data by using linear interpolation; the method comprises the following steps: downsampling original training data and test data, identifying an abnormal value in current data of an intermediate range channel, resetting the abnormal value as a null value, interpolating the null value by using linear interpolation, performing sliding window processing on the data subjected to abnormal value processing, calculating mutual information in pairs of current of each channel of the intermediate range, and calculating the current of each channel of the intermediate range. Fault time identification is carried out by judging mutual information size changes, a current signal of the intermediate range detector is predicted from fault starting time, and a confidence interval is constructed to carry out fault early warning. According to fault time identification based on mutual information, the time characteristics of the current data of the intermediate range channel are effectively utilized.

Description

Technical Field

[0001] The present invention relates to the technical field of fault early warning of an intermediate-range detector of an ex-core nuclear measurement system, and in particular to a fault early warning method of an intermediate-range detector of an ex-core nuclear measurement system. Background Art

[0002] To ensure that nuclear power units operate in a safe and stable state and avoid various possible risks, it is necessary to monitor the operating status of nuclear power units. The off-core nuclear measurement system is the basis for realizing reactor core monitoring, and it plays an important role in ensuring the safety of nuclear power units. The off-core nuclear measurement system detector is used to measure the power of the reactor core. Its importance is self-evident both for the safe operation of nuclear power units and for the economic benefits of nuclear power plants. Fault warning can detect potential abnormal conditions of equipment before it fails completely, preventing equipment failure from developing into serious accidents. For the fault warning of the intermediate-range detector of the off-core nuclear measurement system, it is necessary to identify the fault start time of the channel current data and judge the channel current trend to provide relevant operation and maintenance decision-making information. The data-driven intermediate-range detector fault warning method is a method that meets the needs of safe operation of the nuclear power system and the economic benefits of the nuclear power plant, and has important engineering significance.

[0003] At present, some scholars at home and abroad have realized fault diagnosis of neutron detectors through various methods such as generalized likelihood ratio, principal component analysis, kmeans algorithm, etc. At present, there are some abnormality and fault detection methods for neutron detector equipment in the ex-core nuclear measurement system, which have failed to warn of the fault of the intermediate range detector. Fault warning is a predictive maintenance method, the purpose of which is to take measures before the fault occurs and during the acquisition to avoid equipment damage and production interruption. By analyzing and modeling the historical data of the equipment, a prediction model is established, and future data is predicted, and faults are warned. At present, there is little research on fault warning of intermediate range detectors in ex-core nuclear measurement systems at home and abroad. Summary of the invention

[0004] The purpose of the present invention is to provide a method for early warning of a fault of an intermediate-range detector of an off-core nuclear measurement system, thereby providing a solution for the latent early warning of a fault of the intermediate-range detector.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A method for early warning of a fault of an intermediate range detector of an out-of-core nuclear measurement system comprises the following steps:

[0007] Step 1: Obtain channel current data through the intermediate range detector of the out-of-core nuclear measurement system;

[0008] Step 2: Use linear interpolation to fill the data to ensure that the data sampling frequency is consistent;

[0009] Step 3: Data downsampling: downsampling the original training data and test data;

[0010] Step 4: Identify outliers in the current data of the mid-range channel and reset the outliers to null values;

[0011] Step 5: Use linear interpolation to interpolate the null values ​​to ensure that the processed curve is highly similar to the original curve characteristics;

[0012] Step 6: Perform sliding window processing on the data after outlier processing;

[0013] Step 7: Calculate the mutual information of each channel current in the middle range. The calculation formula is as follows:

[0014]

[0015] Where I(X; Y) represents the mutual information of variables X and Y. When X and Y are completely independent, I(X; Y) is the smallest zero. p(x), p(y), and p(x, y) represent the marginal probability distribution function of variable X, the marginal probability distribution function of Y, and the joint probability distribution function of X and Y, respectively.

[0016] Step 8: Identify the fault time by judging the change in mutual information;

[0017] Step 9: Using the NerualProphet model, predict the current signal of the intermediate range detector from the fault initiation time;

[0018] Step 10: Construct a confidence interval and conduct fault warning.

[0019] In step 3, the data is downsampled using a sliding window.

[0020] Furthermore, a sliding window with a length of 3600 and a step size of 3600 is used to sample the original training data and test data.

[0021] In step 4, outliers are identified using the constant fluctuation detector differencing algorithm.

[0022] In step 6, the left pointer of the window is always fixed at the leftmost position of the array, and the right pointer of the window is shifted 24 bits to the right after each sliding window operation to obtain the current time characteristics of the intermediate range channel.

[0023] In step 7, the formula Calculate information entropy,

[0024] In the formula, H(X) is the information entropy value, p(x) is the probability of taking the value x, and b is usually taken as the base of the natural constant e.

[0025] In step 9, the Neural Prophet model is composed of the Prophet model and the AR-Net model p-order autoregressive model.

[0026] The Prophet model is:

[0027] P(t)=g(t)+s(t)+h(t)+ε t ,

[0028]

[0029] In the formula, g(t) represents the trend term; C(t) represents the carrying capacity, which is a function that changes with time and limits the maximum value of the curve fitting; s(t) represents the seasonal term; h(t) represents the holiday term; κ~Normal(0,v 2 ) obeys the normal distribution, κ i represents the impact of the event on the model, i represents holidays or other events, and D i represents the variable of the ith independent model; ε t represents the residual term and follows a Gaussian distribution, indicating the fluctuations that the model is subject to when fitting;

[0030] The p-order autoregressive model of the AR-Net model is:

[0031]

[0032] In the formula, y t-1 ,…,y t-p represents the p-order lag term, w i is the p-order weight or autoregressive coefficient, e t represents noise, c represents a constant term;

[0033] The MSE calculation formula is as follows, which is an evaluation indicator used to calculate the loss function:

[0034]

[0035] In order to introduce sparsity to the autoregressive coefficients, the regular term R is added to the loss function, and the loss function calculation formula is:

[0036]

[0037] In the formula, s represents the sparsity of the autoregressive coefficient, c λ represents the degree of regularization, c1 and c2 represent the regularization parameters.

[0038] In step 10, six sigma is used (similar to the 3sigema function, the internal parameter 3 is set to 6 to construct a larger confidence interval to avoid false alarms) to construct a confidence interval. By constructing confidence intervals for the maximum channel current prediction value and the fault channel current prediction value, it is inferred that the time when the mid-range channel current ratio is less than 1.9 times.

[0039] Compared with the prior art, the method for early warning of faults of intermediate-range detectors of an out-of-core nuclear measurement system provided by the present invention has the following beneficial effects:

[0040] The data cleaning process provided by the present invention can effectively solve the problem of current data in the intermediate range channel and improve the data quality.

[0041] The present invention recognizes fault time based on mutual information and effectively utilizes the time characteristics of current data of the intermediate range channel.

[0042] Furthermore, the NerualProphet model provided by the present invention introduces AR-Net to construct time series autocorrelation, gradient descent optimization method, feedforward neural network and other functions on the basis of the Prophet model, which effectively improves the prediction accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for the technical description.

[0044] Figure 1 The present invention provides a flowchart of a method for early warning of a fault of an intermediate-range detector of an extra-core nuclear measurement system. DETAILED DESCRIPTION

[0045] The following is further explained in detail through specific implementation methods.

[0046] like Figure 1 As shown, the present invention provides a method for early warning of faults of an intermediate range detector of an out-of-core nuclear measurement system, comprising the following steps:

[0047] Step 1: Obtain channel current data through the intermediate range detector of the extra-core nuclear measurement system.

[0048] Step 2: To address the problem of unstable data sampling frequency, use linear interpolation to fill the data to ensure consistent data sampling frequency.

[0049] Step 3: Data downsampling, convert the original frequency of the data to another frequency, if it is from high frequency to low frequency (conversion of data sampling point frequency, for example, originally one point was sampled every minute, now after sampling, one point can be sampled every ten minutes). To address the data redundancy problem, use a sliding window to downsample the data. Use a sliding window with a length of 3600 and a step size of 3600 to sample the original training data and test data.

[0050] Step 4: To address the issues of large data fluctuations and data drift, use the constant fluctuation detector differential algorithm to identify abnormal values ​​in the current data of the mid-range channel and reset them to null values.

[0051] Step 5: Use linear interpolation to interpolate the null values ​​to ensure that the processed curve is highly similar to the original curve characteristics.

[0052] Step 6: Perform sliding window processing on the data after outlier processing. The left pointer of the window is always fixed to the leftmost side of the array, and the right pointer of the window moves 24 bits to the right after each sliding window operation to obtain the current time characteristics of the intermediate range channel.

[0053] Step 7: The detectors of each channel in the middle range are evenly distributed and have spatial correlation. The mutual information of each channel current in the middle range is calculated pairwise.

[0054] In information theory, information entropy is a basic unit of information that can be used to measure the degree of confusion between variables. The larger the information entropy, the greater the degree of confusion of the variables, and the greater the amount of information contained in the variables. The smaller the information entropy, the smaller the degree of confusion of the variables, and the more limited the amount of information contained. For any variable, the calculation formula for information entropy is:

[0055]

[0056] Where: H(X) is the information entropy value, p(x) is the probability of taking the value x, and b is usually taken as the base of the natural constant e as the logarithm.

[0057] Mutual Information (MI) in information theory can represent the correlation between two random variables X and Y, that is, it measures the degree of dependence between the two variables. It can be regarded as the amount of information that a random variable X contains about another random variable Y, that is, the stronger the correlation between variables X and Y, the greater the mutual information value, and vice versa. The formula for calculating the mutual information of two random variables X and Y is:

[0058]

[0059] Where: I(X; Y) represents the mutual information of variables X and Y. When X and Y are completely independent, I(X; Y) is the smallest, which is zero. p(x), p(y), and p(x, y) represent the marginal probability distribution function of variable X, the marginal probability distribution function of Y, and the joint probability distribution function of X and Y, respectively.

[0060] Step 8: Identify the fault time by judging the change in the mutual information size. When the mutual information size continues to decrease after a certain time point while showing an upward trend, the detector is faulty, otherwise the detector is not faulty.

[0061] Step 9: Use the NerualProphet model to predict from the fault start time.

[0062] It should be noted that the Neural Prophet model is a decomposable time series model developed based on Pytorch. It continues the main functions of Prophet that the Facebook team open-sourced in 2017, and introduces AR-Net to build time series autocorrelation, gradient descent optimization, feedforward neural network and other functions. The Neural Prophet model consists of Prophet and AR-Net. Among them, Prophet is a self-additive combination, which is composed of three items: trend item, seasonal item and holiday item. The basic formula is:

[0063] P(t)=g(t)+s(t)+h(t)+ε t

[0064] g(t) represents the trend term, which represents the non-periodic trend of the time series. The nonlinear curve can be fitted by adjusting the smoothness parameter. The calculation formula is:

[0065]

[0066] Where: C(t) represents the capacity, which is a function that changes over time and limits the maximum value of the curve fitting. k(t) represents the growth rate, which can be changed by algorithms or manual specification when the model faces certain specific moments or potential cycles. m(t) represents the offset. When the growth rate k(t) changes, the corresponding m(t) will also be adjusted.

[0067] s(t) represents the seasonal term, or periodic term, which is usually expressed in weeks or years. The Fourier series is used to approximate the periodic properties. The calculation formula is:

[0068]

[0069] Where: N represents the number of cycle types used in the model. A larger N value can fit a more complex seasonal curve. If it is too large, it is easy to overfit. S Represents a fixed period.

[0070] h(t) represents the holiday term, which indicates the impact of non-fixed-period holidays or major events in the time series on the curve, such as maintenance and repair of substations, typhoon weather and other events. Since different events have different impacts on the curve, the holiday model can treat the impact of each event at different time points as multiple independent models, and can set the range of the time period before and after the event for different events. The calculation formula is:

[0071]

[0072] Where: κ~Normal(0,v 2 ) obeys the normal distribution, κ i represents the impact of the event on the model, i represents holidays or other events, and D i represents the variable of the ith independent model. If the time of event i belongs to this variable, then D i The value is 1 if yes, otherwise it is 0.

[0073] ε t Represents the residual term and follows a Gaussian distribution, indicating the fluctuations to which the model is fitted.

[0074] In addition, Neural Prophet also introduced the AR-Net model. The first layer of the neural network is a p-order autoregressive structure. In order to make the curve fitting more accurate, a certain number of neurons are added to the second layer, and the mean square error (MSE) is defined as the loss function. The basic formula of the p-order autoregressive model is:

[0075]

[0076] Where: y t-1 ,…,y t-p represents the p-order lag term, w i is the p-order weight or autoregressive coefficient, e t represents noise, and c represents a constant term.

[0077] The MSE calculation formula is:

[0078]

[0079] In order to introduce sparsity to the autoregressive coefficients, the regular term R is added to the loss function, and the loss function calculation formula is changed to:

[0080]

[0081] Where: s is the sparsity of the autoregressive coefficient, c λ is the degree of regularization, which depends on the noise of the data, c1, c2 are regularization parameters, which depend on the range of the autoregressive coefficient. When normalized, the autoregressive coefficient range is [0, 1], then c1≈c2≈3.

[0082] Step 10: Use Six Sigma (similar to the 3sigema function, the internal parameter 3 is set to 6 to construct a larger confidence interval to avoid false positives) to construct a confidence interval. By constructing confidence intervals for the maximum channel current prediction value and the fault channel current prediction value, it is inferred that the time when the mid-range channel current ratio is less than 1.9 times.

[0083] The present invention uses a machine learning algorithm to mine the spatiotemporal characteristics of the channel current data of the intermediate-range detector of the ex-core nuclear measurement system, and provides fault warning for the intermediate-range detector of the ex-core nuclear measurement system. Due to the problems of non-fixed sampling frequency, large data redundancy, data fluctuation, data drift, etc. of the intermediate-range channel current data of the ex-core nuclear measurement system, it is difficult to distinguish between fault data and normal data. The present invention proposes a cleaning process for the intermediate-range channel current data to improve data quality.

[0084] In fault warning, determining the fault start time can take timely measures to prevent the fault from further expanding, and can better analyze the cause and mechanism of the fault. However, if the fault start time is identified incorrectly, it will lead to misjudgment of the normal operating state as an abnormal state, thereby generating false alarms and wasting resources. Therefore, the present invention proposes an accurate and fast method for identifying the fault time of the intermediate range channel current of the off-core nuclear measurement system. On the basis of the identification of the intermediate range channel current fault time, the present invention uses the Neuralprophet model to predict the fault current, infer the time when the intermediate range channel current ratio of the off-core nuclear measurement system is less than 1.9 times, and perform fault warning on the intermediate range detector.

[0085] Aiming at the problems of non-fixed sampling frequency, large data redundancy, data fluctuation, data drift, etc. of intermediate-range channel current data, the present invention proposes a data cleaning process, which can effectively improve the quality of intermediate-range channel current data.

[0086] The present invention is based on mutual information fault time recognition, and uses the spatiotemporal characteristics of the intermediate range channel current data to accurately and quickly identify the fault time. On this basis, the NeuralProphet model is used to predict the channel current, construct the confidence interval, and infer the time when the intermediate range channel current ratio is less than 1.9 times, and the intermediate range detector fault warning is performed.

[0087] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for early warning of faults of an intermediate range detector of an out-of-core nuclear measurement system, characterized in that: The following steps are involved: Step 1: Obtain channel current data through the intermediate range detector of the out-of-core nuclear measurement system; Step 2: Use linear interpolation to fill the data to ensure that the data sampling frequency is consistent; Step 3: Data downsampling: downsampling the original training data and test data; Step 4: Identify outliers in the current data of the mid-range channel and reset the outliers to null values; Step 5: Use linear interpolation to interpolate the null values ​​to ensure that the processed curve is highly similar to the original curve characteristics; Step 6: Perform sliding window processing on the data after outlier processing; Step 7: Calculate the mutual information of each channel current in the middle range. The calculation formula is as follows: Where I(X; Y) represents the mutual information of variables X and Y. When X and Y are completely independent, I(X; Y) is the smallest zero. p(x), p(y), and p(x, y) represent the marginal probability distribution function of variable X, the marginal probability distribution function of Y, and the joint probability distribution function of X and Y, respectively. Step 8: Identify the fault time by judging the change in mutual information; Step 9: Using the NerualProphet model, predict the current signal of the intermediate range detector from the fault initiation time; Step 10: Construct a confidence interval and conduct fault warning.

2. The method for early warning of fault of intermediate range detector of an out-of-core nuclear measurement system according to claim 1 is characterized in that: In step 3, the data is downsampled using a sliding window.

3. The method for early warning of fault of intermediate range detector of an out-of-core nuclear measurement system according to claim 2, characterized in that: A sliding window with a length of 3600 and a step size of 3600 is used to sample the original training data and test data.

4. The method for early warning of fault of intermediate range detector of an out-of-core nuclear measurement system according to claim 1, characterized in that: In step 4, outliers are identified using the constant fluctuation detector differencing algorithm.

5. The method for early warning of fault of intermediate range detector of an out-of-core nuclear measurement system according to claim 1, characterized in that: In step 6, the left pointer of the window is always fixed at the leftmost position of the array, and the right pointer of the window is shifted 24 bits to the right after each sliding window operation to obtain the current time characteristics of the intermediate range channel.

6. The method for early warning of fault of intermediate range detector of an out-of-core nuclear measurement system according to claim 1, characterized in that: In step 7, the formula Calculate information entropy, In the formula, H(X) is the information entropy value, p(x) is the probability of taking the value x, and b is usually taken as the base of the natural constant e.

7. The method for early warning of fault of intermediate range detector of an out-of-core nuclear measurement system according to claim 1, characterized in that: In step 9, the Neural Prophet model is composed of the p-order autoregressive model of the Prophet model and the AR-Net model. The Prophet model is: P(t)=g(t)+s(t)+h(t)+ε t , In the formula, g(t) represents the trend term; C(t) represents the carrying capacity, which is a function that changes with time and limits the maximum value of the curve fitting; s(t) represents the seasonal term; h(t) represents the holiday term; κ~Normal(0,v 2 ) obeys the normal distribution, κ i represents the impact of the event on the model, i represents holidays or other events, and D i represents the variable of the ith independent model; ε t represents the residual term and follows a Gaussian distribution, indicating the fluctuations that the model is subject to when fitting; The p-order autoregressive model of the AR-Net model is: In the formula, y t-1 ,…,y t-p represents the p-order lag term, w i is the p-order weight or autoregressive coefficient, e t represents noise, c represents a constant term; The MSE calculation formula is: In order to introduce sparsity to the autoregressive coefficients, the regular term R is added to the loss function, and the loss function calculation formula is: In the formula, s represents the sparsity of the autoregressive coefficient, c λ represents the degree of regularization, c1 and c2 represent the regularization parameters.

8. The method for early warning of fault of intermediate range detector of an out-of-core nuclear measurement system according to claim 1, characterized in that: In step 10, six sigma is used to construct a confidence interval. By constructing a confidence interval for the maximum channel current prediction value and the fault channel current prediction value, the time when the mid-range channel current ratio is less than 1.9 times is inferred.

Images