An intelligent fault detection method for electrical equipment

Abstract

The present invention relates to the technical field of measuring electrical variables, and proposes a fault intelligent detection method for electrical equipment, including: obtaining a three-dimensional imaging device fault monitoring matrix; according to the division results of each row of data obtained from the three-dimensional imaging device fault monitoring matrix, calculating a synchronous elevation fitting factor, a synchronous decrease fitting factor, and a synchronous trend fitting correlation coefficient according to the division results of each row of data; obtaining a trend fitting indirect correlation coefficient and a trend fitting comprehensive correlation coefficient according to the synchronous trend fitting correlation coefficient; obtaining a comprehensive weight correction index according to the trend fitting comprehensive correlation coefficient; using a locally weighted regression algorithm based on the comprehensive weight correction index to obtain a fitting curve for each row of data in the three-dimensional imaging device fault detection matrix, and obtaining a device fault detection result based on the fitting curve. The present invention obtains a fitting curve of device monitoring data through a comprehensive weight correction index, improving the accuracy of device fault detection.

Description

Technical Field

[0001] The present invention relates to the technical field of measuring electrical variables, and particularly to an intelligent fault detection method for electrical equipment. Background Art

[0002] An infrared three-dimensional precise imaging device belongs to electrical equipment. It can use a binocular system and motion compensation technology to obtain an accurate three-dimensional point cloud model of the detection target, achieve accurate measurement and correction of the surface temperature of the detection target, solve the problem of inaccurate temperature measurement at different distances, angles and in different environments at present, and can also perform inversion calculation on the internal temperature field of the detection target, so as to diagnose and analyze the internal heating situation of the detection target. Among them, the electrical parameter data of the infrared three-dimensional precise imaging device can reflect whether its operating state has failed. However, the working environment of the infrared three-dimensional precise imaging device is usually relatively complex, and the device is small and has high precision requirements. The core components are easily affected by the external environment, resulting in large errors in the obtained electrical parameter data, which affects the accuracy of fault detection of the infrared three-dimensional precise imaging device.

[0003] The locally weighted regression Lowess (Locally Weighted Scatterplot Smoothing) algorithm can exclude the influence of other factors according to the non-linear relationship between data sequences, reflect the overall trend of electrical parameter data, obtain accurate electrical parameter fitting values, and improve the accuracy of fault detection of the infrared three-dimensional precise imaging device. However, the Lowess algorithm needs to preset a weight value for the input data. When the preset weight value is inappropriate, it may affect the lower fitting accuracy of the Lowess algorithm and reduce the accuracy of fault detection of the infrared three-dimensional precise imaging device. Summary of the Invention

[0004] The present invention provides an intelligent fault detection method for electrical equipment to solve the problem of low fault detection accuracy of the infrared three-dimensional precise imaging device. The specific technical solutions adopted are as follows:

[0005] An embodiment of the present invention provides an intelligent fault detection method for electrical equipment, which includes the following steps:

[0006] Obtain a fault monitoring matrix of the three-dimensional imaging device;

[0007] Obtain the division result of each row of data according to the change trend characteristics of each row of data in the fault monitoring matrix of the three-dimensional imaging device; obtain the synchronous trend fitting correlation coefficient between different rows of data in the fault monitoring matrix of the three-dimensional imaging device according to the division result of each row of data in the fault monitoring matrix of the three-dimensional imaging device;

[0008] Construct a trend - associated undirected graph of the 3D imaging device fault monitoring matrix according to the synchronization trend fitting correlation coefficient between different - row data of the 3D imaging device fault monitoring matrix; obtain the trend - fitting comprehensive correlation coefficient between non - electrical parameter data and electrical parameter data in the 3D imaging device fault monitoring matrix according to the trend - associated undirected graph of the 3D imaging device fault monitoring matrix; obtain the comprehensive weight correction index of each element in the electrical parameter data sequence in the 3D imaging device fault detection matrix according to the trend - fitting comprehensive correlation coefficient between non - electrical parameter data and electrical parameter data in the 3D imaging device fault monitoring matrix.

[0009] Obtain the fitting curve of each row of data in the 3D imaging device fault monitoring matrix according to the comprehensive weight correction index of each element in the 3D imaging device fault monitoring matrix, and obtain the fault monitoring result of the 3D imaging device based on the fitting curve using the CNN neural network model.

[0010] Preferably, the method for obtaining the division result of each row of data according to the change - trend characteristics of each row of data in the 3D imaging device fault monitoring matrix is as follows:

[0011] For each row of data in the 3D imaging device fault monitoring matrix, take the sequence composed of the data in each row as the monitoring data sequence, take the position serial number of each data in the monitoring data sequence as the abscissa, take the magnitude of each data value in the monitoring data sequence as the ordinate, and take the curve formed by the abscissa and the ordinate as the monitoring data curve of the monitoring data sequence. Obtain all the maximum - value points and minimum - value points of the monitoring data curve, take all the maximum - value points and minimum - value points as the maximum segmentation points and minimum segmentation points of the monitoring data sequence respectively, take the sequence composed of all the maximum segmentation points and minimum segmentation points in ascending order of time as the segmentation sequence, and according to the segmentation sequence, take the sequence composed of the corresponding data in the monitoring data sequence between adjacent maximum segmentation points and minimum segmentation points from left to right as the decreasing - trend sequence of the monitoring data sequence, and take the sequence composed of the corresponding data in the monitoring data sequence between adjacent minimum segmentation points and maximum segmentation points as the increasing - trend sequence of the monitoring data sequence.

[0012] Preferably, the method for obtaining the synchronization trend fitting correlation coefficient between different - row data of the 3D imaging device fault monitoring matrix according to the division result of each row of data in the 3D imaging device fault monitoring matrix is as follows:

[0013] For each row of data in the 3D imaging device fault monitoring matrix, take the sequence composed of the data in each row as the monitoring data sequence, use the Bayesian curve - fitting algorithm to obtain the Bayesian fitting curve corresponding to the monitoring data sequence, and obtain the slope corresponding to each element in the monitoring data sequence according to the Bayesian fitting curve.

[0014] Obtain the synchronous increase fitting factor between different monitoring data sequences according to the difference between the growth trend sequences of different monitoring data sequences in the 3D imaging device fault monitoring matrix;

[0015] Obtain the synchronous decrease fitting factor between different monitoring data sequences according to the difference between the decrease trend sequences of different monitoring data sequences in the 3D imaging device fault monitoring matrix;

[0016] Take the sum of the synchronous increase fitting factor and the synchronous decrease fitting factor between different monitoring data sequences as the first correlation coefficient, take the SBD distance between different monitoring data sequences as the second correlation coefficient, and take the product of the first correlation coefficient and the second correlation coefficient as the synchronous trend fitting correlation coefficient between different monitoring data sequences.

[0017] Preferably, the calculation method for obtaining the synchronous increase fitting factor between different monitoring data sequences according to the difference between the growth trend sequences of different monitoring data sequences in the 3D imaging device fault monitoring matrix is:

[0018]

[0019] In the formula, represents the synchronous increase fitting factor between the th monitoring data sequence and the th monitoring data sequence; and respectively represent the th and the th collection moments corresponding to the th data in the th growth trend sequence in the th and the th monitoring data sequences; and respectively represent the slopes corresponding to the th data in the th growth trend sequence in the th and the th monitoring data sequences; represents the minimum value of the number of data in the th growth trend sequence in the th monitoring data sequence and the number of data in the th growth trend sequence in the th monitoring data sequence; represents the minimum value of the number of growth trend sequences in the th monitoring data sequence and the number of growth trend sequences in the th monitoring data sequence;

[0020] Preferably, the calculation method for obtaining the synchronous reduction fitting factor between different monitoring data sequences according to the difference in the decreasing trend sequences between different monitoring data sequences in the three-dimensional imaging device fault monitoring matrix is as follows:

[0021]

[0022] In the formula, represents the th monitoring data sequence and the th monitoring data sequence; and respectively represent the th and the th monitoring data sequences, the th decreasing trend sequence in them, and the th data collection time; and respectively represent the th and the th monitoring data sequences, the th decreasing trend sequence in them, and the th data corresponding slope; represents the th monitoring data sequence, the th decreasing trend sequence in it, the minimum value of the number of data in the th monitoring data sequence and the th decreasing trend sequence in the th monitoring data sequence; represents the th monitoring data sequence, the minimum value of the number of decreasing trend sequences in the th monitoring data sequence and the th monitoring data sequence;

[0023] Preferably, the method for constructing the trend correlation undirected graph of the three-dimensional imaging device fault monitoring matrix according to the synchronous trend fitting correlation coefficient between different rows of data in the three-dimensional imaging device fault monitoring matrix is as follows:

[0024] For each row of data in the three-dimensional imaging device fault monitoring matrix, take the sequence composed of the data in each row as the monitoring data sequence, and take the sequence formed by sorting the synchronous trend fitting correlation coefficients between the monitoring data sequence and all other monitoring data sequences in ascending order as the synchronous trend fitting correlation sequence of the monitoring data sequence. Construct a binary tree of the monitoring data sequence according to the synchronous trend fitting correlation sequence;

[0025] Take each monitoring data sequence as a node in an undirected graph, take the tree edit distance between the corresponding binary trees of two monitoring data sequences as the initial weight between the corresponding two nodes, and take the undirected graph based on all monitoring data sequences, the initial weights of different monitoring data sequences, and the constructed undirected graph as the trend correlation undirected graph of the three-dimensional imaging device fault monitoring matrix.

[0026] Preferably, the method for obtaining the trend fitting comprehensive correlation coefficient between the non-electrical parameter data and the electrical parameter data in the three-dimensional imaging device fault monitoring matrix according to the trend correlation undirected graph of the three-dimensional imaging device fault monitoring matrix is as follows:

[0027] Take each row of data corresponding to the non-electrical parameter data in the three-dimensional imaging device fault monitoring matrix as a non-electrical parameter data sequence of the three-dimensional imaging device fault monitoring matrix, and take each row of data corresponding to the electrical parameter data in the three-dimensional imaging device fault monitoring matrix as an electrical parameter data sequence of the three-dimensional imaging device fault monitoring matrix;

[0028] Calculate the trend fitting indirect correlation coefficient according to the corresponding relationship between the non-electrical parameter data sequence and the electrical parameter data sequence in the trend correlation undirected graph of the three-dimensional imaging device fault monitoring matrix;

[0029] Take the sum of the synchronous trend fitting correlation coefficient and the trend fitting indirect correlation coefficient between each non-electrical parameter data sequence and each electrical parameter data sequence in the three-dimensional imaging device fault monitoring matrix as the first characteristic coefficient, take the sum of all the first characteristic coefficients as the second characteristic coefficient, and take the ratio of the first characteristic coefficient to the second characteristic coefficient as the trend fitting comprehensive correlation coefficient between each non-electrical parameter data sequence and each electrical parameter data sequence.

[0030] Preferably, the method for calculating the trend fitting indirect correlation coefficient according to the corresponding relationship between the non-electrical parameter data sequence and the electrical parameter data sequence in the trend correlation undirected graph of the three-dimensional imaging device fault monitoring matrix is as follows:

[0031]

[0032] In the formula, represents the th non-electrical parameter data sequence and the th electrical parameter data sequence the trend fitting indirect correlation coefficient between; represents the th non-electrical parameter data sequence and the th electrical parameter data sequence in the trend correlation undirected graph the th path Indicates the edge true weight between the th non - electrical parameter data sequence and the th electrical parameter data sequence in the th path, between the th and the th nodes; Indicates the serial number corresponding to the th non - electrical parameter data sequence and the th electrical parameter data sequence in the th path, for the th node; Indicates the number of nodes in the th non - electrical parameter data sequence and the th electrical parameter data sequence in the th path; Indicates the number of paths between the th non - electrical parameter data sequence and the th electrical parameter data sequence in the trend - associated undirected graph.

[0033] Preferably, the method for obtaining the comprehensive weight correction index of each element in the electrical parameter data sequence in the 3D imaging device fault detection matrix according to the trend - fitting comprehensive correlation coefficient between the non - electrical parameter data and the electrical parameter data in the 3D imaging device fault monitoring matrix is as follows:

[0034] For each electrical parameter data sequence in the 3D imaging device fault monitoring matrix, take the trend - fitting comprehensive correlation coefficient between each electrical parameter data sequence and each other non - electrical parameter data sequence as the fusion weight of each non - electrical parameter data sequence, perform weighted fusion on all the non - electrical parameter data sequences according to the corresponding fusion weights, take the result of the weighted fusion as the non - electrical parameter fusion sequence of each electrical parameter data sequence, and obtain the predicted value of each element in each electrical parameter data sequence based on the non - electrical parameter fusion sequence of each electrical parameter data sequence using the ARIMAX model;

[0035] Obtain the comprehensive weight correction index of each element according to the true value and the predicted value corresponding to each element in the electrical parameter data sequence.

[0036] Preferably, the method for obtaining the comprehensive weight correction index of each element according to the true value and the predicted value corresponding to each element in the electrical parameter data sequence is as follows:

[0037]

[0038] In the formula, Indicates the The comprehensive weight correction index of the th element in the th electrical parameter data sequence; th represents the difference between the true value and the predicted value corresponding to the th element in the th electrical parameter data sequence; th represents the maximum value of the differences between the true values and the predicted values corresponding to all elements in the th electrical parameter data sequence; is the natural exponential function.

[0039] The beneficial effects of the present invention are as follows: By obtaining the synchronous trend fitting correlation coefficient between different types of monitoring data through the synchronous change characteristics between different types of monitoring data in the three-dimensional imaging device fault monitoring matrix, constructing a trend correlation undirected graph of the three-dimensional imaging device fault monitoring matrix according to the synchronous fitting correlation coefficient between different types of monitoring data, obtaining the trend fitting comprehensive correlation coefficient between the non-electrical parameter data sequence and the electrical parameter data sequence in the three-dimensional imaging device fault monitoring matrix according to the trend correlation undirected graph, adopting data fusion and the ARIMAX model to obtain the comprehensive weight correction index based on the trend fitting comprehensive correlation coefficient, and adjusting the weight value of each data in the Lowess algorithm based on the comprehensive weight correction index. The beneficial effect is to avoid the low accuracy of curve fitting using the Lowess algorithm due to unreasonable weight setting, improve the accuracy of processing electrical parameter data of the infrared three-dimensional precise imaging device, and further improve the accuracy of fault monitoring of the infrared three-dimensional precise imaging device. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for description in the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0041] Figure 1 is a schematic flow chart of a fault intelligent detection method for electrical equipment provided by an embodiment of the present invention;

[0042] Figure 2 is a schematic diagram of the process for obtaining a trend correlation undirected graph provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0044] Please refer to Figure 1 , which shows a flowchart of a method for intelligent fault detection of electrical equipment provided by an embodiment of the present invention. The method includes the following steps:

[0045] Step S001, obtain a three-dimensional imaging device fault monitoring matrix.

[0046] During the operation of the infrared three-dimensional precise imaging device, a voltmeter and an ammeter are placed on the device circuit, vibration sensors, temperature sensors, and humidity sensors are arranged in the internal voids of the infrared three-dimensional precise imaging device structure, and sound sensors, outdoor air detectors, and anemometers are arranged on the surface of the infrared three-dimensional precise imaging device to collect data on the current, voltage, vibration, temperature, humidity, audio, dust concentration, wind speed, etc. of the infrared three-dimensional precise imaging device. The time interval for data collection is (the size takes the empirical value 1), and the length of the collected data sequence is (the size takes the empirical value 500). At the same time, since data loss may occur during the data collection process and different data types have different units, the above data is preprocessed. Specifically, the linear interpolation method is used to fill in the missing values, and at the same time, the Z-Score standardization is performed on the filled data sequence. The specific calculation processes of the linear interpolation method and the Z-Score standardization are both well-known technologies and will not be elaborated here.

[0047] Obtain all the monitored data sequences of the preprocessed infrared three-dimensional precise imaging device, and denote the matrix composed of all the monitored data sequences as the three-dimensional imaging device fault monitoring matrix , where is the electrical parameter data sequence, corresponding to the current and voltage data sequences respectively, is the non-electrical parameter data sequence, which are the vibration, temperature, humidity, audio, dust concentration, and wind speed data sequences respectively, represents the transpose of the matrix.

[0048] So far, the three-dimensional imaging device fault monitoring matrix has been obtained.

[0049] Step S002: Obtain the division result of each row of data according to the change trend characteristics of each row of data in the three-dimensional imaging matrix, and calculate the synchronous increase fitting factor, synchronous decrease fitting factor, and synchronous trend fitting correlation coefficient according to the division result of each row of data.

[0050] Infrared three-dimensional precise imaging devices are usually non-active heat dissipation devices. For example, they perform infrared three-dimensional imaging on devices such as cable joints and bushings to measure the accurate surface temperature of the devices. However, the working environment of infrared three-dimensional precise imaging devices is usually relatively complex, and the size of infrared three-dimensional precise imaging devices is small, but the measurement precision of internal parts is relatively high. Therefore, the acquired current and voltage data are easily affected by other factors (such as temperature, wind speed, etc.), resulting in low accuracy of the acquired electrical parameter data and low fault detection accuracy of infrared three-dimensional precise imaging devices.

[0051] Therefore, the purpose of the present invention is to obtain the change characteristics of each electrical parameter data point according to the correlation characteristics between non-electrical parameter data and electrical parameter data, use the locally weighted regression Lowess algorithm to fit the electrical parameter data sequence, reflect the overall trend of the electrical parameter data sequence, and correct the original electrical parameter data. Finally, based on the electrical parameter data curve of the infrared three-dimensional precise imaging device after correction processing, a high-precision device fault detection model is established using a convolutional neural network model to realize the intelligent detection of faults in infrared three-dimensional precise imaging devices.

[0052] Specifically, in an infrared three-dimensional precise imaging device, there may be a high synchronous correlation between two data sequences. For example, when the temperature of the detection target increases, the detection target may emit more infrared radiation. As the temperature of the detection target increases, the temperature of internal components of the infrared three-dimensional imaging device, such as optical elements, increases, resulting in an increase in the acquired temperature data. At the same time, since the infrared three-dimensional precise imaging device can convert the received infrared radiation into a visible image, the detector in the imaging device can receive more energy, resulting in an increase in the current value of the infrared three-dimensional precise imaging device.

[0053] Further, divide the data according to the change characteristics of each row of data in the three-dimensional imaging device fault monitoring matrix. Specifically, for example, the first row of data in the three-dimensional imaging device fault monitoring matrix corresponds to the monitoring data sequence as the current data sequence , indicating the th data in the current data sequence. Take the position serial number of each data in as the abscissa, take the magnitude of each data value in as the ordinate, and take the curve formed by the abscissa and the ordinate as 's monitoring data curve, obtain all the maximum points and minimum points of the monitoring data curve, and take all the maximum points and minimum points as The maximum and minimum split points of The sequence of data between the maximum split point and the minimum split point is The decreasing trend sequence will The sequence of data between the minimum split point and the maximum split point is growth trend sequence.

[0054] Furthermore, the input is a current data sequence , using the Bayesian curve fitting algorithm to obtain The corresponding Bayesian fitting curve is obtained according to the Bayesian fitting curve The slope corresponding to each element in .

[0055] At this point, the division result corresponding to each row of data in the three-dimensional imaging device fault monitoring matrix and the slope corresponding to each element in each row of data can be obtained.

[0056] Furthermore, the synchronization reduction matching factor between different monitoring data sequences in the three-dimensional imaging device fault monitoring matrix is ​​obtained according to the difference in the reduction trend sequence between the different monitoring data sequences. The specific calculation formula is as follows:

[0057]

[0058] In the formula, Indicates monitoring data series and The synchronous increase matching factor between the monitoring data series; and Respectively represent and The monitoring data series The first in the growth trend sequence The collection time corresponding to each data; and Respectively represent and The monitoring data series The first in the growth trend sequence The data corresponds to the slope; Indicates The monitoring data series The number of data in the growth trend series is The monitoring data series The minimum value of the number of data in a growth trend series; Indicates The number of growth trend sequences in the monitoring data series is The minimum value among the numbers of increasing trend sequences in a monitoring data sequence; Denotes a regulation parameter, with an empirical value of 0.001; Is the natural constant.

[0059] If the growth change trends among different monitoring data sequences in the 3D imaging device fault monitoring matrix are close, then the The smaller the value of, The larger the value of, that is, the larger the value of the The larger the value of, it indicates that the growth trend change between the th monitoring data sequence and the th monitoring data sequence in the 3D imaging device fault monitoring matrix is close.

[0060] Furthermore, the synchronous decrease fitting factor between the different monitoring data sequences is obtained according to the differences in the decreasing trend sequences among the different monitoring data sequences in the 3D imaging device fault monitoring matrix. The specific calculation formula is as follows:

[0061]

[0062] In the formula, Denotes the synchronous decrease fitting factor between the th monitoring data sequence and the th monitoring data sequence; And Respectively denote the th and the th acquisition moments corresponding to the th data in the th decreasing trend sequence in the th and the th monitoring data sequences; th and the th monitoring data sequences; th decreasing trend sequence in the th data corresponding slopes; Denotes the th monitoring data sequence in the th decreasing trend sequence in the number of data and the th monitoring data sequence in the th minimum value among the numbers of data in the decreasing trend sequence; Denotes the th monitoring data sequence in the number of decreasing trend sequences and the th minimum value among the numbers of decreasing trend sequences in the monitoring data sequence; Denotes a regulation parameter, with an empirical value of 0.001; Is the natural constant.

[0063] If the decreasing change trends among different monitoring data sequences in the 3D imaging device fault monitoring matrix are close, the smaller the calculated value, and the larger the value. That is, the larger the calculated value indicates that the decreasing trend changes between the th monitoring data sequence and the th monitoring data sequence in the 3D imaging device fault monitoring matrix are close.

[0064] Furthermore, according to the analysis results of the increasing change trends and decreasing change trend characteristics among different monitoring data sequences in the 3D imaging device fault monitoring matrix, calculate the synchronous trend fitting correlation coefficient between different monitoring data sequences. Specifically, take the sum of the synchronous increasing fitting factor and the synchronous decreasing fitting factor between the different monitoring data sequences as the first correlation coefficient, take the SBD distance between the different monitoring data sequences as the second correlation coefficient, and take the product of the first correlation coefficient and the second correlation coefficient as the synchronous trend fitting correlation coefficient between the different monitoring data sequences; if the increasing change trends and decreasing change trend characteristics among different monitoring data sequences are close, the larger the value of the calculated synchronous trend fitting correlation coefficient between different monitoring data sequences.

[0065] So far, the synchronous trend fitting correlation coefficient between different monitoring data sequences in the 3D imaging device fault monitoring matrix has been obtained.

[0066] Step S003, obtain the trend fitting indirect correlation coefficient according to the synchronous trend fitting correlation coefficient, calculate the trend fitting comprehensive correlation coefficient according to the synchronous trend fitting correlation coefficient and the trend fitting indirect correlation coefficient, and obtain the comprehensive weight correction index according to the trend fitting comprehensive correlation coefficient.

[0067] Construct a trend correlation undirected graph of the 3D imaging device fault monitoring matrix based on the synchronization trend fitting correlation coefficient between different monitoring data sequences in the 3D imaging device fault monitoring matrix, and reflect the indirect correspondence relationship between different monitoring data sequences in the 3D imaging device fault monitoring matrix through the trend correlation undirected graph of the 3D imaging device fault monitoring matrix. Specifically, the sequence formed by sorting the synchronization trend fitting correlation coefficients between each monitoring data sequence in the 3D imaging device fault monitoring matrix and all other monitoring data sequences in ascending order is used as the synchronization trend fitting correlation sequence of each monitoring data sequence. Each monitoring data sequence is used as the root node, and all other monitoring data sequences are used as child nodes. According to the arrangement order of the elements in the synchronization trend fitting correlation sequence, the monitoring data sequence corresponding to each element in the synchronization trend fitting correlation sequence is used as a child node, and the binary tree constructed by the root node and all child nodes is used as the binary tree of each monitoring data sequence.

[0068] Further, each monitoring data sequence in the 3D imaging device fault monitoring matrix is used as a node in the initial undirected graph, and the tree edit distance between the binary trees corresponding to different monitoring data sequences in the 3D imaging device fault monitoring matrix is used as the initial weight between different monitoring data sequences. An initial undirected graph of the 3D imaging data matrix is constructed according to the node and the initial weight; the sequence formed by sorting all the initial weights in ascending order is used as the initial weight sequence of the initial undirected graph. The input is the initial weight sequence, and the Otsu threshold segmentation algorithm is used to obtain the segmentation threshold of the initial weight sequence. The elements in the initial weight sequence that are less than or equal to the segmentation threshold are used as the edge true weights in the initial undirected graph. The edge true weights in the initial undirected graph are retained, and the updated result of the initial undirected graph is used as the trend correlation undirected graph of the 3D imaging device fault monitoring matrix. The process of obtaining the trend correlation undirected graph from the initial undirected graph is as Figure 2 shown.

[0069] Further, calculate the trend fitting indirect correlation coefficient according to the correspondence relationship between the non-electrical parameter data sequence and the electrical parameter data sequence in the trend correlation undirected graph of the 3D imaging device fault monitoring matrix. The specific calculation formula is as follows:

[0070]

[0071] In the formula, represents the th non-electrical parameter data sequence and the th electrical parameter data sequence represents the th non-electrical parameter data sequence and the th electrical parameter data sequence in the trend correlation undirected graph. The synchronization trend fitting correlation coefficient between the monitoring data sequences corresponding to the rd and the th nodes in a path; Indicates the th non - electrical parameter data sequence and the th electrical parameter data sequence in the trend - associated undirected graph, and the th real weight of the edge between the th and the th nodes in the th path; Indicates the th non - electrical parameter data sequence and the th electrical parameter data sequence in the trend - associated undirected graph, and the th serial number corresponding to the th node in the th path; Indicates the number of nodes in the th path between the th non - electrical parameter data sequence and the th electrical parameter data sequence in the trend - associated undirected graph; Indicates the number of paths between the

[0072] If the correlation between the th non - electrical parameter data sequence and the th electrical parameter data sequence is relatively high, then the calculated value is smaller, and the value is larger, that is, the calculated trend - fitting indirect correlation coefficient between the th non - electrical parameter data sequence and the th electrical parameter data sequence is larger.

[0073] Furthermore, the sum of the synchronization trend fitting correlation coefficient and the trend - fitting indirect correlation coefficient between each non - electrical parameter data sequence and each electrical parameter data sequence in the three - dimensional imaging device fault monitoring matrix is used as the first characteristic coefficient, the sum of all the first characteristic coefficients is used as the second characteristic coefficient, and the ratio of the first characteristic coefficient to the second characteristic coefficient is used as the trend - fitting comprehensive correlation coefficient between each non - electrical parameter data sequence and each electrical parameter data sequence. If the degree of association between the non - electrical parameter data sequence and the electrical parameter data sequence in the three - dimensional imaging device fault monitoring matrix is relatively large, then the calculated value of the first characteristic coefficient is larger, that is, the calculated trend - fitting comprehensive correlation coefficient between the non - electrical parameter data sequence and the electrical parameter data sequence is larger.

[0074] Further, for each electrical parameter data sequence in the three-dimensional imaging device fault monitoring matrix, the trend fitting comprehensive correlation coefficient between each electrical parameter data sequence and each other non-electrical parameter data sequence is used as the fusion weight of each non-electrical parameter data sequence. All the non-electrical parameter data sequences are weighted and fused according to the corresponding fusion weights, and the result of the weighted fusion is used as the non-electrical parameter fusion sequence of each electrical parameter data sequence. Based on the non-electrical parameter fusion sequence of each electrical parameter data sequence, the prediction value of each element in each electrical parameter data sequence is obtained by using the ARIMAX model. The specific training process of the ARIMAX model is a well-known technology and will not be elaborated here.

[0075] Specifically, for example, the monitoring data sequence corresponding to the first row of data in the three-dimensional imaging device fault monitoring matrix is a current data sequence , and the current data sequence is an electrical parameter data sequence of the three-dimensional imaging device fault monitoring matrix. The trend fitting comprehensive correlation coefficients between and the non-parameter data sequences in the three-dimensional imaging device fault monitoring matrix are respectively used as the fusion weights of . Weighted fusion is performed according to and its corresponding fusion weight, and the calculation result of the weighted fusion is used as the non-electrical parameter fusion sequence of . Obtain the sequence composed of the current data corresponding to the first (the size takes the empirical value of 50) acquisition times of each data in in ascending order of time as the prediction data sequence of each data in . The sequence composed of all elements corresponding to the acquisition times of all elements in the prediction data sequence in the non-electrical parameter fusion sequence in ascending order of time is used as the prediction comparison sequence of each data in

[0076] Further, the input is the prediction comparison sequence and the prediction data sequence of each data in . Among them, the prediction comparison sequence is used as the exogenous variable sequence of the ARIMAX model, and the ARIMAX model is used to obtain

[0077] the prediction value of each data in

[0078]

[0079] In the formula, represents the th electrical parameter data sequence and the Comprehensive weight correction index of an element; Indicates the Difference between the true value and the predicted value corresponding to the th element in the Indicates the Maximum value of the differences between the true values and the predicted values corresponding to all elements in the Indicates the Original weight in the locally weighted regression algorithm corresponding to the th element in the Is the natural exponential function.

[0080] If the difference between the true value and the predicted value corresponding to the th element in the th electrical parameter data sequence in the three-dimensional imaging device fault monitoring matrix is smaller, then the calculated value is larger, that is, the calculated comprehensive weight correction index of the th element in the th electrical parameter data sequence is larger.

[0081] So far, the comprehensive weight correction index of each element in the electrical parameter data sequence in the three-dimensional imaging device fault monitoring matrix has been obtained.

[0082] Step S004, based on the comprehensive weight correction index, use the locally weighted regression algorithm to obtain the fitting curve of each row of data in the three-dimensional imaging device fault detection matrix, and obtain the fault detection result of the three-dimensional imaging device based on the fitting curve of each row of data.

[0083] The electrical parameter data sequences in the three-dimensional imaging device fault monitoring matrix are respectively the current data sequence and the voltage data sequence , and the inputs are respectively and The comprehensive weight correction index corresponding to the elements in, and The comprehensive weight correction index corresponding to the elements in, where the comprehensive weight correction index is used as the weight of the input data, and the Lowess algorithm is used to obtain the fitting curves of and respectively, and the fitting curves of and are used to adjust the data at each acquisition moment in and . The specific calculation process of the Lowess algorithm is a well-known technology and will not be elaborated here.

[0084] Further, taking the adjusted and as inputs, a convolutional neural network model is used to obtain the fault detection results of the infrared three-dimensional precise imaging device. The fault detection results include but are not limited to current anomalies, voltage anomalies, temperature anomalies, etc. The optimization algorithm is the Adam algorithm, and the loss function is the cross-entropy loss function. The specific training process of the convolutional neural network is a well-known technology and will not be elaborated here.

[0085] Thus far, the fault detection results of the infrared three-dimensional precise imaging device have been obtained.

[0086] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for intelligent fault detection of electrical equipment, characterized in that: The method comprises the following steps: Obtaining a three-dimensional imaging equipment fault monitoring matrix; According to the change trend characteristics of each row of data in the three-dimensional imaging device fault monitoring matrix, the division result of each row of data is obtained; according to the division result of each row of data in the three-dimensional imaging device fault monitoring matrix, the synchronization trend matching correlation coefficient between different rows of data in the three-dimensional imaging device fault monitoring matrix is ​​obtained; According to the synchronous trend matching correlation coefficient between different rows of data in the three-dimensional imaging equipment fault monitoring matrix, a trend correlation undirected graph of the three-dimensional imaging equipment fault monitoring matrix is ​​constructed; according to the trend correlation undirected graph of the three-dimensional imaging equipment fault monitoring matrix, a trend matching comprehensive correlation coefficient between non-electrical parameter data and electrical parameter data in the three-dimensional imaging equipment fault monitoring matrix is ​​obtained; according to the trend matching comprehensive correlation coefficient between non-electrical parameter data and electrical parameter data in the three-dimensional imaging equipment fault monitoring matrix, a comprehensive weight correction index of each element in the electrical parameter data sequence in the three-dimensional imaging equipment fault detection matrix is ​​obtained; Obtaining a fitting curve for each row of data in the three-dimensional imaging device fault monitoring matrix according to the comprehensive weight correction index of each element in the three-dimensional imaging device fault monitoring matrix, and obtaining a fault monitoring result of the three-dimensional imaging device using a CNN neural network model based on the fitting curve; The method for obtaining the trend matching comprehensive correlation coefficient between the non-electrical parameter data and the electrical parameter data in the three-dimensional imaging device fault monitoring matrix according to the trend correlation undirected graph of the three-dimensional imaging device fault monitoring matrix is: Each row of data corresponding to the non-electrical parameter data in the three-dimensional imaging device fault monitoring matrix is ​​used as a non-electrical parameter data sequence of the three-dimensional imaging device fault monitoring matrix, and each row of data corresponding to the electrical parameter data in the three-dimensional imaging device fault monitoring matrix is ​​used as an electrical parameter data sequence of the three-dimensional imaging device fault monitoring matrix; Calculate the trend matching indirect correlation coefficient according to the corresponding relationship between the non-electrical parameter data sequence and the electrical parameter data sequence in the trend correlation undirected graph of the three-dimensional imaging equipment fault monitoring matrix; The sum of the synchronous trend matching correlation coefficient and the trend matching indirect correlation coefficient between each non-electrical parameter data sequence and each electrical parameter data sequence in the three-dimensional imaging device fault monitoring matrix is ​​taken as the first characteristic coefficient, the sum of all the first characteristic coefficients is taken as the second characteristic coefficient, and the ratio of the first characteristic coefficient to the second characteristic coefficient is taken as the trend matching comprehensive correlation coefficient between each non-electrical parameter data sequence and each electrical parameter data sequence; The method for calculating the trend matching indirect correlation coefficient according to the correspondence between the non-electrical parameter data sequence and the electrical parameter data sequence in the trend correlation undirected graph of the three-dimensional imaging device fault monitoring matrix is: In the formula, Indicates The non-electrical parameter data series and The trend fit indirect correlation coefficient between the electrical parameter data series; Represents the trend correlation in an undirected graph. The non-electrical parameter data series and The first The first The first The synchronization trend fit correlation coefficient between the monitoring data sequences corresponding to the nodes; Represents the trend correlation in an undirected graph. The non-electrical parameter data series and The first The first The first The actual weight of the edge between nodes; Represents the trend correlation in an undirected graph. The non-electrical parameter data series and The first The first The sequence number corresponding to each node; Represents the trend correlation in an undirected graph. The non-electrical parameter data series and The first The number of nodes in a path; Represents the trend correlation in an undirected graph. The non-electrical parameter data series and The number of paths between electrical parameter data sequences.

2. A method for intelligent fault detection of electrical equipment according to claim 1, characterized in that: The method for obtaining the division result of each row of data according to the change trend characteristics of each row of data in the three-dimensional imaging device fault monitoring matrix is: For each row of data in the three-dimensional imaging device fault monitoring matrix, the sequence composed of each row of data is used as the monitoring data sequence, the position number of each data in the monitoring data sequence is used as the horizontal coordinate, the size of each data value in the monitoring data sequence is used as the vertical coordinate, and the curve formed by the horizontal coordinate and the vertical coordinate is used as the monitoring data curve of the monitoring data sequence, and all the maximum points and minimum points of the monitoring data curve are obtained, and all the maximum points and minimum points are used as the maximum segmentation points and the minimum segmentation points of the monitoring data sequence respectively, and the sequence composed of all the maximum segmentation points and the minimum segmentation points in ascending time order is used as the segmentation sequence, and according to the segmentation sequence, in order from left to right, the sequence composed of the corresponding data in the monitoring data sequence between the adjacent maximum segmentation points and the minimum segmentation points is used as the decreasing trend sequence of the monitoring data sequence, and the sequence composed of the corresponding data in the monitoring data sequence between the adjacent minimum segmentation points and the maximum segmentation points is used as the increasing trend sequence of the monitoring data sequence.

3. The intelligent fault detection method for electrical equipment according to claim 1, characterized in that: The method for obtaining the synchronization trend matching correlation coefficient between different rows of data in the three-dimensional imaging device fault monitoring matrix according to the division result of each row of data in the three-dimensional imaging device fault monitoring matrix is: For each row of data in the three-dimensional imaging device fault monitoring matrix, the sequence composed of each row of data is used as a monitoring data sequence, a Bayesian curve fitting algorithm is used to obtain a Bayesian fitting curve corresponding to the monitoring data sequence, and a slope corresponding to each element in the monitoring data sequence is obtained according to the Bayesian fitting curve; Obtaining a synchronous increase matching factor between different monitoring data sequences according to the difference in growth trend sequences between different monitoring data sequences in the three-dimensional imaging device fault monitoring matrix; Acquire a synchronization reduction fit factor between different monitoring data sequences according to the difference in reduction trend sequence between different monitoring data sequences in the three-dimensional imaging device fault monitoring matrix; The sum of the synchronous increase fit factor and the synchronous decrease fit factor between the different monitoring data sequences is taken as the first correlation coefficient, the SBD distance between the different monitoring data sequences is taken as the second correlation coefficient, and the product of the first correlation coefficient and the second correlation coefficient is taken as the synchronous trend fit correlation coefficient between the different monitoring data sequences.

4. A method for intelligent fault detection of electrical equipment according to claim 3, characterized in that: The calculation method for obtaining the synchronous increase matching factor between different monitoring data sequences according to the difference in the growth trend sequence between different monitoring data sequences in the three-dimensional imaging device fault monitoring matrix is: In the formula, Indicates monitoring data series and The synchronous increase matching factor between the monitoring data series; and Respectively represent and The monitoring data series The first in the growth trend sequence The collection time corresponding to each data; and Respectively represent and The monitoring data series The first in the growth trend sequence The data corresponds to the slope; Indicates The monitoring data series The number of data in the growth trend series is The monitoring data series The minimum value of the number of data in a growth trend series; Indicates The number of growth trend sequences in the monitoring data series is The minimum value of the number of growth trend sequences in the monitoring data series; represents the adjustment parameter; is a natural constant.

5. The intelligent fault detection method for electrical equipment according to claim 3, characterized in that: The calculation method for obtaining the synchronous reduction fit factor between different monitoring data sequences according to the difference in the reduction trend sequence between different monitoring data sequences in the three-dimensional imaging device fault monitoring matrix is: In the formula, Indicates monitoring data series and The synchronization between the monitoring data series reduces the fit factor; and Respectively represent and The monitoring data series The first The collection time corresponding to each data; and Respectively represent and The monitoring data series The first The data corresponds to the slope; Indicates The monitoring data series The number of data in the trend series is reduced by The monitoring data series The minimum value among the number of data in the decreasing trend series; Indicates The number of decreasing trend sequences in the monitoring data series is related to the The minimum value of the number of decreasing trend sequences in the monitored data series; represents the adjustment parameter; is a natural constant.

6. A method for intelligent fault detection of electrical equipment according to claim 1, characterized in that: The method for constructing a trend-related undirected graph of a three-dimensional imaging device fault monitoring matrix according to the synchronization trend matching correlation coefficient between different rows of data in the three-dimensional imaging device fault monitoring matrix is: For each row of data in the three-dimensional imaging device fault monitoring matrix, a sequence composed of each row of data is used as a monitoring data sequence, a sequence composed of synchronization trend matching correlation coefficients between the monitoring data sequence and all other monitoring data sequences is sorted in ascending order as a synchronization trend matching correlation sequence of the monitoring data sequence, and a binary tree of the monitoring data sequence is constructed according to the synchronization trend matching correlation sequence; Each monitoring data sequence is regarded as a node in an undirected graph, the tree edit distance between the binary trees corresponding to the two monitoring data sequences is regarded as the initial weight between the corresponding two nodes, and the undirected graph constructed based on all monitoring data sequences, the initial weights of different monitoring data sequences and the undirected graph is regarded as the trend association undirected graph of the fault monitoring matrix of the three-dimensional imaging equipment.

7. The intelligent fault detection method for electrical equipment according to claim 1, characterized in that: The method for obtaining the comprehensive weight correction index of each element in the electrical parameter data sequence in the three-dimensional imaging device fault detection matrix according to the trend matching comprehensive correlation coefficient between the non-electrical parameter data and the electrical parameter data in the three-dimensional imaging device fault monitoring matrix is: For each electrical parameter data sequence in the three-dimensional imaging device fault monitoring matrix, the trend matching comprehensive correlation coefficient between each electrical parameter data sequence and each other non-electrical parameter data sequence is used as the fusion weight of each non-electrical parameter data sequence, all the non-electrical parameter data sequences are weightedly fused according to the corresponding fusion weights, and the weighted fusion result is used as the non-electrical parameter fusion sequence of each electrical parameter data sequence, and the predicted value of each element in each electrical parameter data sequence is obtained by using the ARIMAX model based on the non-electrical parameter fusion sequence of each electrical parameter data sequence; The comprehensive weight correction index of each element in each electrical parameter data sequence is obtained according to the true value and the predicted value corresponding to each element in each electrical parameter data sequence.

8. A method for intelligent fault detection of electrical equipment according to claim 7, characterized in that: The method for obtaining the comprehensive weight correction index of each element in each electrical parameter data sequence according to the true value and the predicted value corresponding to each element in each electrical parameter data sequence is: In the formula, Indicates The first The comprehensive weight correction index of the elements; Indicates The first The difference between the true value and the predicted value corresponding to the element; Indicates The maximum value of the difference between the true value and the predicted value corresponding to all elements in the electrical parameter data sequence; Indicates The first The original weights in the local weighted regression algorithm corresponding to the elements; is a natural exponential function.

Images