High-voltage cable fault diagnosis method and device

Abstract

The embodiment of the invention relates to a high-voltage cable fault diagnosis method and device, and the method comprises the steps: collecting to-be-detected gas through a non-dispersive infrared optical detector, an electrochemical gas sensor and a temperature and humidity sensor, and obtaining a first collection signal, a second collection signal and a third collection signal; respectively splicing the first acquisition signal and the second acquisition signal with the third acquisition signal to obtain a first spliced acquisition signal and a second spliced acquisition signal; respectively carrying out feature extraction on the first splicing acquisition signal and the second splicing acquisition signal to obtain a first feature signal and a second feature signal; analyzing the first characteristic signal and the second characteristic signal by using a fusion network; and diagnosing the high-voltage cable fault based on an analysis result. According to the technical scheme provided by the embodiment of the invention, fault diagnosis is carried out on the high-voltage cable by fusing the non-dispersive infrared signal and the gas chemical signal, so that the accuracy of fault diagnosis of the high-voltage cable is effectively improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of power detection, and in particular, to a method and device for diagnosing high-voltage cable faults. Background Art

[0002] With the continuous development of the social economy, the increasing urban electricity consumption has led to an increased load on high-voltage cables, and the overheating faults of high-voltage cables have become more prominent. As an important means of power transmission, once a fire accident occurs in a high-voltage cable, it will not only block the power transmission and damage other equipment in the high-voltage cable tunnel, but also pose a great threat to the lives and property safety of the masses. Therefore, being able to detect high-voltage cable outer sheath faults in a timely manner and give early warnings is of great significance for maintaining the stable operation of high-voltage cables.

[0003] Traditional high-voltage cable fault monitoring methods include methods such as infrared thermal imagers and manual inspections. These methods can only detect the temperature on the surface of the cable, are easily blocked by the field of view blind area, and at the same time, manual inspections cannot provide real-time data, have hysteresis and are also affected by subjective experience. Summary of the Invention

[0004] Based on the above situation of the prior art, the purpose of the embodiments of the present invention is to provide a method and device for diagnosing high-voltage cable faults, which reduce problems such as false alarms and low sensitivity easily caused by using a single-modal sensing technology, and effectively improve the accuracy of high-voltage cable fault diagnosis.

[0005] To achieve the above object, according to one aspect of the present invention, a method for diagnosing high-voltage cable faults is provided, including the steps of: Collecting the gas to be measured by using a non-dispersive infrared optical detector, an electrochemical gas sensor, and a temperature and humidity sensor respectively to obtain a first collected signal, a second collected signal, and a third collected signal; Splicing the first collected signal and the second collected signal with the third collected signal respectively to obtain a first spliced collected signal and a second spliced collected signal; Extracting features from the first spliced collected signal and the second spliced collected signal respectively to obtain a first feature signal and a second feature signal; Analyzing the first feature signal and the second feature signal by using a fusion network; Diagnosing the high-voltage cable fault based on the result of the analysis; Wherein, the gas to be measured is the thermally released gas of the high-voltage cable outer sheath.

[0006] Further, a time-domain neural network is used to extract features from the first spliced collected signal, and the first feature signal is expressed as: ; Among them, represents the first feature signal, represents the first activation function, represents the first spliced acquisition signal, represents the weight value of the convolution kernel, represents the bias value, represents the convolution kernel size value, represents the dilation factor of the convolution.

[0007] Furthermore, the gated multi-layer perceptron is used to extract features from the second spliced acquisition signal, and the second feature signal is expressed as: ; Among them, represents the second feature signal, represents the second spliced acquisition signal, represents the gated multi-layer perceptron processing.

[0008] Furthermore, the third acquisition signal includes a temperature acquisition signal and a humidity acquisition signal; the first spliced acquisition signal is expressed as: ; The second spliced acquisition signal is expressed as: ; Among them, represents the data splicing function, represents the first acquisition signal, represents the second acquisition signal, represents the temperature acquisition signal, represents the humidity acquisition signal.

[0009] Furthermore, the fusion network is a multi-modal fusion network based on the Transformer cross-attention mechanism. The multi-modal fusion network includes a multi-head attention module, and the multi-head attention module includes a first cross-attention module and a second cross-attention module.

[0010] Furthermore, analyzing the first feature signal and the second feature signal by using the fusion network includes: Generating a first Q vector, a first K vector, and a first V vector based on the first feature signal, and generating a second Q vector, a second K vector, and a second V vector based on the second feature signal; Generating a first attention score matrix based on the first Q vector and the first K vector, and generating a second attention score matrix based on the second Q vector and the second K vector; Input the first attention score matrix and the first V vector into the first cross-attention module, and input the second attention score matrix and the second V vector into the second cross-attention module; Concatenate the outputs of the first cross-attention module and the second cross-attention module to obtain an attention output vector; Based on the attention output vector and the first feature signal, obtain the concentration value of 2-EH gas in the gas to be measured.

[0011] Further, obtaining the concentration value of 2-EH gas in the gas to be measured based on the attention output vector and the first feature signal includes: Perform residual connection and normalization processing on the attention output vector and the first feature signal to obtain a fused feature; Obtain the output feature of the fusion network according to the fused feature; Based on the output feature, obtain the concentration value of 2-EH gas in the gas to be measured.

[0012] Further, the concentration value of 2-EH gas in the gas to be measured is expressed as: ; where represents the concentration value of 2-EH gas in the gas to be measured, represents the output feature of the fusion network, represents a linear transformation.

[0013] Further, diagnosing the high-voltage cable fault based on the analysis result includes: Perform early warning of high-voltage cable faults based on the concentration value of 2-EH gas in the gas to be measured.

[0014] According to another aspect of the present invention, there is provided a high-voltage cable fault diagnosis device, including: A signal acquisition module, which respectively uses a non-dispersive infrared optical detector, an electrochemical gas sensor, and a temperature and humidity sensor to collect the gas to be measured, and obtains a first acquisition signal, a second acquisition signal, and a third acquisition signal; A signal splicing module, which is used to splice the first acquisition signal and the second acquisition signal with the third acquisition signal respectively to obtain a first spliced acquisition signal and a second spliced acquisition signal; A feature extraction module, which is used to extract features from the first spliced acquisition signal and the second spliced acquisition signal respectively to obtain a first feature signal and a second feature signal; An analysis module, which uses a fusion network to analyze the first feature signal and the second feature signal; A fault diagnosis module, which diagnoses the high-voltage cable fault based on the analysis result; Wherein, the gas to be measured is the thermally released gas from the outer sheath of the high-voltage cable.

[0015] In summary, the embodiments of the present invention provide a high-voltage cable fault diagnosis method and device. The method includes the steps of: respectively collecting the gas to be measured by using a non-dispersive infrared optical detector, an electrochemical gas sensor, and a temperature and humidity sensor to obtain a first collection signal, a second collection signal, and a third collection signal; splicing the first collection signal and the second collection signal with the third collection signal respectively to obtain a first spliced collection signal and a second spliced collection signal; respectively extracting features from the first spliced collection signal and the second spliced collection signal to obtain a first feature signal and a second feature signal; analyzing the first feature signal and the second feature signal by using a fusion network; and diagnosing the high-voltage cable fault based on the analysis result. The technical solution provided by the embodiments of the present invention diagnoses the overheating fault of the outer sheath of the high-voltage cable by fusing the non-dispersive infrared signal and the gas chemical signal, fully utilizes the high sensitivity of the electrochemical gas sensor and the high selectivity of the non-dispersive infrared optical detector, and avoids the problem that the threshold alarm method using only the electrochemical gas sensor is easily affected by other interfering gases and causes false alarms under complex actual working conditions, or the problem that the non-dispersive infrared absorption gas detection method cannot effectively identify low-concentration characteristic gases. The embodiments of the present invention use a fusion network to analyze the fused gas chemical signal and non-dispersive infrared signal, improve the detection range and accuracy, can detect the overheating fault of the outer sheath of the high-voltage cable at an early stage to avoid the occurrence of high-voltage cable fires, and effectively improve the safety of high-voltage cable operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flowchart of the high-voltage cable fault diagnosis method provided by the embodiments of the present invention; Figure 2 is a threshold diagram for dividing the state interval of the thermally released characteristic gas of the outer sheath of the high-voltage cable according to the embodiments of the present invention; Figure 3 is a system block diagram for implementing the high-voltage cable fault diagnosis method according to the embodiments of the present invention; Figure 4 is a response curve of a certain electrochemical gas sensor in the embodiments of the present invention when a series of concentration gradients of 2-EH gas are introduced; Figure 5 is a response curve of the non-dispersive infrared optical detector in the embodiments of the present invention when a series of concentration gradients of 2-EH gas are introduced in an intermittent interfering gas environment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. In addition, in the following descriptions, descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present invention.

[0018] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should have the ordinary meaning understood by those of ordinary skill in the art to which the present invention pertains. The terms "first", "second", and similar terms used in one or more embodiments of the present invention do not denote any order, quantity, or importance, but are merely used to distinguish different components. The terms "comprising", "including", or similar terms mean that the elements or items appearing before this term cover the elements or items listed after this term and their equivalents, without excluding other elements or items. The terms "connected" or "coupled" or similar terms are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect.

[0019] The technical solutions of the present invention will be described in detail below in conjunction with the accompanying drawings. An embodiment of the present invention provides a method for diagnosing high-voltage cable faults. Figure 1 The flowchart of the method for diagnosing high-voltage cable faults according to an embodiment of the present invention is shown in Figure 1 as follows, and the method includes the following steps: S202. Respectively use a non-dispersive infrared optical detector, an electrochemical gas sensor, and a temperature and humidity sensor to collect the gas to be measured, and obtain a first collection signal, a second collection signal, and a third collection signal. Among them, the gas to be measured is the thermally released gas of the outer sheath of the high-voltage cable. In the embodiment of the present invention, by online monitoring and diagnosing the characteristic gases in the early overheated released gas of the outer sheath of the high-voltage cable, timely warning of faults in the outer sheath of the high-voltage cable can be achieved. In the embodiment of the present invention, the non-dispersive infrared optical detector includes, for example, an infrared emission module, a 2-EH sensor module, and a reference sensor module, and is used to collect non-dispersive infrared signals in the gas to be measured; the electrochemical gas sensor is used to collect chemical signals in the gas to be measured; the temperature and humidity sensor is used to collect the current temperature and humidity. The obtained first collection signal is a non-dispersive infrared signal, the second collection signal is a gas chemical signal, and the third collection includes a temperature collection signal and a humidity collection signal. Before collection, each detector and sensor can be preheated so that each detector and sensor can work normally.

[0020] S204. Concatenate the first acquisition signal and the second acquisition signal with the third acquisition signal respectively to obtain a first concatenated acquisition signal and a second concatenated acquisition signal. Before concatenating the signals, the first acquisition signal and the second acquisition signal can be preprocessed first. The first acquisition signal is filtered and then differentially amplified, and then concatenated with the third acquisition signal; the second acquisition signal is filtered and amplified and then concatenated with the third acquisition signal. The first concatenated acquisition signal can be expressed as: ; The second concatenated acquisition signal can be expressed as: ; Wherein, represents a data concatenation function, represents the first acquisition signal, which can be a time series of non-dispersive infrared signals after filtering and differential amplification. In the embodiments of the present invention, is a matrix with a dimension of ; represents the number of sampling points; represents the second acquisition signal, which can be a time series of gas chemical signals after filtering and amplification, is a matrix with a dimension of ; represents the time series of temperature acquisition signals, is a matrix with a dimension of ; represents the time series of humidity acquisition signals, is a matrix with a dimension of ; is a matrix with a dimension of ; is a matrix with a dimension of . By concatenating the non-dispersive infrared signal and the gas chemical signal with the temperature and humidity signals, the changes in the non-dispersive infrared light absorption characteristics and the gas chemical signal response characteristics caused by temperature and humidity changes can be compensated, thereby improving the accuracy of the analysis results.

[0021] S206. Extract features from the first concatenated acquisition signal and the second concatenated acquisition signal respectively to obtain a first feature signal and a second feature signal. In the embodiments of the present invention, a time-domain neural network TCN is used to extract features from the first concatenated acquisition signal, and the first feature signal extracted can be expressed as: ; Wherein, represents the first feature signal, represents the first activation function, represents the first concatenated acquisition signal, represents the weight value of the convolution kernel, represents the offset value, represents the size value of the convolutional kernel, represents the dilation factor of the convolution, which is an adjustable parameter, represents the acquisition time, represents the elements of the input sequence of the temporal convolutional network (TCN). There is a causal relationship between the convolutional network layers of the TCN, which can memorize the dynamic change relationship of the non-dispersive infrared signal. At the same time, it also includes a dropout technique, which can improve the stability of the network.

[0022] In the embodiment of the present invention, the gated multilayer perceptron (gated Multilayer Perceptron, hereinafter referred to as "gMLP") is used to extract features from the second spliced acquisition signal. The second spliced acquisition signal is input into the gMLP module stacked continuously for two layers. The processing process of each gMLP module is shown in the following formula: ; ; ; Among them, and represent transformation matrices with the same dimension as the channel dimension, represents the second spliced acquisition signal, represents a linear mapping with the same dimension as the channel dimension, represents the second activation function, represents the output sequence of the gMLP module. The second feature signal extracted can be expressed as: ; Among them, represents the second feature signal, represents the second spliced acquisition signal, represents the processing of the gated multilayer perceptron. As a feature extraction network, gMLP has the advantages of high computational efficiency and strong feature extraction ability. The mechanism combining gate control and multilayer perceptron it adopts can dynamically adjust the importance of features according to the task.

[0023] S208. Analyze the first feature signal and the second feature signal using a fusion network. In the embodiments of the present invention, the fusion network fuses the non-dispersive infrared signal and the gas chemical signal based on the Transformer cross-attention mechanism, and analyzes the concentration of the characteristic gas using the fused data features, making full use of the high selectivity of the non-dispersive infrared signal and the high sensitivity of the gas chemical signal. The fusion network adopts a multi-modal fusion network based on the Transformer cross-attention mechanism, and this multi-modal fusion network includes a multi-head attention module, where the multi-head attention module includes two cross-attention modules. The composition of other parts of the multi-modal fusion network adopted in the embodiments of the present invention is the same as that of the existing multi-modal fusion network structure, and also includes structures such as a feature fusion layer, a normalization layer, and a multi-layer perceptron layer. This step can be achieved through the following sub-steps: S2081. Generate a first Q vector, a first K vector, and a first V vector based on the first feature signal, and generate a second Q vector, a second K vector, and a second V vector based on the second feature signal. Among them, the Q vector, the K vector, and the V vector are respectively represented as matrix, matrix, and matrix: ; ; ; Among them, is a matrix with a dimension of , is a matrix with a dimension of , is a matrix with a dimension of , respectively represent the weight matrices , corresponding to the matrices generated from the first feature signal and the second feature signal and their related parameters.

[0024] S2082. Generate a first attention score matrix based on the first Q vector and the first K vector, and generate a second attention score matrix based on the second Q vector and the second K vector. In this step, calculate the attention score matrix between the non-dispersive infrared signal and the gas chemical signal based on the Q vector and the V vector. The attention score matrix can be expressed as: ; Among them, represents the attention score matrix, represents the key vector, represents the matrix transpose, and the dimension is, for example, 8, , corresponding to the first attention score matrix and the second attention score matrix and their related parameters.

[0025] S2083. Input the first attention score matrix and the first V vector into the first cross-attention module, input the second attention score matrix and the second V vector into the second cross-attention module, and concatenate the outputs of the first cross-attention module and the second cross-attention module to obtain an attention output vector, which is the output of this multi-head attention module. The output of the cross-attention module can be expressed as: ; where , corresponding to the outputs of the first cross-attention module and the second cross-attention module and their related parameters respectively. The concatenated attention output vector can be expressed as: ; where is a matrix with dimension 2H 1, representing the concatenated attention output vector.

[0026] S2084. Obtain the concentration value of 2-EH gas in the gas to be measured based on the attention output vector and the first feature signal. In this step, perform residual connection and normalization processing on the attention output vector and the first feature signal to obtain a fused feature : ; According to this fused feature obtain the output feature of the fused network : ; where represents layer normalization processing.

[0027] Based on the output feature, obtain the concentration value of 2-EH gas in the gas to be measured. The concentration value of 2-EH gas in the gas to be measured can be expressed as: ; where represents the concentration value of the gas to be measured, represents the output feature of the fused network, represents linear transformation.

[0028] S210. Diagnose the high-voltage cable fault based on the analysis results. In this step, high-voltage cable fault early warning is carried out based on the concentration value of the fault characteristic gas. In the embodiment of the present invention, a gas chromatograph-mass spectrometer is used to measure the volume fractions of the thermally released gas of the outer sheath insulating material of a certain type of high-voltage cable at 100°, 120°, 140°, 160°, 180° and 200° when the specified maximum operating temperature is 90° and above. After analysis, it is found that 2-ethylhexanol (2-EH) gas exists in the entire range of 90° to 200° during the overheating operation of the high-voltage cable, and its concentration has a large variation difference. Therefore, 2-EH gas is selected as the characteristic gas for characterizing the overheating degree of the outer sheath of the high-voltage cable. The threshold diagram based on which the interval is divided is as shown in Figure 2 Figure

[0029] The time-domain neural network TCN, the gated multi-layer perceptron gMLP network, and the Transformer network based on the cross-attention mechanism involved in the above embodiments of the present invention can all be implemented using existing neural network architectures, and their internal structures will be omitted from the description. The high-voltage cable fault diagnosis method provided by the above embodiments of the present invention can be adopted Figure 3The system framework implementation shown in the figure. The system includes a single-chip microcomputer module, a non-dispersive infrared signal acquisition module, a chemical signal acquisition module, a temperature and humidity sensor module, an acoustic-optic alarm module, and a host computer. The chemical signal acquisition module includes a primary detection gas chamber, an electrochemical gas sensor unit, and a signal sampling unit. The electrochemical gas sensor module unit includes multiple electrochemical gas sensors, a matching resistor network supporting the electrochemical gas sensors, a voltage division network for the output voltage of the electrochemical gas sensors, and a primary filtering network composed of capacitors and resistors. The signal sampling unit includes a high-precision operational amplifier chip, a high-speed analog-to-digital sampling chip, and a supporting reference voltage circuit composed of resistors and capacitors. The non-dispersive infrared signal acquisition module includes a non-dispersive infrared gas detection module unit and an infrared control unit. The infrared control unit includes an infrared drive circuit, a differential amplification circuit, and a voltage sampling unit. The infrared drive circuit is used to amplify the 1Hz voltage signal generated by the single-chip microcomputer and drive the infrared light source to radiate infrared light with a period of 1Hz. The differential amplification circuit is used to perform differential processing on the output voltages of the 2-EH sensor and the reference sensor and then collect them by the voltage sampling unit. The temperature and humidity sensor module includes a temperature and humidity sensing chip and a signal conditioning circuit composed of capacitors and resistors supporting the temperature and humidity chip. The host computer is used to receive the preprocessed non-dispersive infrared signal and gas chemical signal, analyze the signals using the deployed feature extraction network and fusion network, and give an early warning level.

[0030] The gas path of the system mainly includes two sections: the first section is from the air inlet to the primary detection gas chamber in the chemical signal acquisition module, and the second section is from the secondary detection gas chamber in the non-dispersive infrared signal acquisition module to the air outlet. The air inlet in this gas path is directly connected to the primary detection gas chamber and is used to introduce the gas to be measured. The air outlet of this gas path is behind the secondary detection gas chamber and is used to discharge the gas after the test. After introducing the gas to be measured into the system, the gas to be measured first contacts the electrochemical gas sensors in the primary detection gas chamber and then contacts the infrared light in the secondary detection gas chamber.

[0031] According to another aspect of the present invention, a high-voltage cable fault diagnosis device is provided, including: A signal acquisition module that respectively uses a non-dispersive infrared optical detector, an electrochemical gas sensor, and a temperature and humidity sensor to collect the gas to be measured, and obtains a first acquisition signal, a second acquisition signal, and a third acquisition signal; A signal splicing module for splicing the first acquisition signal and the second acquisition signal with the third acquisition signal respectively to obtain a first spliced acquisition signal and a second spliced acquisition signal; A feature extraction module for respectively extracting features from the first spliced acquisition signal and the second spliced acquisition signal to obtain a first feature signal and a second feature signal; An analysis module that analyzes the first feature signal and the second feature signal using a fusion network; A fault diagnosis module diagnoses the high-voltage cable fault based on the result of the analysis; Wherein, the gas to be measured is the thermally released gas of the outer sheath of the high-voltage cable.

[0032] In the high-voltage cable fault diagnosis device provided by this embodiment of the present invention, the specific processes for each module to implement its functions are the same as the steps in the high-voltage cable fault method provided by the above embodiment of the present invention, and the repeated description thereof will be omitted here.

[0033] To detect the effectiveness of the high-voltage cable fault diagnosis method of this embodiment of the present invention, a gas-sensing test platform is built for testing. Considering the common interfering gases carbon monoxide and hydrogen sulfide in the high-voltage cable tunnel, a mass flow controller is used to introduce a mixed gas of 2-EH, CO, and H 2 S into the system for collection. The collected non-dispersive infrared signals and gas chemical signals are analyzed for the 2-EH concentration using various algorithms. An error within 5% in the 2-EH concentration detection is considered accurate, and an experiment is carried out. The test results are shown in Table 1.

[0034] Table 1 Analysis of 2-EH concentration by different algorithms

[0035] In this embodiment of the present invention, 2-EH gas with concentration gradients of 10 ppm, 20 ppm, 30 ppm, 40 ppm, and 50 ppm is introduced into the above system, and a certain chemical signal in the electrochemical gas sensor module is directly output from the upper computer. Figure 4 The response curves under a series of concentration gradients of 2-EH gas are shown, and it can be seen that the electrochemical gas sensor has high sensitivity. 2-EH gas with concentration gradients of 50 ppm, 100 ppm, and 200 ppm is set, and a mixed interfering gas of 20 ppm of CO and 10 ppm of H 2 S is continuously introduced for 10 minutes every 5 minutes, and the non-dispersive infrared signal is directly output from the upper computer. Figure 5 The response curves of the non-dispersive infrared optical detector under a series of concentration gradients of 2-EH gas in an intermittent interfering gas environment are shown, and it can be seen that the result is not interfered by the mixed interfering gas.

[0036] To verify the detection ability of the high-voltage cable fault diagnosis method provided by this embodiment of the present invention and other diagnosis methods when the outer sheath of the high-voltage cable is overheated, a certain type of high-voltage cable outer sheath material is selected and tested at different overheating temperatures. The specific test comparison results are shown in Table 2. It can be seen that the non-dispersive infrared (hereinafter referred to as "NDIR") detection method fails to alarm at low concentrations.

[0037] Comparison of Overheating Detection Methods for the Outer Sheath of High-Voltage Cables at Different Temperatures without Interference (Table 2) 70℃ 90℃ 120℃ 140℃ 160℃ 180℃ 200℃ The present invention Normal Normal Warning Warning Alarm Alarm Alarm Electrochemical sensor threshold detection method Normal Normal Warning Warning Alarm Alarm Alarm NDIR detection method No No No No Alarm Alarm Alarm Smoke alarm No No No No No No No

[0038] Considering the possible interfering gases in the actual high-voltage cable tunnel, to verify the detection ability of the high-voltage cable fault diagnosis method provided in the embodiments of the present invention and other fault diagnosis methods for overheating of the outer sheath of high-voltage cables in the presence of interfering gases, a certain type of outer sheath material of high-voltage cable is selected, and 20 ppm of CO and 10 ppm of H 2 S mixed interfering gas is introduced for testing at different overheating temperatures of the outer sheath material of the high-voltage cable. The specific test comparison results are shown in Table 3. It can be seen that the method provided in the embodiments of the present invention is not affected by the interfering gas and accurately gives the alarm level, but the electrochemical sensor threshold detection method is severely affected, and at the same time, the NDIR detection method fails to alarm at low concentrations.

[0039] Comparison of Overheating Detection Methods for the Outer Sheath of High-Voltage Cables at Different Temperatures in the Presence of Interference (Table 3) 70℃ 90℃ 120℃ 140℃ 160℃ 180℃ 200℃ The present invention Normal Normal Warning Warning Alarm Alarm Alarm Electrochemical sensor threshold detection method False alarm False alarm False alarm False alarm Alarm Alarm Alarm NDIR detection method No No No No Alarm Alarm Alarm Smoke alarm No No No No No No No

[0040] In summary, the embodiments of the present invention relate to a high-voltage cable fault diagnosis method and device. The method includes the steps of: respectively collecting the gas to be measured by using a non-dispersive infrared optical detector, an electrochemical gas sensor, and a temperature and humidity sensor to obtain a first collection signal, a second collection signal, and a third collection signal; splicing the first collection signal and the second collection signal with the third collection signal respectively to obtain a first spliced collection signal and a second spliced collection signal; respectively extracting features from the first spliced collection signal and the second spliced collection signal to obtain a first feature signal and a second feature signal; analyzing the first feature signal and the second feature signal by using a fusion network; and diagnosing the high-voltage cable fault based on the analysis result. The technical solution provided by the embodiments of the present invention diagnoses the overheating fault of the outer sheath of the high-voltage cable by fusing the non-dispersive infrared signal and the gas chemical signal, makes full use of the high sensitivity of the electrochemical gas sensor and the high selectivity of the non-dispersive infrared optical detector, and avoids the problem that the electrochemical gas sensor threshold alarm method is easily affected by other interfering gases in complex actual working conditions and causes false alarms, or the non-dispersive infrared absorption gas detection method cannot effectively identify low-concentration characteristic gases. The embodiments of the present invention use a fusion network to analyze the fused gas chemical signal and non-dispersive infrared signal, improve the detection range and accuracy, can detect the overheating fault of the outer sheath of the high-voltage cable at an early stage to avoid the occurrence of high-voltage cable fires, and effectively improve the safety of high-voltage cable operation.

[0041] It should be understood that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the present invention (including the claims) is limited to these examples; within the concept of the present invention, the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations in different aspects of one or more embodiments of the present invention as described above, and they are not provided in detail for the sake of brevity. The above specific embodiments of the present invention are only used for exemplary illustration or explanation of the principle of the present invention and do not constitute a limitation to the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc. made without departing from the spirit and scope of the present invention shall be included within the protection scope of the present invention. In addition, the appended claims of the present invention are intended to cover all variations and modifications that fall within the scope and boundaries of the appended claims, or equivalent forms of such scope and boundaries.

Claims

1. A high voltage cable fault diagnosis method, characterized in that: Includes steps: A non-dispersive infrared optical detector, an electrochemical gas sensor and a temperature and humidity sensor are respectively used to collect the gas to be measured to obtain a first collection signal, a second collection signal and a third collection signal; Splicing the first acquisition signal and the second acquisition signal with the third acquisition signal respectively to obtain a first spliced ​​acquisition signal and a second spliced ​​acquisition signal; Performing feature extraction on the first spliced ​​acquisition signal and the second spliced ​​acquisition signal respectively to obtain a first feature signal and a second feature signal; Analyzing the first characteristic signal and the second characteristic signal by using a fusion network; diagnosing the high voltage cable fault based on the analysis result; Wherein, the gas to be tested is the thermally released gas from the outer sheath of the high-voltage cable.

2. The method according to claim 1, characterized in that The first spliced ​​acquisition signal is subjected to feature extraction using a time domain neural network, and the first feature signal is expressed as: ； in, represents the first characteristic signal, represents the first activation function, represents the first splicing acquisition signal, Indicates the collection time, represents the weight value of the convolution kernel, Indicates the bias value, Indicates the size of the convolution kernel. Represents the dilation factor of the convolution.

3. The method according to claim 2, characterized in that The second spliced ​​acquisition signal is subjected to feature extraction using a gated multilayer perceptron, and the second feature signal is expressed as: ； in, represents the second characteristic signal, Indicates the second splicing acquisition signal, Represents gated multilayer perceptron processing.

4. The method according to claim 3, characterized in that The third acquisition signal includes a temperature acquisition signal and a humidity acquisition signal; the first spliced ​​acquisition signal is expressed as: ； The second spliced ​​acquisition signal is expressed as: ； in, represents the data concatenation function, represents the first acquisition signal, represents the second acquisition signal, Indicates the temperature acquisition signal, Indicates humidity acquisition signal.

5. The method according to claim 4, characterized in that The fusion network is a multimodal fusion network based on the Transformer cross-attention mechanism, and the multimodal fusion network includes a multi-head attention module, and the multi-head attention module includes a first cross-attention module and a second cross-attention module.

6. The method according to claim 5, characterized in that Analyzing the first characteristic signal and the second characteristic signal by using a fusion network includes: Generate a first Q vector, a first K vector, and a first V vector based on the first characteristic signal, and generate a second Q vector, a second K vector, and a second V vector based on the second characteristic signal; Generate a first attention score matrix based on the first Q vector and the first K vector, and generate a second attention score matrix based on the second Q vector and the second K vector; Input the first attention score matrix and the first V vector into a first cross-attention module, and input the second attention score matrix and the second V vector into a second cross-attention module; Concatenate the outputs of the first cross-attention module and the second cross-attention module to obtain an attention output vector; The concentration value of 2-EH gas in the gas to be measured is obtained based on the attention output vector and the first characteristic signal.

7. The method according to claim 6, characterized in that Obtaining a concentration value of 2-EH gas in the gas to be measured based on the attention output vector and the first characteristic signal, including: Perform residual connection and normalization on the attention output vector and the first feature signal to obtain a fused feature; Obtaining output features of the fusion network according to the fusion features; The concentration value of 2-EH gas in the gas to be measured is obtained based on the output characteristics.

8. The method according to claim 7, characterized in that The concentration value of 2-EH gas in the gas to be measured is expressed as: ； in, Indicates the concentration value of 2-EH gas in the gas to be tested. represents the output features of the fusion network, Represents a linear transformation.

9. The method according to claim 8, characterized in that The high voltage cable fault is diagnosed based on the analysis result, including: A high-voltage cable fault warning is performed based on the concentration value of the 2-EH gas in the gas to be tested.

10. A high voltage cable fault diagnosis device, characterized in that: include: The signal acquisition module uses a non-dispersive infrared optical detector, an electrochemical gas sensor, and a temperature and humidity sensor to collect the gas to be measured, and obtains a first acquisition signal, a second acquisition signal, and a third acquisition signal; A signal splicing module, used for splicing the first acquisition signal and the second acquisition signal with the third acquisition signal respectively to obtain a first spliced ​​acquisition signal and a second spliced ​​acquisition signal; A feature extraction module, used to extract features from the first spliced ​​acquisition signal and the second spliced ​​acquisition signal respectively to obtain a first feature signal and a second feature signal; An analysis module, using a fusion network to analyze the first characteristic signal and the second characteristic signal; A fault diagnosis module, diagnosing the high-voltage cable fault based on the analysis result; Wherein, the gas to be tested is the thermally released gas from the outer sheath of the high-voltage cable.

Images