Gas insulation fault identification method based on improved support vector machine algorithm

Abstract

The invention discloses a gas insulation fault identification method based on an improved support vector machine algorithm, and relates to the technical field of power equipment partial discharge detection, and the method comprises the steps: carrying out the simulation and partial discharge test of a typical fault type in gas insulation electrical equipment, and collecting a partial discharge signal; extracting statistical characteristic quantities, and performing dimensionality reduction and screening on the extracted characteristic quantities by utilizing a chi-square test method; and processing the optimized feature set by using an improved support vector machine algorithm, and identifying the type of the gas insulation fault. According to the method, the partial discharge multi-physical feature chi-square test optimization set is established through extraction and optimization of the partial discharge multi-physical features on the data level, compared with a traditional single physical model, the recognition effect is improved, and the problem that the misjudgment rate between corona discharge and suspension discharge is high in the traditional method is solved.

Description

Technical Field

[0001] The invention relates to the technical field of partial discharge detection of electric power equipment, and in particular to a gas insulation fault identification method based on an improved support vector machine algorithm. Background Art

[0002] Gas Insulated Switchgear (GIS) usually uses SF6 and other gases as insulating media. It has the advantages of high electrical strength, small footprint, and less maintenance workload, and has been widely used. However, due to the influence of some unavoidable factors in the production, transportation, assembly, and long-term service of GIS equipment, there are sometimes insulation defects such as metal spikes, surface abnormalities of insulating components, and floating potential inside the equipment, which bring different degrees of safety hazards to the operating equipment. In recent years, with the continuous improvement of partial discharge detection methods and the progress of information technology and machine learning technology, it is possible to use a variety of physical measurement methods to jointly detect partial discharge of power equipment, obtain multi-modal partial discharge information, and realize the state maintenance work mode of comprehensive diagnosis and decision-making.

[0003] After detecting the occurrence of partial discharge inside the equipment, it is an important and valuable task to identify the type of partial discharge and assess the severity of the discharge. Experienced operation and maintenance personnel can distinguish the type of partial discharge based on the pulse waveform characteristics of the partial discharge or the shape of the PRPD spectrum, so as to further determine the insulation status and disposal measures of the equipment. With the technological progress and scenario expansion in the field of machine learning, its application in partial discharge pattern recognition has also received more and more attention and research. At present, the mainstream practice in the field is to use the measured PRPD spectrum for statistical feature extraction and other methods, and input the feature quantity into the algorithm model for learning, training and recognition. However, in the current field application, due to the inherent limitations of the detection methods and equipment, the phenomenon of misdiagnosis and false alarm still occurs from time to time, and there is still a lack of relevant research and practical experience in the joint detection work using multiple detection methods. In particular, there are application difficulties in the following two specific scenarios: there are multiple physical forms of detection equipment on site, and the multi-physical information of partial discharge is obtained through measurement, but it is difficult to unify the data format and perform data reuse and comprehensive diagnosis; there are multiple state maintenance equipment with diagnostic functions on site, but different equipment has obtained contradictory identification results. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: the existing gas insulation fault detection method has low fault type identification accuracy, frequent misdiagnosis and false alarm phenomena, and difficulty in realizing comprehensive diagnosis of multimodal information, as well as how to improve the accuracy of partial discharge type identification and diagnostic efficiency.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a gas insulation fault identification method based on an improved support vector machine algorithm, comprising simulating and performing partial discharge tests on typical fault types inside gas-insulated electrical equipment, and collecting partial discharge signals; extracting statistical feature quantities, and using the chi-square test method to reduce the dimension and screen the extracted feature quantities; using the improved support vector machine algorithm to process the optimized feature set and identify the gas insulation fault type.

[0007] As a preferred solution of the gas insulation fault identification method based on the improved support vector machine algorithm described in the present invention, the typical fault types include corona discharge defects, surface discharge defects, and suspended discharge defects.

[0008] As a preferred solution of the gas insulation fault identification method based on the improved support vector machine algorithm described in the present invention, wherein: the collection of partial discharge signals includes synchronous collection of partial discharge pulse current signals, ultra-high frequency signals, optical signals and ultrasonic signals through Rogowski coils, ultra-high frequency sensors, optical sensors, acoustic emission sensors and digital acquisition cards; the detection bandwidth of the Rogowski coil is 460kHz~120MHz, the maximum allowable current of the optical sensor is 10mA, the photon detection efficiency is 50% when the external 30V bias power supply is connected, the dark count rate is 150kHz / mm2, and the residual pulse is 5%, the ultra-high frequency sensor adopts a double-helix ultra-high frequency sensor, the receiving antenna working bandwidth is 300MHz-1.5GHz, the ultrasonic sensor uses a narrow-band acoustic emission sensor, and the detection frequency range is 20kHz-180kHz.

[0009] As a preferred solution of the gas insulation fault identification method based on the improved support vector machine algorithm described in the present invention, wherein: the extraction of statistical feature quantities includes selecting the average discharge amplitude A of the statistical feature quantity for the pulse current method, the optical measurement method and the ultra-high frequency method. av , average discharge frequency N, skewness Sk, steepness Ku, phase asymmetry φ, discharge asymmetry Q, cross-correlation factor Cc and ultrasonic signal dispersion entropy DE k ; The average value of all pulse amplitudes in the positive half cycle within the measurement time range of an integral multiple of the power frequency cycle is recorded as A av+ The average value of all pulse amplitudes in the positive half cycle is recorded as A av- , the number of discharge pulses occurring in the positive half cycle within the unit measurement time t is recorded as N +, the number of discharge pulses occurring in the negative half cycle is recorded as N - .

[0010] As a preferred solution of the gas insulation fault identification method based on the improved support vector machine algorithm described in the present invention, the pulse current method includes calculating the skewness, which is expressed as:

[0011]

[0012] Among them, A i is the amplitude of the ith discharge pulse, σ is the standard deviation of the discharge pulse amplitude within the statistical range, n is the number of discharge pulses within the statistical range, and the steepness is calculated as:

[0013]

[0014] Calculate the phase asymmetry, expressed as:

[0015] φ=φ in- / φ in+

[0016] Among them, φ in+ is the difference between the positive half-cycle discharge starting phase and 0 degrees, φ in- It is the difference between the starting phase of the negative half-cycle discharge and 180 degrees.

[0017] As a preferred solution of the gas insulation fault identification method based on the improved support vector machine algorithm described in the present invention, the pulse current method also includes calculating the cross-correlation factor, which is expressed as:

[0018]

[0019] Among them, x i' is the sum of the discharge pulse amplitudes in the i'th phase window of the positive half cycle, y i' It is the sum of the discharge pulse amplitudes in the i'th phase window in the negative half cycle, and n' is the number of phase windows in the positive and negative half cycles.

[0020] As a preferred solution of the gas insulation fault identification method based on the improved support vector machine algorithm described in the present invention, the optical measurement method includes calculating the dispersion entropy based on the Shannon entropy quantification signal distribution characteristics in frequency or time, which is expressed as:

[0021]

[0022] Where n” represents the total number of possible states of the random variable k, p i” represents the probability that the random variable k is in the i'th state.

[0023] Another object of the present invention is to provide a gas insulation fault identification system based on an improved support vector machine algorithm, which can process the optimized feature set using the improved support vector machine algorithm to identify the type of gas insulation fault, thereby solving the problem of low accuracy of current partial discharge detection technology for power equipment.

[0024] As a preferred solution of the gas insulation fault identification system based on the improved support vector machine algorithm described in the present invention, it includes: a signal acquisition module, a feature extraction module, and a fault identification module;

[0025] The signal acquisition module is used to simulate and perform partial discharge tests on typical fault types inside gas-insulated electrical equipment, and to collect partial discharge signals; the feature extraction module is used to extract statistical feature quantities, and to reduce the dimension and screen the extracted feature quantities using the chi-square test method; the fault identification module is used to process the optimized feature set using an improved support vector machine algorithm, and to identify gas insulation fault types.

[0026] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a gas insulation fault identification method based on an improved support vector machine algorithm.

[0027] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a gas insulation fault identification method based on an improved support vector machine algorithm.

[0028] Beneficial effects of the present invention: The gas insulation fault identification method based on the improved support vector machine algorithm provided by the present invention adopts multiple sensors to synchronously measure partial discharge data. Different sensors can detect different types of signals generated by partial discharge, such as ultrasound, light, electrical signals and chemical substances. By comprehensively analyzing these signals, partial discharge can be more accurately identified and located, and the sensitivity and reliability of detection can be improved. Partial discharge signals may be affected by external electromagnetic interference. The use of multiple sensors can utilize their respective resistance to interference. For example, UHF sensors have a high signal-to-noise ratio and are resistant to external electromagnetic interference, thereby improving the accuracy of detection signals. Due to the limited training samples, too many feature quantities It is easy to cause overfitting in the model training process, which negatively affects the generalization and accuracy of the partial discharge diagnosis model. Therefore, the present invention extracts and optimizes the multi-physical features of partial discharge at the data level, and establishes a partial discharge multi-physical feature chi-square test optimization set. Under the same algorithm model, the local discharge type recognition accuracy of the multi-physical feature chi-square optimization set can reach 96.6%, and the local discharge severity can be recognized with an accuracy of 92.7%. Compared with the traditional single physical model, the recognition effect is improved, and the problem of high misjudgment rate between corona discharge and suspension discharge in traditional methods is solved. The present invention achieves better results in accuracy, efficiency and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0030] Figure 1 An overall flow chart of a gas insulation fault identification method based on an improved support vector machine algorithm provided in the first embodiment of the present invention.

[0031] Figure 2 A diagram of a gas-insulated partial discharge multi-physics experimental platform for a gas-insulated fault identification method based on an improved support vector machine algorithm provided in the second embodiment of the present invention.

[0032] Figure 3 A diagnostic model effect diagram of a gas insulation fault identification method based on an improved support vector machine algorithm provided in the second embodiment of the present invention.

[0033] Figure 4 An overall flow chart of a gas insulation fault identification system based on an improved support vector machine algorithm provided in the third embodiment of the present invention. DETAILED DESCRIPTION

[0034] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0035] Example 1, reference Figure 1 , as an embodiment of the present invention, provides a gas insulation fault identification method based on an improved support vector machine algorithm, comprising:

[0036] S1: Simulate typical fault types inside gas-insulated electrical equipment and perform partial discharge tests to collect partial discharge signals.

[0037] Furthermore, typical fault types include corona discharge defects, surface discharge defects, and suspended discharge defects.

[0038] It should be noted that the collection of local discharge signals includes the synchronous collection of local discharge pulse current signals, UHF signals, optical signals and ultrasonic signals through Rogowski coils, UHF sensors, optical sensors, acoustic emission sensors and digital acquisition cards; the detection bandwidth of the Rogowski coil is 460kHz~120MHz, the maximum allowable current of the optical sensor is 10mA, the photon detection efficiency is 50% when the external 30V bias power supply is connected, the dark count rate is 150kHz / mm2, and the residual pulse is 5%. The UHF sensor adopts a double-helix UHF sensor, and the working bandwidth of the receiving antenna is 300MHz-1.5GHz. The ultrasonic sensor uses a narrow-band acoustic emission sensor with a detection frequency range of 20kHz-180kHz.

[0039] S2: Extract statistical features and use the chi-square test method to reduce the dimension and screen the extracted features.

[0040] Furthermore, the statistical characteristic quantity extraction includes selecting the average discharge amplitude A of the statistical characteristic quantity for the pulse current method, the optical measurement method and the ultra-high frequency method. av , average discharge frequency N, skewness Sk, steepness Ku, phase asymmetry φ, discharge asymmetry Q, cross-correlation factor Cc and ultrasonic signal dispersion entropy DE k ; The average value of all pulse amplitudes in the positive half cycle within the measurement time range of an integral multiple of the power frequency cycle is recorded as A av+ The average value of all pulse amplitudes in the positive half cycle is recorded as A av- , the number of discharge pulses occurring in the positive half cycle within the unit measurement time t is recorded as N +, the number of discharge pulses occurring in the negative half cycle is recorded as N - .

[0041] It should be noted that the pulse current method includes the calculation of the skewness, expressed as:

[0042]

[0043] Among them, A i is the amplitude of the ith discharge pulse, σ is the standard deviation of the discharge pulse amplitude within the statistical range, n is the number of discharge pulses within the statistical range, and the steepness is calculated as:

[0044]

[0045] Calculate the phase asymmetry, expressed as:

[0046] φ=φ in- / φ in+

[0047] Among them, φ in+ is the difference between the positive half-cycle discharge starting phase and 0 degrees, φ in- It is the difference between the starting phase of the negative half-cycle discharge and 180 degrees.

[0048] It should also be noted that the pulse current method also includes the calculation of the cross-correlation factor, which is expressed as:

[0049]

[0050] Among them, x i' is the sum of the discharge pulse amplitudes in the i'th phase window of the positive half cycle, y i' It is the sum of the discharge pulse amplitudes in the i'th phase window in the negative half cycle, and n' is the number of phase windows in the positive and negative half cycles.

[0051] It should also be noted that the optical measurement method includes calculating the dispersion entropy based on the Shannon entropy to quantify the distribution characteristics of the signal in frequency or time, expressed as:

[0052]

[0053] Where n” represents the total number of possible states of the random variable k, p i” represents the probability that the random variable k is in the i'th state.

[0054] It should also be noted that, due to limited training samples, too many features can easily cause overfitting in the model training process, which negatively affects the generalization and accuracy of the partial discharge diagnosis model; therefore, it is also necessary to reduce the dimension of the features and remove redundant features with high correlation to obtain better diagnostic results; for this purpose, the feature screening method of the chi-square test was used for dimensionality reduction. The chi-square test is a widely used hypothesis testing method, which is often used for comparison of multiple rates or multiple components, as well as correlation analysis of classified data. The chi-square test statistics the deviation between the actual observed value of the sample and the theoretical inference value. The larger the chi-square value, the stronger the orthogonality between the variables, and the lower the chi-square value, the stronger the dependence between the variables; therefore, the chi-square value of each feature and the classification label is calculated, that is, the feature value with the score as the screening criterion.

[0055] S3: Use the improved support vector machine algorithm to process the optimized feature set and identify the gas insulation fault type.

[0056] Furthermore, the main idea of ​​the support vector machine algorithm classification is to solve the convex quadratic programming problem for the distribution of samples in the feature space, that is, to find an optimal classification hyperplane so that samples of different categories are exactly located on both sides of the hyperplane, and in the process, the distance between the samples and the hyperplane is made as far as possible; and for multi-classification problems, it can be solved by a combination of multiple binary classification support vector machine models.

[0057] It should be noted that in a binary support vector machine, for a linearly separable data set, there is a separating hyperplane with the largest interval, expressed as:

[0058] ω T x+b=0

[0059] Among them, ω represents the normal vector of the hyperplane, x represents the feature vector, b represents the bias of the hyperplane, T represents the transposition operation, and the classification decision function corresponding to the hyperplane is expressed as:

[0060] f(x)=ω T x+b

[0061] If we divide the data points (x i' ,y i' ) is correctly classified, then the sum of the distances γ from the heterogeneous support vectors to the hyperplane is expressed as:

[0062]

[0063] To find a hyperplane that makes the sample as far away from the hyperplane as possible, the constraints should be met:

[0064]

[0065] y i' (ω T x i' +b)≥1, i'=1,2,…,m

[0066] Among them, m represents the number of training samples, and the constraints that need to be met are:

[0067]

[0068] y i' (ω T x i' +b)≥1, i'=1,2,…,m

[0069] Through the Lagrange multiplier α i' ≥0 solves the constraint conditions, expressed as:

[0070]

[0071] According to the KKT condition, let the partial derivative of L(ω,b,α) with respect to ω and b be zero, expressed as:

[0072]

[0073] In order to reduce the impact of abnormal data points in training samples, allow a certain degree of uncertainty in classification applications, and reduce system losses during training, a slack variable ξ ≥ 0 and a penalty parameter C are introduced, which can be expressed as:

[0074]

[0075] y i' (ω T x i' +b)≥1-ξ i'

[0076] Introducing the Lagrange multiplier μ, it is expressed as:

[0077]

[0078] Solving with the Lagrange multiplier method, we get the expressions for ω and b, which are expressed as:

[0079]

[0080] Obtain the classification hyperplane in the SVM classifier. Since the classification of partial discharge is a nonlinear problem, it is necessary to introduce the kernel function φ(x i' ,x j' ), mapping nonlinear data into linear space, expressed as:

[0081]

[0082] It can be seen that the original feature set has too many parameters, and the overfitting phenomenon in model training is quite obvious; after the chi-square test feature screening determines the optimized feature set, the recognition effect of the model is improved.

[0083] Example 2, reference Figure 2-Figure 3 , which is an embodiment of the present invention, provides a gas insulation fault identification method based on an improved support vector machine algorithm. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0084] Firstly, three typical discharge defect electrode models were prepared, and an experimental system was built on the gas-insulated partial discharge multi-physics experimental platform. Figure 2 This is a diagram of the multi-physics experimental platform for gas-insulated partial discharge. The Rogowski coil, UHF sensor, optical sensor and acoustic emission sensor are used in conjunction with a digital acquisition card to synchronously collect partial discharge signals. The synchronously collected partial discharge signals are analyzed, and statistical feature quantities are selected as input parameters of the partial discharge identification model. The selected statistical feature quantities include: average discharge amplitude, average discharge frequency, skewness, steepness, phase asymmetry, discharge asymmetry, cross-correlation factor and ultrasonic signal dispersion entropy. The feature quantities are extracted separately through different detection methods (pulse current method, optical measurement method, and UHF method), and their sources are distinguished in subsequent analysis. The chi-square test method is used to reduce the dimension of the extracted feature quantities and remove redundant features. Reduce the number of features to prevent model overfitting. Filter out features with strong orthogonality to classification labels through chi-square value calculation to form an optimized feature set. Based on the optimized feature set, use the support vector machine (SVM) algorithm for model training. By finding the optimal classification hyperplane in the feature space, classify fault samples of different categories to ensure that the distance between the samples and the hyperplane is as far as possible, thereby improving the accuracy of recognition. For multi-classification problems, multiple binary support vector machine models are combined to achieve the recognition of gas insulation fault types (corona discharge, surface discharge, and suspended discharge) with an accuracy rate of 96.6%. The severity of partial discharge can also be recognized with an accuracy rate of 92.7%. The effect of the diagnostic model can be seen in Figure 1. Figure 3 .

[0085] Experiments show that the use of an improved support vector machine algorithm combined with the simultaneous collection and analysis of local discharge data from multiple physical sensors can improve the accuracy of identifying internal fault types in gas-insulated electrical appliances. Compared with the traditional single physical model, our invention has reduced the misjudgment rate between corona discharge and suspended discharge, improved the overall recognition effect, and has better practical application value.

[0086] Example 3, reference Figure 4, which is an embodiment of the present invention, provides a gas insulation fault identification system based on an improved support vector machine algorithm, including a signal acquisition module, a feature extraction module, and a fault identification module.

[0087] The signal acquisition module is used to simulate and perform partial discharge tests on typical fault types inside gas-insulated electrical equipment and collect partial discharge signals; the feature extraction module is used to extract statistical feature quantities and use the chi-square test method to reduce the dimension and screen the extracted feature quantities; the fault identification module is used to use the improved support vector machine algorithm to process the optimized feature set and identify the gas insulation fault type.

[0088] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0089] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0090] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0091] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

[0092] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A gas insulation fault identification method based on an improved support vector machine algorithm, characterized in that: include: Simulate typical fault types inside gas-insulated electrical equipment and conduct partial discharge tests to collect partial discharge signals; Extract statistical features, and use the chi-square test method to reduce the dimension and screen the extracted features; The optimized feature set is processed using an improved support vector machine algorithm to identify the gas insulation fault type.

2. The gas insulation fault identification method based on the improved support vector machine algorithm according to claim 1 is characterized in that: The typical fault types include corona discharge defects, surface discharge defects, and suspended discharge defects.

3. The gas insulation fault identification method based on the improved support vector machine algorithm according to claim 2 is characterized in that: The collecting of partial discharge signals includes synchronously collecting partial discharge pulse current signals, ultra-high frequency signals, optical signals and ultrasonic signals through Rogowski coils, ultra-high frequency sensors, optical sensors, acoustic emission sensors and digital acquisition cards; The detection bandwidth of the Rogowski coil is 460kHz~120MHz, the maximum allowable current of the optical sensor is 1OmA, the photon detection efficiency is 50% when the external 30V bias power supply is connected, the dark count rate is 150kHz / mm2, and the residual pulse is 5%. The UHF sensor adopts a double-helix UHF sensor, and the receiving antenna working bandwidth is 300MHz-1.5GHz. The ultrasonic sensor uses a narrow-band acoustic emission sensor with a detection frequency range of 20kHz-180kHz.

4. The gas insulation fault identification method based on the improved support vector machine algorithm according to claim 3 is characterized in that: The extraction of statistical characteristic quantities includes selecting the average discharge amplitude A of the statistical characteristic quantity for the pulse current method, the optical measurement method and the ultra-high frequency method. av , average discharge frequency N, skewness Sk, steepness Ku, phase asymmetry φ, discharge asymmetry Q, cross-correlation factor Cc and ultrasonic signal dispersion entropy DE k ; The average value of all pulse amplitudes in the positive half cycle within the measurement time range of an integral multiple of the power frequency cycle is recorded as A av+ The average value of all pulse amplitudes in the positive half cycle is recorded as A av- , the number of discharge pulses occurring in the positive half cycle within the unit measurement time t is recorded as N + , the number of discharge pulses occurring in the negative half cycle is recorded as N-.

5. The gas insulation fault identification method based on the improved support vector machine algorithm according to claim 4 is characterized in that: The pulse current method involves calculating the skewness, expressed as: Among them, A i is the amplitude of the ith discharge pulse, σ is the standard deviation of the discharge pulse amplitude within the statistical range, n is the number of discharge pulses within the statistical range, and the steepness is calculated as: Calculate the phase asymmetry, expressed as: φ=φi n - / fi n+ Among them, φi n+ is the difference between the positive half-cycle discharge starting phase and 0 degrees, φi n - is the difference between the starting phase of the negative half-cycle discharge and 180 degrees.

6. The gas insulation fault identification method based on the improved support vector machine algorithm according to claim 5, characterized in that: The pulse current method also includes calculating the cross-correlation factor, expressed as: Among them, x i ′ is the sum of the discharge pulse amplitudes in the i′th phase window of the positive half cycle, y i ′ is the sum of the discharge pulse amplitudes in the i′th phase window in the negative half cycle, and n′ is the number of phase windows in the positive and negative half cycles.

7. The gas insulation fault identification method based on the improved support vector machine algorithm according to claim 4 is characterized in that: The optical measurement method includes calculating the dispersion entropy based on the Shannon entropy to quantify the distribution characteristics of the signal in frequency or time, expressed as: Where n″ represents the total number of possible states of the random variable k, p i″ represents the probability that the random variable k is in the i″th state.

8. A system using the gas insulation fault identification method based on the improved support vector machine algorithm as claimed in any one of claims 1 to 7, characterized in that: It includes signal acquisition module, feature extraction module and fault identification module; The signal acquisition module is used to simulate typical fault types inside gas-insulated electrical equipment and perform partial discharge tests to collect partial discharge signals; The feature extraction module is used to extract statistical feature quantities and use the chi-square test method to reduce the dimension and screen the extracted feature quantities; The fault identification module is used to process the optimized feature set using an improved support vector machine algorithm to identify the gas insulation fault type.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the gas insulation fault identification method based on the improved support vector machine algorithm described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the gas insulation fault identification method based on the improved support vector machine algorithm described in any one of claims 1 to 7 are implemented.

Images