Method for identifying characteristics of metal foreign matters in GIS (Geographic Information System)

Abstract

The invention provides a method for identifying the characteristics of a metal foreign body in a GIS, and the method comprises the steps: building an electric field simulation model, and setting the metal foreign body; a multi-modal sensor is adopted to collect signals generated by partial discharge of the metal foreign matter so as to obtain a multi-modal atlas database; performing multi-domain feature extraction on the multi-modal atlas database, and generating a lightweight fingerprint feature vector through feature dimension reduction; the lightweight fingerprint feature vector sequentially passes through all steps of a combinatorial algorithm to solve identification items of the metal foreign matter; and selecting a globally optimal group of identification items to be combined into a metal foreign matter feature identification scheme. According to the invention, fine analysis and processing are carried out on partial discharge signals through a signal processing and pattern recognition algorithm, so that the detection and classification precision of metal foreign matters can be effectively improved; real-time monitoring and quick response under complex working conditions can be realized, a quantitative basis can be provided for fault early warning and equipment maintenance, and the operation reliability and safety of GIS equipment are further enhanced.

Description

Technical Field

[0001] The present invention relates to the field of power technology, and particularly relates to a method for identifying the characteristics of metal foreign objects in GIS. Background Art

[0002] Gas Insulated Switchgear (GIS) equipment, as a key equipment in high-voltage power systems, is widely used in urban power distribution, high-voltage power transmission, substations and other occasions due to its advantages such as small floor area, reliable operation, and simple maintenance.

[0003] However, the problem of metal foreign objects existing inside GIS has always been one of the main hidden dangers affecting its operation safety and reliability. Metal foreign objects may originate from defects in the manufacturing process, wear during equipment operation, misoperations during maintenance, or even intrusion of external environmental factors. Under the action of the complex electric field inside GIS, these metal foreign objects will cause partial discharge phenomena, which will further lead to a decrease in insulation performance and local overheating of the equipment. In severe cases, it may even trigger safety accidents such as equipment failures or fires.

[0004] Currently, the detection of metal foreign objects inside GIS mainly relies on partial discharge monitoring technologies, such as Ultra High Frequency (UHF) sensors, ultrasonic sensors, and partial discharge detectors. These devices attempt to judge the quantity and location of metal foreign objects by collecting and analyzing partial discharge signals. However, due to the complex electric field distribution inside GIS, the partial discharge signals caused by metal foreign objects are often diverse and complex. Traditional signal analysis methods face great challenges in identifying metal foreign objects of different types, positions, and quantities. Furthermore, there are many deficiencies in the existing technologies in terms of identification accuracy, real-time monitoring ability, and feature extraction and processing methods, which cannot fully reflect the complex characteristics of metal foreign objects, resulting in frequent misjudgments and missed judgments, and it is difficult to meet the requirements of modern power systems for efficient and intelligent equipment management.

[0005] In summary, there is still a need to develop a method that can achieve refined identification of metal foreign objects inside GIS to improve the identification accuracy and real-time monitoring ability. Summary of the Invention

[0006] The present invention aims at the problems existing in the prior art, and provides an effective and rapid method for identifying the characteristics of metal foreign objects in GIS equipment, which is used to detect the type, position, and quantity of metal foreign objects.

[0007] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0008] The present invention provides a method for identifying the characteristics of metal foreign objects in GIS, which includes the following steps:

[0009] A three-dimensional electric field model of GIS equipment is established based on the finite element method, and the electric field distortion area is determined through numerical simulation. Metal foreign objects of different sizes are set in the electric field distortion area;

[0010] Multimodal sensors are used to collect the signals generated by partial discharge of the metal foreign objects to obtain a multimodal atlas library;

[0011] Multidomain feature extraction is performed on the multimodal atlas library, and lightweight fingerprint feature vectors are generated through feature dimensionality reduction; the lightweight fingerprint feature vectors are successively passed through an ensemble learning algorithm based on decision trees, an iterative optimization algorithm, and a fuzzy logic algorithm to solve the identification items of the metal foreign objects; the identification items include the type of metal foreign object, the location of the metal foreign object, and the number of metal foreign objects;

[0012] Through the adaptive test sorting method, the sequential combinations of the identification items are exhausted, and a globally optimal set is selected as the identification scheme for the metal foreign object features.

[0013] Optionally, the calculation of the electric field distortion area includes the following steps:

[0014] The surface distribution curve of the electric field is calculated through numerical simulation, and the location with the maximum electric field distortion is solved;

[0015] The electric field distortion area includes the middle part of the insulator, low potential, high potential, and the location with the maximum electric field distortion.

[0016] Optionally, the multimodal sensors include ultra-high frequency sensors, ultrasonic sensors, and partial discharge detectors.

[0017] Optionally, the multidomain feature extraction includes the following steps:

[0018] The partial discharge data corresponding to different types of metal foreign objects are selected from the multimodal atlas library, and fingerprint feature vectors are extracted through time-domain and frequency-domain signal processing algorithms;

[0019] The extracted fingerprint feature vectors are reversibly reduced in order through the feature dimensionality reduction to generate the lightweight fingerprint feature vectors.

[0020] Optionally, the extraction of the fingerprint feature vectors includes the following steps:

[0021] Based on the time-domain signal of the partial discharge data, the pulse characteristics of the discharge signal are extracted;

[0022] Based on the partial discharge data, wavelet transform is used for multi-scale signal decomposition to extract local energy features;

[0023] The time-domain signal of the partial discharge data is converted into a frequency-domain signal through Fourier transform to extract spectral features;

[0024] Combine the pulse characteristics, local energy characteristics, and the spectral characteristics to obtain the fingerprint feature vector, and label and file the fingerprint feature vector.

[0025] Optionally, the algorithm for feature dimensionality reduction is the RST algorithm.

[0026] Optionally, the ensemble learning algorithm based on decision trees adopts the random forest algorithm. By constructing multiple decision trees, and each decision tree is trained based on different training subsets, and finally the first result is determined by voting.

[0027] Optionally, the iterative optimization algorithm is the repeated clipping algorithm. Based on the first result, multiple weak classifiers are iteratively trained to output the second result.

[0028] Optionally, the fuzzy logic algorithm is the fuzzy recognition algorithm. Based on the second result, the result features are mapped into a fuzzy set, and the third result is output through a fuzzy inference system.

[0029] Optionally, the obtaining of the recognition scheme includes the following steps:

[0030] Set the globally optimal evaluation metrics as accuracy and response speed;

[0031] Traverse each set of exhaustive results in a simulation scenario, and obtain the evaluation metrics respectively;

[0032] Solve the globally optimal sequential combination through a multi-objective optimization algorithm, and output the globally optimal sequential combination as the recognition scheme.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] Through signal processing and pattern recognition algorithms, the present invention conducts refined analysis and processing on partial discharge signals, which can effectively improve the detection and classification accuracy of metal foreign objects; it can not only achieve real-time monitoring and rapid response under complex working conditions, but also provide a quantitative basis for fault warning and equipment maintenance, further enhancing the operation reliability and safety of GIS equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] 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 in the embodiments. Obviously, the drawings in the following description 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.

[0036] Figure 1 It is the method flow chart in a specific embodiment of the present invention;

[0037] Figure 2 It is the flowchart for obtaining the recognition scheme in a specific embodiment of the present invention. Specific Embodiments

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0039] It should be noted that similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0040] It is worth noting that the methods used in the present invention are all conventional methods unless otherwise specified; the raw materials and devices used are all conventional commercially available products unless otherwise specified, and their sources are not specifically limited.

[0041] It also needs to be noted that in the specific embodiments of the present invention, for the convenience of understanding, the method steps are described in a certain order. However, those skilled in the art can adjust the order of the steps according to actual needs. Therefore, this cannot be used as a limiting condition. Further, in the description of the following specific embodiments, unless otherwise specified, the upper and lower subscripts of each parameter should be understood as the difference marks of similar identifiers according to common interpretations, representing the parameters of the components related to the subscripts or corresponding to each other, and cannot be understood as specific models or special marks.

[0042] Please refer to Figure 1 and Figure 2 As shown, a method for identifying the characteristics of metal foreign objects in GIS needs to improve the calculation method of existing detection means. Therefore, the method is improved under simulated test conditions. Thus, a three-dimensional electric field model of GIS equipment is established based on the finite element method, and the electric field distortion region is determined through numerical simulation. Specifically, the electric field of the GIS pot insulator is numerically simulated, and the finite element analysis software is used to simulate the electric field distribution.

[0043] Metal foreign objects of different sizes are set in the electric field distortion region. Optionally, the calculation steps of the electric field distortion region in this embodiment are as follows:

[0044] Calculate the surface distribution curve of the electric field through numerical simulation and solve for the location with the maximum electric field distortion. Through formula analysis, it is found that there are two positions with the most severe electric field distortion along the surface electric field curve. Among them, the electric field in the middle is relatively stable, and the electric field drops significantly at both ends. The electric field calculation formula is simplified as:

[0045]

[0046] In the formula, E(x) is the electric field strength at position x, E0 is the initial electric field strength, and L is the length of the insulator.

[0047] Therefore, the electric field distortion region includes the middle of the insulator, low potential, high potential, and the location with the maximum electric field distortion. At the above key positions, the middle of the insulator, and the three combination points of the ground potential and high potential, metal particles are arranged for experiments according to the requirements of different lengths and positions from the micron level to the millimeter level. These metal particles simulate various possible metal foreign object situations to ensure the comprehensiveness and representativeness of the experimental data.

[0048] Use a multi-modal sensor to collect the signals generated by partial discharge of metal foreign objects to obtain a multi-modal atlas library. Optionally, the multi-modal sensor includes an ultra-high frequency sensor, an ultrasonic sensor, and a partial discharge detector. The ultra-high frequency sensor, ultrasonic sensor, and partial discharge detector are used to synchronously collect experimental data to ensure the diversity and accuracy of the data. Through multiple repeated experiments, the partial discharge signal data caused by different types, positions, and quantities of metal foreign objects are obtained, and a partial discharge multi-modal atlas library containing various discharge signal types is formed.

[0049] Through the above steps, a comprehensive atlas library covering partial discharge signals caused by different types, positions, and quantities of metal foreign objects is constructed, providing a rich data basis for subsequent feature extraction and recognition.

[0050] Perform multi-domain feature extraction on the multi-modal atlas library. The multi-domain feature extraction includes the following steps:

[0051] Select the partial discharge data corresponding to different types of metal foreign objects from the multi-modal atlas library, and extract fingerprint feature vectors through time-domain and frequency-domain signal processing algorithms. The steps are as follows:

[0052] (1) Time-domain feature extraction;

[0053] Based on the time-domain signal of the partial discharge data, extract the pulse characteristics of the discharge signal. Specifically, use the time-domain signal analysis method to extract the pulse characteristics of the discharge signal. For example, calculate the peak voltage V peak and the pulse width T pulse :

[0054] V peak = max(V(t));

[0055] T pulse = t end - t start ;

[0056] Where V(t) is the discharge signal, t end is the signal end time, and t start is the signal start time. These time-domain features can reflect the basic pulse form of the partial discharge signal and provide a preliminary basis for identification.

[0057] (2) Wavelet transform;

[0058] Based on the partial discharge data, wavelet transform is used for multi-scale signal decomposition to extract local energy features; specifically, wavelet transform is used for multi-scale signal decomposition to extract local energy features. The wavelet transform formula is as follows:

[0059]

[0060] Where a is the scale parameter, b is the translation parameter, and ψ(t) is the mother wavelet. Through wavelet coefficients at different scales, the energy distribution of the signal at different frequencies can be captured.

[0061] (3) Fourier transform;

[0062] The time-domain signal of the partial discharge data is converted into a frequency-domain signal through Fourier transform to extract spectral features; specifically, the time-domain signal is converted into a frequency-domain signal through Fourier transform to extract spectral features. For example, calculate the spectral density S(f) of the signal:

[0063]

[0064] The spectral features reflect the frequency components of the signal and can identify the frequency feature differences caused by different metal foreign objects.

[0065] (4) Fingerprint feature vector;

[0066] The pulse characteristics, local energy features, and spectral features are combined to obtain the fingerprint feature vector, and the fingerprint feature vector is labeled and archived. Specifically, combining the time-domain and frequency-domain features, a comprehensive fingerprint feature vector F is formed, and its dimension is 128:

[0067] F = [F1, F2, …, F 128 ;

[0068] Where F i represents different time-domain and frequency-domain feature parameters. The comprehensive feature vector can more comprehensively describe the characteristics of the partial discharge signal and improve the accuracy of subsequent identification.

[0069] The extracted fingerprint feature vector is reversibly reduced in order through feature dimension reduction to generate a lightweight fingerprint feature vector; optionally, the algorithm for feature dimension reduction is the RST (rough set) algorithm. The RST algorithm is used to reduce the order of the extracted fingerprint features and construct an information system:

[0070] S = (U, A ∪ {d});

[0071] In the formula, U is the object set, A is the attribute set, and {d} is the decision attribute. By calculating the positive region of the attribute subset :

[0072]

[0073] And select to satisfy:

[0074] POS B (d) = POS A (d);

[0075] Thus, a lightweight fingerprint feature vector is generated through feature dimension reduction. The lightweight fingerprint feature vector is the minimum attribute subset B; a feature set B with a lower operation dimension is obtained, thereby reducing the computational complexity and improving the processing efficiency.

[0076] The lightweight fingerprint feature vector passes through an ensemble learning algorithm based on a decision tree, an iterative optimization algorithm, and a fuzzy logic algorithm in sequence to solve the recognition items of metallic foreign objects. Optionally, the recognition items include the type of metallic foreign object, the location of the metallic foreign object, and the quantity of the metallic foreign object. Since the order of the sequence of solutions for each item can be adjusted during the solution process, and different solution accuracies and response speeds are generated, the solution sequence of type - location - quantity is described first.

[0077] First, the ensemble learning algorithm based on a decision tree adopts the random forest algorithm. By constructing multiple decision trees, and each decision tree is trained based on different training subsets, and finally, the voting method is used to determine the first result. Specifically, in this embodiment, based on the reduced - order feature set, the random forest algorithm is used for preliminary classification. This algorithm constructs multiple decision trees {h1(x), h2(x), …, h N (x)} and uses the voting method to determine the first result, that is, the type H(x) of the metallic foreign object:

[0078] H(x) = mode{h1(x), h2(x), …, h N (x)};

[0079] The above - mentioned method can improve the classification accuracy and robustness, and effectively reduce the over - fitting phenomenon.

[0080] After that, the iterative optimization algorithm is the Boosting algorithm. Based on the first result, multiple weak classifiers are iteratively trained to output the second result. Specifically, in combination with the Boosting algorithm, the position of the metal foreign object is accurately located. Boosting gradually improves the overall classification performance by iteratively training multiple weak classifiers.

[0081] Its core arithmetic formula is:

[0082]

[0083] In the formula, ∈ t is the error rate of the t-th classifier, α t is the classifier weight, w i is the sample weight, y i is the true label of the sample, h t (x i ) is the prediction result of the t-th classifier. Taking the prediction result as the second result, that is, the positioning of the metal foreign object.

[0084] Finally, the fuzzy logic algorithm is the fuzzy recognition algorithm. Based on the second result, the result features are mapped into a fuzzy set, and the third result is output through a fuzzy inference system. Specifically, the fuzzy recognition algorithm is used to estimate the number of metal foreign objects. Fuzzy recognition is based on fuzzy logic rules, maps the quantity features into a fuzzy set, and outputs the quantity estimation result through a fuzzy inference system.

[0085] The fuzzy logic inference arithmetic formula is as follows:

[0086] μ output (z) = sup{min(μ A (x), μ B (y)) | z = f(x, y)};

[0087] In the formula, μ A (x) and μ B (y) are the membership functions of the input variables respectively, and f(x, y) is the inference function of the fuzzy rule.

[0088] Thus, through the above calculation process, the type of metal foreign object, the positioning of the metal foreign object, and the number of metal foreign objects are output layer by layer in sequence.

[0089] As described above, under the condition that the algorithm sequence remains unchanged, the accuracy and response speed can be improved by changing the content to be solved. Therefore, in order to obtain a better detection scheme, in this embodiment, the adaptive test sorting method is used to exhaustively list the order combinations of the recognition items, that is, for different signals collected, multiple recognition order schemes are designed and tested, such as the first result - the second result - the third result corresponding to the solution items being "type - position - quantity" and "quantity - type - position", etc. The necessity of designing with different hierarchical orders and selecting different recognition algorithms for each layer lies in that different hierarchical orders can comprehensively analyze the complex characteristics of partial discharge signals from multiple perspectives, and selecting the most suitable recognition algorithm for each layer's characteristics helps to give full play to the advantages of various algorithms in dealing with different signal characteristics, so as to achieve precise positioning, classification, and quantity estimation of metal foreign objects.

[0090] Among them, the globally optimal evaluation indicators are set as accuracy and response speed; in the simulation scenario, each set of exhaustive results is traversed, and the evaluation indicators are obtained respectively. By recording the recognition accuracy and recognition speed of each scheme, the optimal scheme with the highest recognition accuracy and the fastest response speed is finally selected.

[0091] Since there are two evaluation indicators in this embodiment, the method for solving the optimal solution can solve the globally optimal order combination through a multi-objective optimization algorithm and output the globally optimal order combination as the recognition scheme; however, in order to further simplify the solution steps, this embodiment optimizes the recognition order by constructing a correlation formula, and the formula is:

[0092]

[0093] In the formula, sequence is the set of order combinations, Accuracy is the accuracy of the order combination, and Time is the corresponding time of the order combination.

[0094] Through statistical analysis, the optimal recognition order and algorithm combination are determined to ensure the best recognition effect and response speed in practical applications, and this combination is selected as the recognition scheme output for the characteristics of metal foreign objects, and then deployed in the actual measurement link. When the corresponding data is collected in the actual measurement, the calculation process of this group is used to identify metal foreign objects, and the final calculation result is output as the recognition result of the actual measurement (including the type, position, and quantity of metal foreign objects).

[0095] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, rather than a limitation on the protection scope of the present invention. Any simple modification or equivalent replacement made by those of ordinary skill in the art to the technical solution of the present invention does not depart from the essence and scope of the technical solution of the present invention.

Claims

1. A method for identifying the characteristics of metal foreign objects in GIS, characterized in that: It includes the following steps: Based on the finite element method, establish a three-dimensional electric field model of the GIS device, determine the electric field distortion area through numerical simulation, and set metal foreign objects of different sizes in the electric field distortion area; Use a multi-modal sensor to collect the signals generated by the partial discharge of the metal foreign object to obtain a multi-modal atlas library; Extract multi-domain features from the multi-modal atlas library, and generate a lightweight fingerprint feature vector through feature dimensionality reduction; the lightweight fingerprint feature vector sequentially passes through an ensemble learning algorithm based on a decision tree, an iterative optimization algorithm, and a fuzzy logic algorithm to solve the identification items of the metal foreign object; the identification items include the type of metal foreign object, the location of the metal foreign object, and the number of metal foreign objects; Through the adaptive test sorting method, exhaustively list the sequential combinations of the identification items, and select the globally optimal set as the identification scheme for the characteristics of the metal foreign object.

2. The method for identifying the characteristics of metallic foreign objects in GIS according to claim 1, wherein: The calculation of the electric field distortion area includes the following steps: Calculate the surface distribution curve of the electric field through numerical simulation, and solve the location with the maximum electric field distortion; The electric field distortion area includes the middle part of the insulator, low potential, high potential, and the location with the maximum electric field distortion.

3. The method for identifying the characteristics of metallic foreign objects in GIS according to claim 1, wherein: The multi-modal sensor includes a UHF sensor, an ultrasonic sensor, and a partial discharge detector.

4. The method for identifying the characteristics of metallic foreign objects in GIS according to claim 3, characterized in that: The multi-domain feature extraction includes the following steps: Select the partial discharge data corresponding to different types of metal foreign objects from the multi-modal atlas library, and extract the fingerprint feature vector through time-domain and frequency-domain signal processing algorithms; Perform reversible order reduction processing on the extracted fingerprint feature vector through the feature dimensionality reduction to generate the lightweight fingerprint feature vector.

5. The method for identifying the characteristics of metallic foreign objects in GIS according to claim 4, characterized in that: The extraction of the fingerprint feature vector includes the following steps: Extract the pulse characteristics of the discharge signal based on the time-domain signal of the partial discharge data; Perform multi-scale signal decomposition on the partial discharge data using wavelet transform to extract local energy characteristics; Convert the time-domain signal of the partial discharge data into a frequency-domain signal through Fourier transform to extract spectral characteristics; Combine the pulse characteristics, local energy characteristics, and the spectral characteristics to obtain the fingerprint feature vector, and label and file the fingerprint feature vector.

6. The method for identifying the characteristics of metal foreign objects in GIS according to any one of claims 1-5, characterized in that: The algorithm for feature dimensionality reduction is the RST algorithm.

7. The method for identifying the characteristics of metallic foreign objects in GIS according to claim 1, characterized in that: The ensemble learning algorithm based on a decision tree uses the random forest algorithm. By constructing multiple decision trees, and each decision tree is trained based on different training subsets, and finally a voting method is used to determine the first result.

8. The method for identifying the characteristics of metallic foreign objects in GIS according to claim 7, characterized in that: The iterative optimization algorithm is the repeated clipping algorithm. Based on the first result, multiple weak classifiers are trained iteratively to output the second result.

9. The method for identifying the characteristics of metallic foreign objects in GIS according to claim 8, wherein: The fuzzy logic algorithm is the fuzzy recognition algorithm. Based on the second result, the result features are mapped into a fuzzy set, and a fuzzy inference system is used to output the third result.

10. The method for identifying the characteristics of metallic foreign objects in GIS according to claim 1, wherein: The acquisition of the identification scheme includes the following steps: Set the globally optimal evaluation indicators as accuracy and response speed; Traverse each set of exhaustive results in the simulation scenario and obtain the evaluation indicators respectively; Solve the globally optimal sequential combination through a multi-objective optimization algorithm, and output the globally optimal sequential combination as the identification scheme.

Images