Method for evaluating health state of GIS (Gas Insulated Switchgear) equipment

Abstract

The invention relates to a GIS equipment health state assessment method, and belongs to the technical field of feature extraction and analysis. The method comprises the following steps: firstly, setting scoring intervals of four health states of the GIS equipment, and then generating a historical mixed map of four historical health states of each component according to historical parameters in a historical detection report of the GIS equipment. Acquiring historical scores of four historical health states of each component, and inputting the historical scores and all historical mixed maps of all the components into the CKAN model for training; and finally, collecting real-time parameter information of the components to obtain a real-time mixed map, inputting the real-time mixed map into the trained ckan model which can extract key features highly associated with the health state, and reconstructing a key feature mixed map of each key component. And finally, respectively calculating the similarity between the key feature mixed map of each component and the map when the historical health state of each component is normal and serious to obtain a real-time score, and comparing the score with a score interval to judge the health state of each component.

Description

Technical Field

[0001] The invention relates to a method for evaluating the health status of GIS equipment, and belongs to the technical field of feature extraction and analysis (G06F18 / 20). Background Art

[0002] With the development of the national economy and industry and commerce, the demand for electric energy in all walks of life is increasing, and power transmission and transformation equipment needs to be continuously updated. At present, GIS equipment has the characteristics of compact structure and small footprint, which is particularly in line with the needs of the current new power system. However, GIS equipment contains a variety of key components, such as disconnectors, grounding switches, circuit breakers, current transformers and voltage transformers, which play an important role in the safe operation of the power system. The failure modes of different key components have different impacts on the operation of the equipment and the frequency of failures. Therefore, how to accurately and reasonably evaluate the health status of GIS equipment is crucial to improving the safety of power grid operation. Summary of the invention

[0003] The technical problem to be solved by the present invention is to achieve accurate evaluation of the health status of GIS equipment.

[0004] The technical solution proposed by the present invention to solve the above technical problems is: a method for evaluating the health status of GIS equipment, comprising the following steps:

[0005] Step 1: Set the score ranges corresponding to the four health states of the GIS equipment as a whole and its key components as follows:

[0006] Normal [0, 0.25],

[0007] Note that (0.25, 0.50],

[0008] Abnormal (0.50, 0.75],

[0009] severe (0.75, 1.00];

[0010] Step 2: Collect historical electrical quantities, historical mechanical quantities, historical environmental quantities, and historical appearance images of each key component in four historical health states from the historical inspection reports of the GIS equipment;

[0011] Convert the historical electrical quantity, historical mechanical quantity and historical environmental quantity of each key component in the four historical health states into the electrical characteristic map, mechanical characteristic map and environmental characteristic map of each key component in the four historical health states respectively;

[0012] The historical electrical characteristic maps, historical mechanical characteristic maps, historical environmental characteristic maps, and historical appearance photographed images of each key component in the four historical health states are fused to generate a historical mixed map of each key component in the four historical health states;

[0013] Step 3: Use word2vec to obtain the historical scores of the four health states of all key components;

[0014] Step 4: The historical mixed graphs of all key components in the four historical health states obtained in step 2 and the historical scores of all key components in the four historical health states obtained in step 3 are input into the CKAN model for training;

[0015] Step 4.1: Set the initial hyperparameters of the CKAN model;

[0016] Step 4.2: The CKAN model extracts the first feature point of the historical mixed graph of each key component under the four historical health states according to the initial hyperparameters, and generates a prediction score of each key component under the four historical health states based on the extracted first feature point;

[0017] Step 4.3: Calculate the loss function value of the predicted scores and historical scores of the four historical health states of all key components according to equations (1), (2), and (3):

[0018] (1);

[0019] In formula (1), m is the number of key components of GIS equipment; i is the key component;

[0020] It is the key component i history of the historical scores of the four health states and;

[0021] It is the key component i the predicted score of the four historical health states and;

[0022] (2);

[0023] In formula (2), is the mean of the historical scores of the four health states of all key components;

[0024] (3);

[0025] In formula (3), =2;

[0026] Step 4.4: Use the SWO algorithm to continuously iterate and update the initial hyperparameters, and repeat steps 4.2-4.3 using the new hyperparameters generated in each iteration as the initial hyperparameters;

[0027] Step 4.5: Stop the iteration after reaching the number of iterations. The hyperparameter with the smallest value is input into the CKAN model to complete the training;

[0028] Step 5: Collect real-time electrical quantities, real-time mechanical quantities, real-time environmental quantities and real-time appearance images of all key components in the GIS equipment;

[0029] Convert the real-time electrical quantity, real-time mechanical quantity and real-time environmental quantity of each key component into a real-time electrical characteristic map, a real-time mechanical characteristic map and a real-time environmental characteristic map respectively;

[0030] The real-time electrical characteristic map, real-time mechanical characteristic map, real-time environmental characteristic map, and real-time appearance shooting image of each key component are respectively fused to obtain a real-time mixed map of each key component;

[0031] Step 6: Input the real-time mixed maps of all key components into the CKAN model trained in step 4. The CKAN model constructs the key feature mixed maps of the corresponding key components according to the second feature points extracted from the real-time mixed maps of each key component. ;

[0032] Step 7: Calculate the real-time score of the health status of each key component in the GIS equipment according to formula (4)

[0033] (4);

[0034] In formula (1), j is the health status, j=1,2;

[0035] When j=1, It is the historical mixed spectrum when the historical health status of key component i is normal;

[0036] When j=2, It is the historical mixed spectrum when the historical health status of key component i is severe;

[0037] is the key feature mixed map of key component i;

[0038] Real-time scoring of the health status of each key component Compare with the four scoring value intervals in step 1 to determine which health status each key component belongs to.

[0039] Further, after step 7, the following steps are performed:

[0040] Step 8: Calculate the weight of each key component in the same health state according to equations (5) and (6).

[0041] (5);

[0042] (6);

[0043] Step 9: Calculate the information entropy of the key components of GIS equipment in normal and severe health states according to formula (7).

[0044] (7);

[0045] Step 10: Calculate the weights of the key components of GIS equipment in normal and serious health states according to formula (8)

[0046] (8);

[0047] Step 11: Calculate the overall score of GIS equipment according to formula (9)

[0048] (9);

[0049] Compare the overall score Z of the GIS equipment with the four score value intervals in step 1 to determine the overall health status of the GIS equipment.

[0050] Furthermore, the hyperparameters are learning rate, batch size and regularization parameter.

[0051] The beneficial effects of the present invention are as follows: since the CKAN model is trained by utilizing the historical mixed maps and historical scores of all key components in the four historical health states, the trained CKAN model extracts important features that are strongly correlated with the health state in the mixed maps, thereby obtaining more detailed and valuable information about key components, effectively reducing the impact of extreme data on the prediction results, reducing the prediction error, and improving the robustness and accuracy of the model; in addition, by fusing the maps generated by electrical quantities, environmental quantities, and mechanical quantities with their respective appearance photographed images, the formed mixed map can fully retain all information of key parts. This comprehensive perspective makes the prediction more accurate and can better reflect the actual operating conditions of the equipment. Using the fused mixed map as data input can significantly improve the accuracy of the prediction model for the health state of key components. By analyzing these comprehensive information, the prediction model can more effectively identify potential problems and provide more reliable prediction scores. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The GIS equipment health status evaluation method of the present invention is further described below with reference to the accompanying drawings.

[0053] Figure 1 It is a real-time electrical quantity characteristic map of the pot-type insulator of the GIS equipment in the embodiment of the method for evaluating the health status of the GIS equipment.

[0054] Figure 2 It is a real-time mechanical quantity characteristic map of the pot-type insulator of the GIS equipment in the evaluation method of the health status of the GIS equipment in the embodiment.

[0055] Figure 3 It is a real-time environmental quantity characteristic map of the pot-type insulator of the GIS equipment in the evaluation method of the health status of the GIS equipment in the embodiment.

[0056] Figure 4 It is a real-time mixed map of the pot insulator of the GIS equipment in the method for evaluating the health status of the GIS equipment in the embodiment.

[0057] Figure 5 It is a mixed map of key features of pot-type insulators of GIS equipment in the method for evaluating the health status of GIS equipment in the embodiment.

[0058] Figure 6 This is the ckan model structure in the method for evaluating the health status of GIS equipment in the embodiment. DETAILED DESCRIPTION

[0059] Example 1

[0060] A method for evaluating the health status of a GIS device in this embodiment includes the following steps:

[0061] Step 1: Set the score ranges corresponding to the four health states of the GIS equipment as a whole and its key components as follows:

[0062] Normal [0, 0.25],

[0063] Note that (0.25, 0.50],

[0064] Abnormal (0.50, 0.75],

[0065] Severe (0.75, 1.00].

[0066] Step 2: Collect historical electrical quantities, historical mechanical quantities, historical environmental quantities, and historical appearance images of each key component in four historical health states from the historical inspection reports of the GIS equipment;

[0067] Convert the historical electrical quantity, historical mechanical quantity and historical environmental quantity of each key component in the four historical health states into the electrical characteristic map, mechanical characteristic map and environmental characteristic map of each key component in the four historical health states respectively;

[0068] This embodiment uses the existing technology Gram angle field to convert the extracted electrical quantity, mechanical quantity and environmental quantity into characteristic spectra respectively;

[0069] The historical electrical characteristic maps, historical mechanical characteristic maps, historical environmental characteristic maps, and historical appearance photographed images of each key component in the four historical health states are fused to generate a historical mixed map of each key component in the four historical health states;

[0070] This embodiment adopts the existing technology of Co-attention of Multi-view (CAMV) to fuse the three historical feature maps with the historical appearance shooting images to construct a historical mixed map.

[0071] Step 3: Use word2vec to obtain the historical scores of the four health states of all key components;

[0072] Word2Vec is an existing technology that maps words into vector space. It can capture the semantic relationship between words, map each health status name into a vector, and then map the vector of the health status name to a specific historical score.

[0073] Step 4: The historical mixed graphs of all key components in the four historical health states obtained in step 2 and the historical scores of all key components in the four historical health states obtained in step 3 are input into the CKAN model for training;

[0074] like Figure 6 As shown in the figure, the CKAN model is an existing convolutional neural network model for graph processing, which aims to improve the effect of feature extraction by aligning the convolution kernels. It optimizes the alignment of its convolution kernels by optimizing the hyperparameters of the CKAN model to better capture the key features in the graph, reduce the features of the spatial dimension to a smaller dimension, and then flatten the feature map, converting it into a one-dimensional feature vector, which is mapped to the specific historical ratings;

[0075] Step 4.1: Set the initial hyperparameters of the CKAN model; the initial hyperparameters are learning rate, batch size, and regularization parameters;

[0076] Step 4.2: The CKAN model extracts the first feature point of the historical mixed graph of each key component under the four historical health states according to the initial hyperparameters, and generates a prediction score of each key component under the four historical health states based on the extracted first feature point;

[0077] Step 4.3: Calculate the loss function value of the predicted scores and historical scores of the four historical health states of all key components according to equations (1), (2), and (3):

[0078] (1);

[0079] In formula (1), m is the number of key components of GIS equipment; i is the key component;

[0080] It is the key component i history of the historical scores of the four health states and;

[0081] It is the key component i the predicted score of the four historical health states and;

[0082] (2);

[0083] In formula (2), is the mean of the historical scores of the four health states of all key components;

[0084] (3);

[0085] In formula (3), =2;

[0086] Step 4.4: Use the SWO algorithm to continuously iterate and update the initial hyperparameters, and repeat steps 4.2-4.3 using the new hyperparameters generated in each iteration as the initial hyperparameters;

[0087] Step 4.5: Stop the iteration after reaching the number of iterations. The hyperparameters with the smallest value are input into the CKAN model to complete the training.

[0088] Step 5: Collect real-time electrical quantities, real-time mechanical quantities, real-time environmental quantities and real-time appearance images of all key components in the GIS equipment;

[0089] Convert the real-time electrical quantity, real-time mechanical quantity and real-time environmental quantity of each key component into a real-time electrical characteristic map, a real-time mechanical characteristic map and a real-time environmental characteristic map respectively;

[0090] In this embodiment, three real-time characteristic graphs of one of the key components (pot insulator) of GIS equipment are as follows: Figure 1 , 2 , as shown in 3;

[0091] The real-time electrical characteristic map, real-time mechanical characteristic map, real-time environmental characteristic map, and real-time appearance shooting image of each key component are respectively fused to obtain a real-time mixed map of each key component;

[0092] In this embodiment, the real-time mixed map of one of the key components of the GIS equipment (pot insulator) is as follows: Figure 4 shown.

[0093] Step 6: Input the real-time mixed maps of all key components into the CKAN model trained in step 4. The CKAN model constructs the key feature mixed maps of the corresponding key components according to the second feature points extracted from the real-time mixed maps of each key component. ;

[0094] In the embodiment, the key feature hybrid map of one of the key components (pot insulator) of the GIS equipment is as follows: Figure 5 As shown;

[0095] Step 7: Calculate the real-time score of the health status of each key component in the GIS equipment according to formula (4)

[0096] (4);

[0097] In formula (1), j is the health status, j=1,2;

[0098] When j=1, It is the historical mixed spectrum when the historical health status of key component i is normal;

[0099] When j=2, It is the historical mixed spectrum when the historical health status of key component i is severe;

[0100] is the key feature mixed map of key component i;

[0101] Real-time scoring of the health status of each key component Compare with the four scoring value intervals in step 1 to determine which health state each key component belongs to;

[0102] Real-time scoring of the health status of each key component in this embodiment The health status is shown in Table 1 below.

[0103] Key Components Real-time scoring Health status breaker 0.22 normal Closing resistance 0.21 normal Disconnector 0.19 normal Earthing switch 0.18 normal Current transformer 0.17 normal lightning arrester 0.16 normal Voltage transformer 0.15 normal Pot insulator 0.19 normal Mechanical structure 0.15 normal

[0104] Table 1

[0105] Example 2

[0106] The method for evaluating the health status of a GIS device in this embodiment is basically the same as that in Embodiment 1, except that after step 7, the following steps are performed:

[0107] Step 8: Calculate the weight of each key component in the same health state according to equations (5) and (6).

[0108] (5);

[0109] (6).

[0110] Step 9: Calculate the information entropy of the key components of GIS equipment in normal and severe health states according to formula (7).

[0111] (7).

[0112] Step 10: Calculate the weights of the key components of GIS equipment in normal and serious health states according to formula (8)

[0113] (8).

[0114] Step 11: Calculate the overall score of GIS equipment according to formula (9)

[0115] (9);

[0116] Compare the overall score Z of the GIS equipment with the four score value intervals in step 1 to determine which health state the GIS equipment as a whole belongs to;

[0117] The overall score of GIS equipment calculated in this embodiment =0.19, the health status is healthy.

[0118] The above description is only a preferred embodiment of the present invention, but the present invention is not limited thereto. All equivalent replacements or equivalent changes of the concepts and technical solutions according to the present invention should be included in the protection scope of the present invention.

Claims

1. A method for evaluating the health status of GIS equipment, characterized in that The steps include: Step 1: Set the score ranges corresponding to the four health states of the GIS equipment as a whole and its key components as follows: Normal [0, 0.25], Note that (0.25, 0.50], Abnormal (0.50, 0.75], severe (0.75, 1.00]; Step 2: Collect historical electrical quantities, historical mechanical quantities, historical environmental quantities, and historical appearance images of each key component in four historical health states from the historical inspection reports of the GIS equipment; Convert the historical electrical quantity, historical mechanical quantity and historical environmental quantity of each key component in the four historical health states into the electrical characteristic map, mechanical characteristic map and environmental characteristic map of each key component in the four historical health states respectively; The historical electrical characteristic maps, historical mechanical characteristic maps, historical environmental characteristic maps, and historical appearance photographed images of each key component in the four historical health states are fused to generate a historical mixed map of each key component in the four historical health states; Step 3: Use word2vec to obtain the historical scores of the four health states of all key components; Step 4: The historical mixed graphs of all key components in the four historical health states obtained in step 2 and the historical scores of all key components in the four historical health states obtained in step 3 are input into the CKAN model for training; Step 4.1: Set the initial hyperparameters of the CKAN model; Step 4.2: The CKAN model extracts the first feature point of the historical mixed graph of each key component under the four historical health states according to the initial hyperparameters, and generates a prediction score of each key component under the four historical health states based on the extracted first feature point; Step 4.3: Calculate the loss function value of the predicted scores and historical scores of the four historical health states of all key components according to equations (1), (2), and (3): （1）； In formula (1), m is the number of key components of GIS equipment; i is the key component; It is the key component i history of the historical scores of the four health states and; It is the key component i the predicted score of the four historical health states and; （2）； In formula (2), is the mean of the historical scores of the four health states of all key components; （3）； In formula (3), =2; Step 4.4: Use the SWO algorithm to continuously iterate and update the initial hyperparameters, and repeat steps 4.2-4.3 using the new hyperparameters generated in each iteration as the initial hyperparameters; Step 4.5: Stop the iteration after reaching the number of iterations. The hyperparameter with the smallest value is input into the CKAN model to complete the training; Step 5: Collect real-time electrical quantities, real-time mechanical quantities, real-time environmental quantities and real-time appearance images of all key components in the GIS equipment; Convert the real-time electrical quantity, real-time mechanical quantity and real-time environmental quantity of each key component into a real-time electrical characteristic map, a real-time mechanical characteristic map and a real-time environmental characteristic map respectively; The real-time electrical characteristic map, real-time mechanical characteristic map, real-time environmental characteristic map, and real-time appearance shooting image of each key component are respectively fused to obtain a real-time mixed map of each key component; Step 6: Input the real-time mixed maps of all key components into the CKAN model trained in step 4. The CKAN model constructs the key feature mixed maps of the corresponding key components according to the second feature points extracted from the real-time mixed maps of each key component. ; Step 7: Calculate the real-time score of the health status of each key component in the GIS equipment according to formula (4) （4）； In formula (1), j is the health status, j=1,2; When j=1, It is the historical mixed spectrum when the historical health status of key component i is normal; When j=2, It is the historical mixed spectrum when the historical health status of key component i is severe; is the key feature mixed map of key component i; Real-time scoring of the health status of each key component Compare with the four scoring value intervals in step 1 to determine which health status each key component belongs to.

2. The method for evaluating the health status of GIS equipment according to claim 1, characterized in that: After step 7, the following steps are performed: Step 8: Calculate the weight of each key component in the same health state according to equations (5) and (6). （5）； （6）； Step 9: Calculate the information entropy of the key components of GIS equipment in normal and severe health states according to formula (7). （7）； Step 10: Calculate the weights of the key components of GIS equipment in normal and serious health states according to formula (8) （8）； Step 11: Calculate the overall score of GIS equipment according to formula (9) （9）； Compare the overall score Z of the GIS equipment with the four score value intervals in step 1 to determine the overall health status of the GIS equipment.

3. The method for evaluating the health status of GIS equipment according to claim 1, characterized in that: The hyperparameters are learning rate, batch size and regularization parameter.

Images