A GIS device fault prediction method and system

Abstract

The application provides a GIS device fault prediction method and system, the method comprises the following steps: obtaining partial discharge state detection data and electrical equipment state text information containing the partial discharge state detection data, and extracting a plurality of characteristic parameters which will affect the device fault; using a Word2Vec method to encode the plurality of characteristic parameters to obtain a word vector; inputting the word vector into a pre-constructed insulation defect classification model to enable the insulation defect classification model to capture the dependency relationship and influence between the plurality of characteristic parameters in the word vector, and outputting a prediction text of the GIS device fault type. The GIS device fault prediction method provided by the application comprehensively considers the influence of the plurality of characteristic parameters in the partial discharge state detection data and the electrical equipment state text information on the device fault, fully excavates the dependency relationship between the characteristic parameters, realizes comprehensive evaluation of the GIS device, and can improve the effectiveness and accuracy of the partial discharge severity evaluation.

Description

Technical Field

[0001] This invention relates to the technical field of partial discharge fault diagnosis of gas-insulated switchgear (GIS), and in particular to a fault prediction method and system for GIS equipment. Background Technology

[0002] As a key piece of equipment in the power system, the operating status of gas-insulated switchgear (GIS) directly affects the stable operation of the power grid. With the establishment of power system big data platforms, data centers have accumulated a large amount of condition monitoring data and textual information on electrical equipment status, including GIS condition monitoring data and operating parameters. This data and information exhibits typical characteristics of multi-source heterogeneous big data. Much important partial discharge information is recorded in text form, while traditional fault probability prediction methods typically only target structured data, making it difficult to perform data mining and feature learning on multi-format data and to complete fault probability prediction for partial discharge in GIS. Summary of the Invention

[0003] This invention aims to provide a method and system for predicting GIS equipment faults, in order to solve the above-mentioned technical problems. It comprehensively considers the impact of multiple feature parameters in the discharge status detection data and electrical equipment status text information on equipment faults, fully explores the dependencies between feature parameters, realizes a comprehensive evaluation of GIS equipment, and improves the effectiveness and accuracy of assessing the severity of partial discharge.

[0004] To address the aforementioned technical problems, this invention provides a method for predicting GIS equipment faults, comprising the following steps:

[0005] Acquire partial discharge status detection data and text information containing electrical equipment status, and extract multiple feature parameters that may affect equipment failure;

[0006] The Word2Vec method is used to encode multiple feature parameters to obtain word vectors;

[0007] Word vectors are input into a pre-built insulation defect classification model, which allows the model to capture the dependencies and influences between multiple feature parameters in the word vectors and output predicted text for GIS equipment fault types.

[0008] The above scheme comprehensively considers the impact of multiple feature parameters in the discharge status detection data and electrical equipment status text information on equipment failure, fully explores the dependencies between feature parameters, and realizes a comprehensive evaluation of GIS equipment, which can improve the effectiveness and accuracy of assessing the severity of partial discharge.

[0009] Furthermore, in the process of acquiring partial discharge state detection data and text information containing electrical equipment status, and extracting multiple feature parameters that may affect equipment failure, the feature parameters include PRPS spectrum data, partial discharge location, partial discharge amplitude development trend, voltage level, and operating time.

[0010] Furthermore, the partial discharge state detection data is PRPS map data; the word vectors are obtained by encoding multiple feature parameters using the Word2Vec method, including:

[0011] The PRPS map data is normalized and its dimensions are transformed to obtain one-dimensional matrix data;

[0012] Pre-encode one-dimensional matrix data to obtain string data;

[0013] The Word2Vec method is used to encode the feature parameters in string data and text information containing electrical equipment status to obtain word vectors.

[0014] Furthermore, before encoding multiple feature parameters using the Word2Vec method to obtain word vectors, the multiple feature parameters are sorted in the order of PRPS map data, partial discharge location, partial discharge amplitude development trend, voltage level, and running time.

[0015] Furthermore, the step of inputting word vectors into a pre-built insulation defect classification model, so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors, and output predicted text for GIS equipment fault types, specifically involves:

[0016] Word vectors are input into a pre-built insulation defect classification model based on the transformer framework, so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors and output predicted text for GIS equipment fault types.

[0017] The predictive text obtained by the above scheme can be used for effective early warning and risk assessment, formulate and implement corresponding maintenance strategies, ensure the safe and reliable operation of equipment, and avoid equipment failures from causing more adverse effects on the power system and society.

[0018] This invention provides a GIS equipment fault prediction system, comprising:

[0019] The feature parameter extraction module is used to acquire partial discharge state detection data and text information containing electrical equipment status, and extract multiple feature parameters that may affect equipment failure.

[0020] The word vector acquisition module is used to encode multiple feature parameters using the Word2Vec method to obtain word vectors;

[0021] The fault prediction module is used to input word vectors into a pre-built insulation defect classification model, so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors and output predicted text of GIS equipment fault types.

[0022] The above system architecture is simple and highly applicable. It can realize a GIS equipment fault prediction method, which comprehensively considers the impact of multiple feature parameters in the discharge status detection data and electrical equipment status text information on equipment faults, fully explores the dependencies between feature parameters, realizes a comprehensive evaluation of GIS equipment, and can improve the effectiveness and accuracy of assessing the severity of partial discharge.

[0023] Furthermore, in the feature parameter extraction module, the feature parameters include PRPS spectral data, partial discharge location, partial discharge amplitude development trend, voltage level, and running time.

[0024] Furthermore, the word vector acquisition module is used to encode multiple feature parameters using the Word2Vec method to obtain word vectors, including:

[0025] The PRPS map data is normalized and its dimensions are transformed to obtain one-dimensional matrix data;

[0026] Pre-encode one-dimensional matrix data to obtain string data;

[0027] The Word2Vec method is used to encode the feature parameters in string data and text information containing electrical equipment status to obtain word vectors.

[0028] Furthermore, in the word vector acquisition module, the Word2Vec method is used to encode multiple feature parameters. Before acquiring word vectors, the multiple feature parameters need to be sorted in the order of PRPS map data, partial discharge location, partial discharge amplitude development trend, voltage level, and running time.

[0029] Furthermore, the fault prediction module is used to input word vectors into a pre-built insulation defect classification model, so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors, and output predicted text for the fault type of GIS equipment, specifically:

[0030] Word vectors are input into a pre-built insulation defect classification model based on the transformer framework, so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors and output predicted text for GIS equipment fault types. Attached Figure Description

[0031] Figure 1 This is a schematic flowchart of a GIS equipment fault prediction method according to an embodiment of the present invention;

[0032] Figure 2 A schematic diagram showing the content before and after encoding the feature parameters provided in an embodiment of the present invention;

[0033] Figure 3 This is a schematic diagram of a transformer framework structure provided in an embodiment of the present invention;

[0034] Figure 4 This is a schematic diagram of a GIS equipment fault prediction system architecture provided in an embodiment of the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Please see Figure 1 This embodiment provides a method for predicting GIS equipment faults, including the following steps:

[0037] S1: Acquire partial discharge state detection data and text information containing electrical equipment status, and extract multiple feature parameters that will affect equipment failure;

[0038] S2: The Word2Vec method is used to encode multiple feature parameters to obtain word vectors;

[0039] S3: Input the word vectors into the pre-built insulation defect classification model so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors and output the predicted text of the GIS equipment fault type.

[0040] This embodiment comprehensively considers the impact of multiple feature parameters in the discharge status detection data and electrical equipment status text information on equipment failure, fully explores the dependencies between feature parameters, and realizes a comprehensive evaluation of GIS equipment, which can improve the effectiveness and accuracy of assessing the severity of partial discharge.

[0041] It should be noted that in step S1, the partial discharge status detection data and the text information containing the electrical equipment status together constitute the GIS fault case dataset. The partial discharge status detection data is mainly the PRPS map data obtained from UHF detection, and the text information containing the electrical equipment status is obtained through the partial discharge detection report, which mainly includes the partial discharge location, the development trend of the partial discharge amplitude, the voltage level, and the operating time.

[0042] Furthermore, in the process of acquiring partial discharge state detection data and text information containing electrical equipment status, and extracting multiple feature parameters that may affect equipment failure, the feature parameters include PRPS spectrum data, partial discharge location, partial discharge amplitude development trend, voltage level, and operating time.

[0043] Furthermore, the partial discharge state detection data is PRPS map data; the word vectors are obtained by encoding multiple feature parameters using the Word2Vec method, including:

[0044] The PRPS map data is normalized and its dimensions are transformed to obtain one-dimensional matrix data, which is then stored in the form of a one-dimensional array.

[0045] Pre-encode one-dimensional matrix data to obtain string data;

[0046] The Word2Vec method is used to encode the feature parameters in string data and text information containing electrical equipment status to obtain word vectors.

[0047] Furthermore, before encoding multiple feature parameters using the Word2Vec method to obtain word vectors, the multiple feature parameters are sorted in the order of PRPS map data, partial discharge location, partial discharge amplitude development trend, voltage level, and running time.

[0048] It should be noted that the Word2Vec method is based on deep semantic learning to train word vectors. Through training, words can be mapped into N-dimensional real vectors, thereby encoding the feature parameters of partial discharge. The resulting N-dimensional real vectors are the input data of the insulation defect classification model. One of the final outputs of the insulation defect classification model is the failure probability value of electrical equipment that may have one of four types of defects: tip discharge, floating discharge, particle discharge, and insulation-related discharge.

[0049] Furthermore, the step of inputting word vectors into a pre-built insulation defect classification model, so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors, and output predicted text for GIS equipment fault types, specifically involves:

[0050] Word vectors are input into a pre-built insulation defect classification model based on the transformer framework, so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors and output predicted text for GIS equipment fault types.

[0051] The predicted text obtained in this embodiment can be used for effective early warning and risk assessment, formulate and activate corresponding maintenance strategies, ensure the safe and reliable operation of equipment, and avoid equipment failures from causing more adverse effects on the power system and society.

[0052] To further illustrate the technical features and highlight the advantages of this technical solution, this embodiment provides a specific application of the GIS equipment fault prediction method. Specifically, it includes:

[0053] Step 1: Acquire partial discharge status detection data and text information containing electrical equipment status to establish a GIS fault case dataset, and extract multiple feature parameters that will affect equipment failure. The data comes from on-site operation monitoring at substations. All samples in the established GIS fault case dataset confirm internal partial discharge signals, containing partial discharge status detection data and partial discharge detection reports. A GIS equipment failure is defined as a breakdown accident occurring within that interval within one month after the detection of a partial discharge signal, or a serious defect, approaching insulation failure, discovered after disassembly and inspection. If the GIS equipment is found to have only minor defects that do not affect operation after disassembly, it is considered a counterexample. Among these:

[0054] Partial discharge status detection data consists of PRPS spectrum data obtained from UHF detection, represented by a 72×50 two-dimensional matrix. The two dimensions of the matrix represent the phase and period of the partial discharge, respectively, and the value of each point represents the discharge amplitude. The partial discharge status text information is the textual information recorded in the detection report, including the partial discharge location, the development trend of the partial discharge amplitude, the voltage level, and the operating time. The partial discharge location is determined in the record using a positioning method to identify the discharging component, specifically: basin insulator, grounding switch, isolating switch, circuit breaker, current transformer (CT), current transformer (PT), and busbar. The development trend of the partial discharge amplitude is categorized into three types: basically stable, increasing trend, and decreasing trend. Voltage levels include 110kV, 220kV, 330kV, and 500kV. Operating time includes 1-5 years, 5-10 years, 10-15 years, and over 15 years. The filtered data includes 154 positive examples and 1838 negative examples.

[0055] Finally, based on the partial discharge status detection data and text information containing electrical equipment status, PRPS map data, partial discharge location, partial discharge amplitude development trend, voltage level, and operating time are extracted as feature parameters.

[0056] Step 2: Encode the partial discharge state detection (PRPS) spectrum data of the real-valued matrix into a string to enhance the bit information of the real-valued data, which is beneficial for the transformer model to extract information from the PRPS data. The preprocessing method for the partial discharge state detection (PRPS) spectrum data is as follows:

[0057] 1) The partial discharge state detection PRPS spectrum data is first normalized, as shown in Formula 1:

[0058]

[0059] In the formula: x max x is the maximum value in the two-dimensional matrix; min is the minimum value in the two-dimensional matrix; x' is the linearly normalized data, and the calculation result is retained to 3 significant figures.

[0060] 2) After normalization, the PRPS data x' represented by the 72×50 two-dimensional matrix is ​​transformed into a one-dimensional matrix x″ with a length of 3600.

[0061] 3) Encode each data point in x″ as a string of the form v1_p1, v2_p2, ..., vn_pn, where vn_ represents the value of the nth digit of the data, and pn represents the number of scientific notation digits corresponding to that value. For example, 0.532 would be encoded as: 0_0,5_-1,3_-2,2_-3. See [link to documentation] for details. Figure 2 The PRPS data string in the file.

[0062] Step 3: To comprehensively assess the equipment status by considering the impact of multiple feature parameters in partial discharge status detection data and electrical equipment status text information on equipment failure, such as... Figure 2 The "Encoded Data" section shows that this embodiment encodes each set of partial discharge state detection PRPS map data and partial discharge state text information from the GIS fault case dataset according to... Figure 2 The rules in this system prioritize PRPS map data, partial discharge location, partial discharge amplitude development trend, voltage level, and operating time. This expands the parameter selection range for GIS equipment fault prediction by more effectively learning from PRPS data. The preprocessing steps for each group of data after sorting are shown below:

[0063] 1) Use the Word2Vec method to... Figure 2Each word represented by a rounded rectangle (e.g., "circuit breaker") is encoded into a word vector. The gensim toolkit in Python is used for implementation. During training, the number of words scanned before and after the initial scan is 3, the word frequency threshold is 0, the word vector dimension is 18, and the random number generator is set to 1. Finally, all word vectors are represented by a single 1×18 vector. The number of words in each data set is fixed at 3613, without zero-padding. For each data set X1 encoded into a word vector, its size is 3613×18.

[0064] 2) Calculate the position embedding representation X2 for each element in X1, using the following formula:

[0065]

[0066]

[0067] In the formula, pos is the sequence number of the word in each data set, with values ​​ranging from 1 to 3613; in formula (2), i ranges from 0 to 8, and 2i+1 represents odd numbers from 1 to 18; in formula (3), i ranges from 1 to 9, and 2i represents even numbers from 1 to 18. The advantage of using trigonometric functions to represent position embeddings is that the position embedding of the word at position i can be linearly represented by the position embedding of the word at position i+k, reflecting the relative positional relationship between the two words. Furthermore, the inner product of the position embeddings at positions i and i+k decreases as the relative positions increase, thus representing the relative distance between the positions.

[0068] 3) Add X1 and X2 together to obtain the input data X for the prediction model in the next step. The PRPS data encoding and parameter sorting processes used in this embodiment to obtain the input data X ensure that the input data X has a certain degree of comparability and consistency, contains sufficient information required for GIS equipment fault prediction, and meets the requirements of the transformer framework for input data.

[0069] Step 4: In order to explore the influence of the interrelationships among multiple feature parameters in the partial discharge state detection data and electrical equipment state text information, this embodiment selects the Transformer framework with attention mechanism to capture the dependencies and influences between feature parameters, thereby improving the accuracy of fault prediction. Figure 3 This is a transformer framework for predicting faults in GIS equipment. The transformer framework consists of two parts: an encoder and a decoder.

[0070] The encoder's input data X has a size of 3613×18, and its output data Z also has a size of 3613×18, serving as one of the decoder's inputs. The encoder consists of a multi-head attention layer, a residual and normalization layer, a feedforward neural network, and another residual and normalization layer. The multi-head attention layer comprises three separate self-attention modules, each with an input and output size of 3613×6. The outputs of the three self-attention modules are concatenated to obtain the 3613×18 output of the multi-head attention layer. The feedforward neural network contains one fully connected layer and one output layer, with 18 neurons in each layer. The activation function is the ReLU function, and the output size of the feedforward neural network is 3613×18. The encoder encodes the input information into a continuous representation with attention information, which helps the decoder focus on appropriate words in the input during the decoding process.

[0071] The decoder's task is to generate a text sequence predicting GIS device malfunctions. The decoder consists of two multi-head attention layers, a feedforward neural network layer, residual connections and normalization after each sub-layer, a linear neural network layer for scaling, and a softmax layer for calculating the probability values ​​of the output text sequence. The decoder is autoregressive, outputting a 1×18 vector representing a given text character each time. The updated output sequence Y1 is then used as one of the inputs (the other input data is the encoder's output data Z) until the model gives... <end>Until the terminator.

[0072] Step 5: Following the order of Steps 1 to 4, input the 1992 sets of GIS fault case data containing label values ​​into... Figure 3 The model is trained in the transformer framework. After inputting the data to be predicted, the trained model can output the predicted text of the fault type of GIS equipment, including "point discharge", "particle discharge", "suspended discharge", "insulation discharge" and "no partial discharge".

[0073] This embodiment proposes a GIS fault probability prediction method based on the fusion of partial discharge state detection data and electrical equipment status text information. It comprehensively considers the impact of multiple feature parameters in the partial discharge state detection data and electrical equipment status text information on equipment faults, mines the dependencies between feature parameters, and comprehensively assesses equipment status, thereby improving the effectiveness and accuracy of partial discharge severity assessment. In its specific implementation, this embodiment encodes the real-number matrix partial discharge state detection (PRPS) map data into strings, enhancing the bit information of the real-number data itself, which is beneficial for the prediction model to extract information from the PRPS data. The GIS partial discharge state detection (PRPS) map data and partial discharge status text information are sorted according to specific rules, expanding the parameter selection range for GIS equipment fault prediction based on more effective learning of PRPS data information, making the input data of the prediction model comparable and consistent, and containing sufficient information required for GIS equipment fault prediction. A transformer model with an attention mechanism is used to mine and capture the interrelationships and influences between multiple feature parameters in the partial discharge state detection data and electrical equipment status text information, thereby improving the accuracy of fault prediction.

[0074] Please see Figure 4 This embodiment provides a GIS equipment fault prediction system for implementing a GIS equipment fault prediction method. The system includes:

[0075] The feature parameter extraction module is used to acquire partial discharge state detection data and text information containing electrical equipment status, and extract multiple feature parameters that may affect equipment failure.

[0076] The word vector acquisition module is used to encode multiple feature parameters using the Word2Vec method to obtain word vectors;

[0077] The fault prediction module is used to input word vectors into a pre-built insulation defect classification model, so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors and output predicted text of GIS equipment fault types.

[0078] The system architecture provided in this embodiment is simple and highly applicable. It can realize a GIS equipment fault prediction method. It comprehensively considers the impact of multiple feature parameters in the discharge status detection data and electrical equipment status text information on equipment faults, fully explores the dependencies between feature parameters, realizes a comprehensive evaluation of GIS equipment, and can improve the effectiveness and accuracy of assessing the severity of partial discharge.

[0079] Furthermore, in the feature parameter extraction module, the feature parameters include PRPS spectral data, partial discharge location, partial discharge amplitude development trend, voltage level, and running time.

[0080] Furthermore, the word vector acquisition module is used to encode multiple feature parameters using the Word2Vec method to obtain word vectors, including:

[0081] The PRPS map data is normalized and its dimensions are transformed to obtain one-dimensional matrix data;

[0082] Pre-encode one-dimensional matrix data to obtain string data;

[0083] The Word2Vec method is used to encode the feature parameters in string data and text information containing electrical equipment status to obtain word vectors.

[0084] Furthermore, in the word vector acquisition module, the Word2Vec method is used to encode multiple feature parameters. Before acquiring word vectors, the multiple feature parameters need to be sorted in the order of PRPS map data, partial discharge location, partial discharge amplitude development trend, voltage level, and running time.

[0085] Furthermore, the fault prediction module is used to input word vectors into a pre-built insulation defect classification model, so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors, and output predicted text for the fault type of GIS equipment, specifically:

[0086] Word vectors are input into a pre-built insulation defect classification model based on the transformer framework, so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors and output predicted text for GIS equipment fault types.

[0087] The predicted text obtained in this embodiment can be used for effective early warning and risk assessment, formulate and activate corresponding maintenance strategies, ensure the safe and reliable operation of equipment, and avoid equipment failures from causing more adverse effects on the power system and society.

[0088] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.< / end>

Claims

1. A method for predicting faults in GIS equipment, characterized in that, Includes the following steps: Acquire partial discharge status detection data and text information containing electrical equipment status, and extract multiple feature parameters that may affect equipment failure; The Word2Vec method is used to encode multiple feature parameters to obtain word vectors; Word vectors are input into a pre-built insulation defect classification model based on the transformer framework, so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors and output predicted text for GIS equipment fault types. The partial discharge state detection data is PRPS map data; the word vectors are obtained by encoding multiple feature parameters using the Word2Vec method, including: The PRPS map data is normalized and its dimensions are transformed to obtain one-dimensional matrix data; Pre-encode one-dimensional matrix data to obtain string data; The Word2Vec method is used to encode the feature parameters in string data and text information containing electrical equipment status to obtain word vectors.

2. The GIS equipment fault prediction method according to claim 1, characterized in that, In the process of acquiring partial discharge status detection data and text information containing electrical equipment status, and extracting multiple feature parameters that may affect equipment failure, the feature parameters include PRPS spectrum data, partial discharge location, partial discharge amplitude development trend, voltage level, and operating time.

3. The GIS equipment fault prediction method according to claim 1, characterized in that, Before encoding multiple feature parameters using the Word2Vec method to obtain word vectors, the multiple feature parameters are sorted in the order of PRPS map data, partial discharge location, partial discharge amplitude development trend, voltage level, and running time.

4. A GIS equipment fault prediction system, characterized in that, include: The feature parameter extraction module is used to acquire partial discharge state detection data and text information containing electrical equipment status, and extract multiple feature parameters that may affect equipment failure. The word vector acquisition module is used to encode multiple feature parameters using the Word2Vec method to obtain word vectors; The fault prediction module is used to input word vectors into a pre-built insulation defect classification model so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors and output predicted text for the fault type of GIS equipment. The word vector acquisition module is used to encode multiple feature parameters using the Word2Vec method to obtain word vectors, including: The PRPS map data is normalized and its dimensions are transformed to obtain one-dimensional matrix data; Pre-encode one-dimensional matrix data to obtain string data; The Word2Vec method is used to encode the feature parameters in string data and text information containing electrical equipment status to obtain word vectors.

5. A GIS equipment fault prediction system according to claim 4, characterized in that, In the feature parameter extraction module, the feature parameters include PRPS map data, partial discharge location, partial discharge amplitude development trend, voltage level, and running time.

6. A GIS equipment fault prediction system according to claim 4, characterized in that, The word vector acquisition module uses the Word2Vec method to encode multiple feature parameters. Before acquiring word vectors, the multiple feature parameters need to be sorted in the order of PRPS map data, partial discharge location, partial discharge amplitude development trend, voltage level, and running time.

7. A GIS equipment fault prediction system according to any one of claims 4 to 6, characterized in that, The fault prediction module is used to input word vectors into a pre-built insulation defect classification model, so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors, and output predicted text for the fault type of GIS equipment, specifically: Word vectors are input into a pre-built insulation defect classification model based on the transformer framework, so that the insulation defect classification model can capture the dependencies and influences between multiple feature parameters in the word vectors and output predicted text for GIS equipment fault types.

Images