Multi-system combined diagnosis method for evaluating state of extra-high voltage power transformation equipment

Abstract

The invention discloses a multi-system combined diagnosis method for evaluating the state of extra-high voltage power transformation equipment. The method comprises the following steps: acquiring diagnosis results of expert systems for daily inspection, online monitoring, live detection and offline test of the extra-high voltage power transformation equipment; aiming at different time intervals and data of different data sources, establishing a uniform interface model reflecting the uniformity of time and data; determining the accuracy and the adaptation domain of the diagnosis result of each expert system by using the existing event library of the extra-high voltage power transformation equipment, and constructing a differentiated weight system for different equipment problems; and training and obtaining a neural network model by utilizing a machine learning technology based on the differentiated weight system and the unified interface model, and deploying the neural network model at an edge device side to realize multi-system combined diagnosis. According to the method, joint analysis can be performed on diagnosis conclusions of multiple expert systems, quantitative evaluation can be performed on credibility, operators can be assisted to judge indexes, and misjudgment is reduced.

Description

Technical Field

[0001] The present invention relates to the field of ultra-high voltage intelligent diagnosis, and in particular to a multi-system joint diagnosis method for ultra-high voltage substation equipment status evaluation. Background Art

[0002] UHV substation equipment is relatively rarely used worldwide, and currently only China is using it on a large scale. Due to the lack of manufacturing and operating experience of such equipment, the failure rate of major UHV substation equipment represented by UHV transformers, reactors and GIS is significantly higher than that of similar ultra-high voltage equipment.

[0003] At present, there are the most numerous and complete expert systems in UHV substations, but there is a lack of joint diagnosis methods, so judgments can only rely on the experience of operators. There is a high degree of dependence on personnel and a lack of comprehensive and intelligent decision-making support systems, which can easily lead to misjudgments and missed judgments.

[0004] In addition, since different expert systems often do not have complete status information and their internal diagnostic logic and decision-making methods are also different, different detection conclusions or even completely contradictory maintenance recommendations may be produced.

[0005] At present, accidents have occurred in which equipment has been shut down due to relying on a single indicator and a single system diagnostic result, which has had a significant negative impact on the safe operation of the UHV power grid. Summary of the invention

[0006] In order to solve the above technical problems, the present invention discloses a multi-system joint diagnosis method for ultra-high voltage substation equipment status evaluation, characterized in that the method comprises the following steps: S100, collects the diagnostic results of various expert systems including daily inspection, online monitoring, live detection and offline testing of UHV substation equipment; S200, determining the time section of diagnosis for different time intervals, and establishing a unified interface model that reflects the uniformity of time and data for data from different data sources; S300, using the existing event library of the UHV substation equipment, determining the accuracy and adaptability of the diagnostic results of each expert system for daily inspection, online monitoring, live detection and offline testing of the UHV substation equipment, and constructing a differentiated weight system for different equipment problems, wherein each weight in the differentiated weight system is a weight of the credibility of the diagnostic results of each expert system; S400: Based on the differentiated weight system and unified interface model, a neural network model is trained and obtained by using machine learning technology, and the neural network model is deployed on the edge device side to realize multi-system joint diagnosis of UHV substation equipment status evaluation.

[0007] Preferably, The UHV substation equipment includes: UHV transformers, UHV reactors, UHV GIS and other main equipment.

[0008] Preferably, After step S400, the following steps are also included: S500 and the edge device side further utilize the power communication protocol to connect with the substation integrated automation system to become a standard form of UHV substation smart service, so that the edge device side can be accessed at any time on the platform of the substation integrated automation system.

[0009] Preferably, Step S100 also includes the following steps: Receive various field data generated by various expert systems, including: data with continuous time series characteristics generated by the online monitoring system, irregular data generated by manual inspection and automatic inspection, and preventive status test data with a long period (more than 1 year).

[0010] Preferably, Step S100 also includes the following steps: Further collect and manage the on-site equipment information and automatically attribute the various on-site data to the corresponding UHV substation equipment.

[0011] Preferably, The equipment information includes basic information of major equipment such as UHV transformers, UHV reactors and UHV GIS.

[0012] Preferably, The diagnostic results of each expert system include but are not limited to: the diagnostic conclusions of dissolved gas in oil (DGA) of oil-immersed equipment, the diagnostic conclusions of fast-reactive hydrogen detection devices, the diagnostic conclusions of comprehensive monitoring devices (partial discharge, ultrasound, vibration), the diagnostic conclusions of GIS gas composition analysis, etc.

[0013] Preferably, In step S200, the time section of diagnosis is determined based on the equipment characteristics of the UHV substation equipment, the detection items and the credibility of the diagnostic results of the corresponding expert system, so as to provide meaningful, reliable field data and diagnostic results within a specific time section as a training set for the training of the neural network model, wherein the field data is equivalent to the input and the diagnostic result is equivalent to the output, and the input, output and differentiated value system determined in this way are used to train the neural network model.

[0014] Preferably, For oil-immersed equipment, the test items are monitoring and diagnosis of dissolved gas types in oil, and the reliability regression period is 1 month; among them, For high-quality oil-immersed equipment, consider the performance degradation time of more than 8 years including 8 years; For inferior oil-immersed equipment, allow for 2 years of degradation.

[0015] Preferably, For the devices used in offline tests, by default, testing is carried out once a year; When the device used in the offline test encounters short circuit, discharge, overheating, or oil replenishment events, the previous historical data is no longer of reference value and offline testing must be repeated.

[0016] Preferably, The following model is used to characterize the change in the credibility of the diagnosis results of different expert systems over time:

[0017] Cn is the credibility of the nth expert system, An is the normalized coefficient of credibility, kn is the sensitivity system, N is the calibration period (N>1), T is the credibility decline period (in days), and t is the actual running time.

[0018] Preferably, For the integration of credibility, in terms of oil sampling DGA and online DGA monitoring: The DGA test of oil sampling uses a desktop instrument with high accuracy and a good traceability guarantee system, so the absolute value of the measurement is more reliable. However, when the relevant equipment has suffered from short circuit, overheating or internal discharge defects, the previous test data cannot be used as a basis for judgment, but can be used as a reference for the benchmark value; Online monitoring equipment can provide continuous detection data, which is very helpful for the diagnosis of development trends. However, its measurement accuracy will gradually deviate from the accurate value as the running time increases, and regular calibration is required. Therefore, trend judgment should be used as the main reference. In order to organically integrate the calibration cycle, based on the objective law of the accuracy drift of different detection devices corresponding to the relevant expert systems for the same substation equipment object, the following credibility fusion model is used for processing:

[0019] Where C is the comprehensive data credibility, mi(t) is the credibility weight of the i-th detection device that changes with time, Ci(t) is the credibility of the i-th detection device that changes with time, t is time, and M is the number of all detection devices; Credibility fusion is to obtain comprehensive data credibility, which means: for the same substation equipment object, the fused credibility reflected by the results of the different detection devices, the fused credibility is associated with time, and the fused credibility is further normalized to determine the corresponding weight in the differentiated weight system.

[0020] Preferably, The weights corresponding to each expert system can be determined by regression calculation and normalization.

[0021] Beneficial effects of the present invention: The present invention can not only jointly analyze the diagnostic conclusions of multiple expert systems and give a quantitative evaluation of the degree of credibility, thereby assisting operating personnel in judging indicators and reducing the occurrence of misjudgments, but also standardize the smart services of UHV substations to make them consistent in different environments and applications, thereby improving the operating efficiency and maintainability of UHV equipment products. Standardization also helps to reduce training costs because different substations can share similar conclusion application methods.

[0022] The present invention can be integrated as a part of intelligent inspection, or it can be a module of intelligent decision-making and command system. The present invention embodies the following key points: First, the present invention can take into account the fact that the credibility of the detection device gradually decreases, and perform fusion modeling of multi-diagnosis system time sections; second, the present invention can give initial weight values ​​for different fields, and support self-learning of weights, such as supervised artificial self-learning; third, the present invention can model the method disclosed by it with a standard universal interface, build virtual and intelligent equipment, and incorporate it into the substation communication and automation monitoring system or the substation integrated automation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings illustrate exemplary embodiments of the present invention and together with the description serve to explain the principles of the present invention. These drawings are included to provide a further understanding of the present invention and are incorporated in and constitute a part of this specification.

[0024] Figure 1 It is a flow chart of a multi-system joint diagnosis method for ultra-high voltage substation equipment status evaluation in one embodiment of the present invention; Figure 2 It is a schematic diagram of different credibility caused by different detection items of the same object in one embodiment of the present invention; Figure 3 is a schematic diagram of the integrated credibility after fusion with a time section in one embodiment of the present invention; Figure 4It is a schematic diagram of the mapping relationship between the sensing technology and the diagnostic results of the expert system corresponding to the weights in one embodiment of the present invention. Implementation

[0025] The present invention will be further described in detail below in conjunction with the accompanying drawings and implementations. It is to be understood that the specific implementations described herein are only used to explain the relevant content, rather than to limit the present invention. It is also necessary to explain that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings.

[0026] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0027] Unless otherwise specified, the exemplary embodiments / embodiments shown will be understood as exemplary features that provide various details of some ways in which the technical concept of the present invention can be implemented in practice. Therefore, unless otherwise specified, the features of the various embodiments / embodiments can be further combined, separated, interchanged and / or rearranged without departing from the technical concept of the present invention.

[0028] The use of cross-hatching and / or shading in the accompanying drawings is generally used to make the boundaries between adjacent components clear. As such, unless otherwise specified, the presence or absence of cross-hatching or shading does not convey or indicate any preference or requirement for the specific materials, material properties, dimensions, proportions, commonalities between the components shown, and / or any other characteristics, attributes, properties, etc. of the components. In addition, in the accompanying drawings, the sizes and relative sizes of the components may be exaggerated for clarity and / or descriptive purposes. When the exemplary embodiments can be implemented differently, the specific process sequence can be performed in a different order than described. For example, two successively described processes can be performed substantially simultaneously or in an order opposite to the described order. In addition, the same figure numbers represent the same components.

[0029] When a component is referred to as being "on" or "over," "connected to," or "coupled to" another component, the component may be directly on, directly connected to, or directly coupled to the other component, or intervening components may be present. However, when a component is referred to as being "directly on," "directly connected to," or "directly coupled to" another component, there are no intervening components. For this purpose, the term "connected" may refer to a physical connection, an electrical connection, etc., with or without intervening components.

[0030] For descriptive purposes, the present invention may use spatially relative terms such as "below", "beneath", "under", "lower", "above", "upper", "above", "higher", and "side (e.g., as in "sidewall")" to describe the relationship of one component to another (other) component as shown in the accompanying drawings. The spatially relative terms are intended to encompass different orientations of the device in use, operation, and / or manufacture in addition to the orientation depicted in the accompanying drawings. For example, if the device in the accompanying drawings is turned over, components described as "below" or "beneath" other components or features would then be positioned "above" the other components or features. Thus, the exemplary term "below" can include both the "above" and "below" orientations. Furthermore, the device may be otherwise oriented (e.g., rotated 90 degrees or at other orientations), and the spatially relative descriptors used herein should be interpreted accordingly.

[0031] The terms used here are for the purpose of describing specific embodiments, and are not intended to be restrictive. As used here, unless the context clearly indicates otherwise, the singular forms "one (kind, person)" and "said (the)" are also intended to include plural forms. In addition, when the terms "include" and / or "include" and their variations are used in this specification, it is explained that there are stated features, integral bodies, steps, operations, parts, assemblies and / or their groups, but one or more other features, integral bodies, steps, operations, parts, assemblies and / or their groups are not excluded from existence or addition. It should also be noted that, as used here, the terms "substantially", "approximately" and other similar terms are used as approximate terms and not as degree terms, so that they are used to explain the inherent deviations of the measured values, calculated values ​​and / or the values ​​provided that will be recognized by those of ordinary skill in the art.

[0032] See also Figure 1 In one embodiment, the present invention discloses a multi-system joint diagnosis method for ultra-high voltage substation equipment status evaluation, characterized in that the method comprises the following steps: S100, collects the diagnostic results of various expert systems including daily inspection, online monitoring, live detection and offline testing of UHV substation equipment; S200, determining the time section of diagnosis for different time intervals, and establishing a unified interface model that reflects the uniformity of time and data for data from different data sources; S300, using the existing event library of the UHV substation equipment, determining the accuracy and adaptability of the diagnostic results of each expert system for daily inspection, online monitoring, live detection and offline testing of the UHV substation equipment, and constructing a differentiated weight system for different equipment problems, wherein each weight in the differentiated weight system is a weight of the credibility of the diagnostic results of each expert system; S400: Based on the differentiated weight system and unified interface model, a neural network model is trained and obtained by using machine learning technology, and the neural network model is deployed on the edge device side to realize multi-system joint diagnosis of UHV substation equipment status evaluation.

[0033] It can be understood that accuracy corresponds to credibility and the weight of credibility.

[0034] In another embodiment, The UHV substation equipment includes: UHV transformers, UHV reactors, UHV GIS and other main equipment.

[0035] In another embodiment, S100 specifically includes the following steps: S101. Define the data source of multi-system joint diagnosis for UHV substation equipment status evaluation and obtain data, for example, confirm that data will be collected from the online oil chromatography monitoring system of the UHV transformer, the partial discharge monitoring system of the UHV reactor, the gas composition analysis system of the GIS, and regular manual inspection reports; it can be understood that this step can be manually read from the data files stored in each system itself or automatically collected by developing an API interface, and is not limited to the form of obtaining data; S102. Preprocessing the data of multi-system joint diagnosis used for the status evaluation of UHV substation equipment, such as identifying and processing abnormal values ​​(such as readings outside the normal range) and filling missing values ​​(such as using the nearest neighbor interpolation method); S103. Classify and label the pre-processed multi-system joint diagnosis data for UHV substation equipment status evaluation, for example, label the oil chromatography data as "chemical stability" and label the partial discharge data as "electrical performance". At the same time, experts in the field can also be invited to label some key data points, such as "normal", "potential risk" (including what kind of potential risk), etc.

[0036] In another embodiment, S200 specifically includes the following steps: S201, analyzing equipment characteristics through data classified and labeled in S103, such as analyzing different failure modes of UHV transformers through dissolved gas in oil; S202, establish a time frame, for example, collect data snapshots once at 8 am, ..., and 8 pm every day to capture changes in device status within 24 hours a day; S203. Unified data format: All collected data are converted into a unified time series format to ensure comparability between data from different sources.

[0037] S204, construct a time section, for example, select the data snapshot at 8 o'clock every day as the "section", and make the section contain all relevant data in the past 24 hours, forming a complete status assessment basis.

[0038] In another embodiment, S300 specifically includes the following steps: S301. Establish an event database, for example, integrate the fault records and maintenance reports of all equipment in the past five years to form an event database; S302, verifying the accuracy of the diagnosis, for example, by comparing the actual fault conditions in the event database with the diagnosis results provided by each system, it is found that the oil chromatography monitoring system is most sensitive to the detection of early faults, and other expert systems are similar; S303, determining the applicable domain, for example, clarifying that the oil chromatography monitoring system is particularly suitable for detecting overheating faults of ultra-high voltage transformers, while the partial discharge monitoring system is more adept at detecting electrical insulation defects, thereby determining corresponding applicable domains for different systems; S304. Design weights, for example, assigning a higher weight (eg, 0.7) to the diagnostic results of the oil chromatography monitoring system and a lower weight (eg, 0.3) to the partial discharge monitoring system, to reflect their effectiveness in detecting specific fault types.

[0039] In another embodiment, S400 specifically includes the following steps: S401. Construct a suitable neural network model. For example, consider the complexity of the scenarios and problems targeted by the present invention, the diversity of data, and the influence of historical information, and select a suitable neural network model.

[0040] S402, prepare a training set, for example, randomly select 80% of the data from the past year as a training set, and the remaining 20% ​​as a test set; S403, preliminarily training the model, for example, training the model through a training set and performing multiple iterations, so that the accuracy of the model on the test set reaches 95%; S404, model validation: ensure that the model is not overfitted through cross validation, and verify the generalization ability of the model on new data to meet the standards, thereby completing the training; S405, edge computing deployment: Deploy the trained model to the edge computing server in the substation to achieve multi-system joint diagnosis of the status of UHV equipment.

[0041] In another embodiment, the neural network model includes an input layer, a multi-unit LSTM layer, a hidden layer and an output layer, wherein: Input layer: Assuming that the existing expert system has a total of m sensing and monitoring quantities (such as temperature, pressure, gas composition of xx equipment, temperature of yy equipment, current of zz equipment, etc.), the dimension of the input layer is m; for example, m is 50; since the present invention obtains data according to time, the input dimension of each time step is m; illustratively, the data input to the input layer is: (batch_size, 24, 50), where batch_size is the number of samples processed each time, 24 is the number of time steps (note: corresponding to 24 hours in S202 above), and 50 is the number of features in each time step; the batch_size is the batch size of the model processing data, which is consistent with the batch size when training the model, for example, the batch size is set to 32; The multi-unit LSTM layer includes n LSTM units, wherein the value of n is based on capturing the complex dependencies in the data of multi-system joint diagnosis for the status evaluation of the UHV substation equipment, wherein the dependencies include the dynamic changes in the status of the UHV substation equipment and the mutual influence between different sensors; and the multi-unit LSTM layer only returns the output of the last time step, and the output data is: (batch_size, n); Exemplarily, n is 64 by default; it should be noted that, considering the complexity of the scenarios and problems of the present invention, when the verification of step S404 shows that the model is overfitting, n is reduced, for example, from 64 to 32; when the verification of step S404 shows that the model is underfitting, n is increased, for example, from 64 to 128; A hidden layer, whose input is the output of the multi-unit LSTM layer, wherein the hidden layer includes multiple hidden layers, for example, including: The first hidden layer: contains 128 neurons, and the activation function is ReLU, which performs nonlinear transformation based on the output of the multi-unit LSTM layer; The second hidden layer: contains 64 neurons, and the activation function is ReLU, which extracts high-level features based on the output of the first hidden layer and reduces the dimension of the features; The third hidden layer: contains 32 neurons, and the activation function is ReLU, which extracts higher-level features based on the output of the second hidden layer and further reduces the dimension of the features; Output layer: contains p neurons. The activation function of each neuron is Sigmoid. The output value represents the health probability of the device status (0 represents unhealthy and 1 represents healthy).

[0042] It should be noted that the value of p depends on the total categories of multi-system joint diagnosis and the number of states in each category. For example, assuming there are 3 categories in total (a, b, c) and each category has two states (healthy or unhealthy), then p is equal to 3*2, which means a total of 6 output neurons are required.

[0043] In another embodiment, in step S403, The loss function uses the binary cross-entropy loss function, which is suitable for binary classification problems. This is because the number of states of each category diagnosed by multi-system joint diagnosis is healthy and unhealthy, which belongs to binary classification; The optimizer uses the Adam optimizer with a learning rate of 0.001; Set the batch size to 32, that is, process 32 samples each time during training; The number of training rounds is set to 100.

[0044] In another embodiment, Use K-fold cross validation (for example, 5-fold) to verify the generalization ability of the model.

[0045] Exemplarily, the training set is further divided into 5 parts, 4 parts of the data are used to train the model each time, and the remaining 1 part of the data is used for verification. This is repeated 5 times, and the average verification result is taken.

[0046] It should be noted that in the initial training and verification stages of S403 and S404, in addition to accuracy evaluation, precision, recall and F1 score can also be used to comprehensively evaluate the performance of the model.

[0047] In another embodiment, Each hidden layer is a fully connected layer.

[0048] In another embodiment, After step S400, the following steps are also included: S500 and the edge device side further utilize the power communication protocol to connect with the substation integrated automation system to become a standard form of UHV substation smart service, so that the edge device side can be accessed at any time on the platform of the substation integrated automation system.

[0049] Exemplarily, the power communication protocol may be a variety of power communication protocols such as IEC61850 protocol and IEC104 protocol.

[0050] In another embodiment, Step S100 also includes the following steps: Receive various field data generated by various expert systems, including: data with continuous time series characteristics generated by online monitoring systems, irregular data generated by manual inspections and automatic inspections, and preventive status test data with a long period (more than 1 year).

[0051] In another embodiment, Step S100 also includes the following steps: Further collect and manage the on-site equipment information and automatically attribute the various on-site data to the corresponding UHV substation equipment.

[0052] In another embodiment, The equipment information includes basic information of major equipment such as UHV transformers, UHV reactors and UHV GIS.

[0053] In another embodiment, The diagnostic results of each expert system include but are not limited to: the diagnostic conclusions of dissolved gas in oil (DGA) of oil-immersed equipment, the diagnostic conclusions of fast-reactive hydrogen detection devices, the diagnostic conclusions of comprehensive monitoring devices (partial discharge, ultrasound, vibration), the diagnostic conclusions of GIS gas composition analysis, etc.

[0054] In another embodiment, In step S200, the time section of diagnosis is determined based on the equipment characteristics of the UHV substation equipment, the detection items and the credibility of the diagnostic results of the corresponding expert system, so as to provide meaningful, reliable field data and diagnostic results within a specific time section as a training set for the training of the neural network model, wherein the field data is equivalent to the input and the diagnostic result is equivalent to the output, and the input, output and differentiated value system determined in this way are used to train the neural network model.

[0055] In fact, since the time for online monitoring, live detection and offline testing is not the same, it is necessary to establish a time section.

[0056] See also Figure 2 Since the drift of chemical sensor systems is large, the detection equipment needs to be continuously calibrated in actual use to ensure the accuracy of the device and diagnostic results. Therefore, the monitoring and diagnostic device of dissolved gas in oil needs to be constantly calibrated to have high credibility. For example, if monthly calibration is adopted, the regression period of credibility is 1 month. Moreover, with the aging of the monitoring device or the development of factors such as sensor poisoning, the credibility of the monitoring device will gradually decrease.

[0057] In another embodiment, For oil-immersed equipment, the test items are monitoring and diagnosis of dissolved gas types in oil, and the reliability regression period is 1 month; among them, For high-quality oil-immersed equipment, consider the performance degradation time of more than 8 years including 8 years; For inferior oil-immersed equipment, allow for 2 years of degradation.

[0058] In another embodiment, For the devices used in offline tests, by default, testing is carried out once a year; When the device used in the offline test encounters short circuit, discharge, overheating, or oil replenishment events, the previous historical data is no longer of reference value and offline testing must be repeated.

[0059] In another embodiment, The following model is used to characterize the change in the credibility of the diagnosis results of different expert systems over time:

[0060] Cn is the credibility of the nth expert system, An is the normalized coefficient of credibility, kn is the sensitivity system, N is the calibration period (N>1), T is the credibility decline period (in days), and t is the actual running time.

[0061] In another embodiment, For the integration of credibility, in terms of oil sampling DGA and online DGA monitoring: The DGA test of oil sampling uses a desktop instrument with high accuracy and a good traceability guarantee system, so the absolute value of the measurement is more reliable. However, when the relevant equipment has suffered from short circuit, overheating or internal discharge defects, the previous test data cannot be used as a basis for judgment, but can be used as a reference for the benchmark value; Online monitoring equipment can provide continuous detection data, which is very helpful for the diagnosis of development trends. However, its measurement accuracy will gradually deviate from the accurate value as the running time increases, and regular calibration is required. Therefore, trend judgment should be used as the main reference. In order to organically integrate the calibration cycle, based on the objective law of the accuracy drift of different detection devices corresponding to the relevant expert systems for the same substation equipment object, the following credibility fusion model is used for processing:

[0062] Where C is the comprehensive data credibility, mi(t) is the credibility weight of the i-th detection device that changes with time, Ci(t) is the credibility of the i-th detection device that changes with time, t is time, and M is the number of all detection devices; Credibility fusion is to obtain comprehensive data credibility, which means: for the same substation equipment object, the fused credibility reflected by the results of the different detection devices, the fused credibility is associated with time, and the fused credibility is further normalized to determine the weights in the differentiated weight system. For example, in the example of step S304 above: a higher weight (for example, 0.7) is assigned to the diagnostic results of the oil chromatography monitoring system, while a lower weight (for example, 0.3) is assigned to the partial discharge monitoring system, and the weights are further optimized as follows: the fused credibility is further normalized to determine the weights in the differentiated weight system.

[0063] See also Figure 3 , which takes DGA data online monitoring and offline detection as an example, illustrates the situation of obtaining the comprehensive data credibility after fusion. It can be understood that the credibility of the fused data means that the credibility of the relevant detection device is improved through the present invention.

[0064] It should be noted that the integration of credibility and the retest mentioned above are all aimed at obtaining meaningful, reliable field data and diagnostic results within a specific time section as training sets, where field data is equivalent to input and diagnostic results are equivalent to output. The neural network model is trained with the input, output and differentiated value system determined in this way.

[0065] In another embodiment, The weights corresponding to each expert system can be determined by regression calculation and normalization.

[0066] In another embodiment, through practical actual cases, based on the existing event library, the optimal weight is obtained by means of data mining. For example, data mining involves analyzing historical data such as historical data of field data, historical data of diagnosis results, and detection modes of detection devices and / or relevant modes of expert systems, as well as prediction trends of expert systems.

[0067] In another embodiment, the weights of the expert systems may further represent the adaptability factors of the expert systems to different diagnostic targets; for example, For UHV GIS equipment, internal partial discharge is an early sign of failure. Usually, a combined diagnosis method of UHF and ultrasound is used. The UHF method is highly sensitive but difficult to resist interference, while ultrasound is reliable but has relatively low sensitivity and a short effective distance. For UHV transformers, partial discharge is also important, but the result it brings is not sudden, so it is not necessarily necessary to take corresponding measures immediately. However, the analysis results of gas in oil are more important.

[0068] In another embodiment, the weight is further optimized by mapping the sensing technology or detection technology to the common fault domain of the UHV equipment, thereby determining the mapping relationship between the sensing technology and the diagnosis result of the expert system corresponding to the weight, such as Figure 4 As shown. It should be noted that the present invention classifies electrical equipment detection technology according to partial discharge, chemical detection, micro-water (insulation dryness) detection, vibration noise and physical detection, but is not limited to the listed classifications. The fields referred to in the present invention mainly include the fault diagnosis needs of major electrical equipment, especially ultra-high voltage equipment, which are classified according to the urgency of equipment and diagnosis, and are compared to the "emergency" and "outpatient" of the medical system, forming six types of application fields. The present invention proposes these types, but is not limited to them.

[0069] In another embodiment, for the credibility corresponding to each expert system diagnosis result, the initial value of the corresponding weight can be determined by the mapping relationship between the detection technology and the application field. For example, the initial value of the weight included in the applicable relationship is 1, and the initial value of the weight not included in the applicable relationship is 0. In the subsequent training process, the deviation range is specified to be no more than 20% of the initial value, and the sum of the weights is positive and less than or equal to 1. Under such conditions, a self-learning method for the weight can be formed. It can be understood that if the weight does not need to be self-learned again, the initial value of the weight is used as the value determined by each weight in the differentiated weight system, then the field data is equivalent to the input, and the diagnosis result is equivalent to the output. The neural network model is trained with this input, output and differentiated value system. If the weight needs to be self-learned, the final value of the weight after self-learning is used as the value determined by each weight in the differentiated weight system, then the field data is equivalent to the input, and the diagnosis result is equivalent to the output. The neural network model is trained with this input, output and differentiated value system.

[0070] In another embodiment, When the diagnostic results are used as the output of the trained neural network model, the diagnostic results of each expert system can be further regularized: by using a unified interface model for modeling description, the semantic unification of multiple systems can be achieved. This unification process requires the diagnostic results to be unified to the IEC61970 (DL / T 890.1) standard system using modeling tools. This modeling process can be implemented using the CIM model, so that the neural network model can be regarded as an intelligent device or virtual device that provides intelligent services in a substation.

[0071] In another embodiment, the modeling process includes the following steps: Definition of CIM model: Based on the unified interface model, define appropriate CIM models to represent the diagnostic conclusions generated by different expert systems. This includes creating CIM categories, attributes, and relationships to meet the needs of diagnostic information.

[0072] Mapping to CIM Model: Use CIM modeling tools to map the Unified Interface Model to the CIM model. This involves mapping the elements of the Unified Interface Model to CIM categories and attributes, ensuring that the information can be consistently represented in the CIM model.

[0073] Data import and export: Implement data import and export functions for each expert system to interact its diagnostic conclusions with the representation of the CIM model.

[0074] The following is a specific XML example: <?xml version="1.0" encoding="UTF-8"?> <scl xmlns="http: / / www.iec.ch / 61850 / 2003 / SCL" xmlns:sac="http: / / www.sac-china.com" xmlns:xsi="http: / / www.w3.org / 2001 / XMLSchema-instance" xsi:schemalocation="http: / / www.iec.ch / 61850 / 2003 / SCL SCL.xsd"> <header id="XRSCL" version="1.0.0" revision="1" toolID="AMITool" nameStructure="IEDName">< / header> <communication> <subnetwork name="S1" type="8-MMS"> <connectedap apname="P1" iedname="JFJCGW1001"> <address> 172.20.2.47 < / address> < / connectedap> < / subnetwork> < / communication> <ied configversion="1.0" desc="1000kV长泰Ⅰ线路高抗A相" manufacturer="上海某某有限公司" name="JFJC1" type="S61850"> <services> <dynassociation / > <getdirectory / > <getdataobjectdefinition / > <dataobjectdirectory / > <getdatasetvalue / > <setdatasetvalue / > <datasetdirectory / > <confdataset max="1" modify="false" maxAttributes="200" / > <dyndataset max="42" / > <readwrite / > <confreportcontrol bufConf="false" max="5" / > <getcbvalues / > <reportsettings bufTime="Dyn" intgPd="Dyn" optFields="Dyn" resvTms="true" rptID="Dyn" trgOps="Dyn" / > <conflns fixLnInst="true" fixPrefix="true" / > <goose max="0" / > <gsse max="0" / > <filehandling / > < / services> <accesspoint name="P1"> <server> <authentication none="true" / > <ldevice desc="监控设备" inst="MONT"> <ln0 lntype="JFJCGW1001.LLN0" lnclass="LLN0" inst=""> <dataset name="dsDin1" desc="新数据集"> <fcda ldInst="MONT" lnClass="GGIO" lnInst="1" doName="YX1" fc="ST" / > <fcda ldInst="MONT" lnClass="GGIO" lnInst="1" doName="YX2" fc="ST" / > <fcda ldInst="MONT" lnClass="GGIO" lnInst="1" doName="YX3" fc="ST" / > <fcda ldInst="MONT" lnClass="GGIO" lnInst="1" doName="YX4" fc="ST" / > <fcda ldInst="MONT" lnClass="GGIO" lnInst="1" doName="YX5" fc="ST" / > <fcda ldInst="MONT" lnClass="GGIO" lnInst="1" doName="YX6" fc="ST" / > <fcda ldInst="MONT" lnClass="GGIO" lnInst="1" doName="YX7" fc="ST" / > < / dataset> <reportcontrol buftime="200" buffered="true" confrev="1" datset="dsDin1" intgpd="10000" name="brcbDn1p" rptid="JFJCGW1001MONT / LLN0$BR$brcbDn1p"> <trgops dchg="true" dupd="false" gi="true" period="true" qchg="true" / > <optfields configRef="true" dataRef="true" dataSet="true" entryID="true" reasonCode="true" seqNum="true" timeStamp="true" / > <rptenabled max="5" / > < / reportcontrol> < / ln0> <ln lnType="JFJCGW1001.LPHD" lnClass="LPHD" inst="" desc="物理设备信息" / > <ln lntype="JFJCGW1001.GGIO1" lnclass="GGIO" inst="1" desc="1000kV长泰Ⅰ线路高抗A相"> <doi name="YX1" desc="通道1放电类型外部"> <dai name="dU"> <val> Channel 1 discharge type external< / val> < / dai> < / doi> <doi name="YX2" desc="通道1放电类型悬浮"> <dai name="dU"> <val> Channel 1 discharge type suspension< / val> < / dai> < / doi> <doi name="YX3" desc="通道1放电类型沿面"> <dai name="dU"> <val> Channel 1 discharge type surface< / val> < / dai> < / doi> <doi name="YX4" desc="通道1放电类型内部绝缘"> <dai name="dU"> <val> Channel 1 discharge type internal insulation< / val> < / dai> < / doi> <doi name="YX5" desc="通道1放电类型油纸"> <dai name="dU"> <val> Channel 1 discharge type oil paper< / val> < / dai> < / doi> <doi name="YX6" desc="通道1放电类型杂质"> <dai name="dU"> <val> Channel 1 discharge type impurities< / val> < / dai> < / doi> <doi name="YX7" desc="通道1放电类型噪声"> <dai name="dU"> <val> Channel 1 discharge type noise< / val> < / dai> < / doi> < / ln> < / ldevice> < / server> < / accesspoint> < / ied> <datatypetemplates> <lnodetype id="JFJCGW1001.LLN0" lnclass="LLN0" desc="逻辑节点0"> <do name="Mod" type="JFJCGW1001.Mod" desc="模式" / > <do name="Beh" type="JFJCGW1001.Beh" desc="行为" / > <do name="Health" type="JFJCGW1001.Health" desc="健康" / > <do name="NamPlt" type="JFJCGW1001.NamPlt" desc="铭牌" / > < / lnodetype> <lnodetype id="JFJCGW1001.LPHD" lnclass="LPHD" desc="物理设备信息"> <do name="PhyHealth" type="JFJCGW1001.Health" / > <do name="PhyNam" type="JFJCGW1001.NamPlt" / > <do name="Proxy" type="JFJCGW1001.SPS" / > < / lnodetype> <lnodetype id="JFJCGW1001.GGIO1" lnclass="GGIO"> <do name="Mod" type="JFJCGW1001.Mod" desc="模式" / > <do name="Beh" type="JFJCGW1001.Beh" desc="行为" / > <do name="Health" type="JFJCGW1001.Health" desc="健康" / > <do name="NamPlt" type="JFJCGW1001.NamPlt" desc="铭牌" / > <do name="YX1" type="JFJCGW1001.SPS" desc="通道1放电类型外部" / > <do name="YX2" type="JFJCGW1001.SPS" desc="通道1放电类型悬浮" / > <do name="YX3" type="JFJCGW1001.SPS" desc="通道1放电类型沿面" / > <do name="YX4" type="JFJCGW1001.SPS" desc="通道1放电类型内部绝缘" / > <do name="YX5" type="JFJCGW1001.SPS" desc="通道1放电类型油纸" / > <do name="YX6" type="JFJCGW1001.SPS" desc="通道1放电类型杂质" / > <do name="YX7" type="JFJCGW1001.SPS" desc="通道1放电类型噪声" / > < / lnodetype> <dotype id="JFJCGW1001.Mod" cdc="INC" desc="模式"> <da name="stVal" fc="ST" bType="Enum" type="Mod" dchg="true" / > <da name="q" fc="ST" bType="Quality" qchg="true" / > <da name="t" fc="ST" bType="Timestamp" / > <da name="ctlModel" fc="CF" bType="Enum" type="ctlModel" / > < / dotype> <dotype id="JFJCGW1001.Beh" cdc="INS" desc="行为"> <da name="stVal" fc="ST" bType="Enum" type="Beh" dchg="true" / > <da name="q" fc="ST" bType="Quality" qchg="true" / > <da name="t" fc="ST" bType="Timestamp" / > < / dotype> <dotype id="JFJCGW1001.Health" cdc="INS" desc="健康"> <da name="stVal" fc="ST" bType="Enum" type="Health" dchg="true" / > <da name="q" fc="ST" bType="Quality" qchg="true" / > <da name="t" fc="ST" bType="Timestamp" / > < / dotype> <dotype id="JFJCGW1001.NamPlt" cdc="DPL" desc="铭牌"> <da name="vendor" fc="DC" bType="VisString255" / > <da name="swRev" fc="DC" bType="VisString255" / > < / dotype> <dotype id="JFJCGW1001.MV_float" cdc="MV" desc="带时标的浮点数测量值"> <da name="mag" fc="MX" bType="Struct" type="JFJCGW1001.MV_float.mag" / > <da name="q" fc="MX" bType="Quality" qchg="true" / > <da name="t" fc="MX" bType="Timestamp" / > <da name="dU" fc="DC" bType="Unicode255" / > < / dotype> <dotype id="JFJCGW1001.MV_int" cdc="MV" desc="带时标的整数测量值"> <da name="mag" fc="MX" bType="Struct" type="JFJCGW1001.MV_int.mag" / > <da name="q" fc="MX" bType="Quality" qchg="true" / > <da name="t" fc="MX" bType="Timestamp" / > <da name="dU" fc="DC" bType="Unicode255" / > < / dotype> <dotype id="JFJCGW1001.SPS" cdc="SPS" desc="单点信息"> <da name="stVal" fc="ST" bType="BOOLEAN" dchg="true" / > <da name="q" fc="ST" bType="Quality" qchg="true" / > <da name="t" fc="ST" bType="Timestamp" / > <da name="iStatus" fc="ST" bType="INT32" / > <da name="dU" fc="DC" bType="Unicode255" / > < / dotype> <dotype id="JFJCGW1001.SPC" cdc="SPC" desc="单点控制"> <da name="stVal" fc="ST" bType="BOOLEAN" dchg="true" / > <da name="q" fc="ST" bType="Quality" qchg="true" / > <da name="t" fc="ST" bType="Timestamp" / > <da name="Oper" fc="CO" bType="Struct" type="JFJCGW1001.Oper" / > <da name="ctlModel" fc="CF" bType="Enum" type="ctlModel" / > <da name="dU" fc="DC" bType="Unicode255" / > < / dotype> <datype id="JFJCGW1001.MV_float.mag" desc="浮点数"> <bda name="f" bType="FLOAT32" / > < / datype> <datype id="JFJCGW1001.MV_int.mag" desc="整数"> <bda name="i" bType="INT32" / > < / datype> <datype id="JFJCGW1001.Oper.origin"> <bda name="orCat" bType="Enum" type="orCategory" / > <bda name="orIdent" bType="Octet64" / > < / datype> <datype id="JFJCGW1001.Oper"> <bda name="ctlVal" bType="BOOLEAN" / > <bda name="origin" bType="Struct" type="JFJCGW1001.Oper.origin" / > <bda name="ctlNum" bType="INT8U" / > <bda name="T" bType="Timestamp" / > <bda name="Test" bType="BOOLEAN" / > <bda name="Check" bType="Check" / > < / datype> <enumtype id="ctlModel"> <enumval ord="0"> status-only< / enumval> <enumval ord="1"> direct-with-normal-security< / enumval> <enumval ord="2"> sbo-with-normal-security< / enumval> <enumval ord="3"> direct-with-enhanced-security< / enumval> <enumval ord="4"> sbo-with-enhanced-security< / enumval> < / enumtype> <enumtype id="orCategory"> <enumval ord="0"> not-supported< / enumval> <enumval ord="1"> bay-control< / enumval> <enumval ord="2"> station-control< / enumval> <enumval ord="3"> remote-control< / enumval> <enumval ord="4"> automatic-bay< / enumval> <enumval ord="5"> automatic-station< / enumval> <enumval ord="6"> automatic-remote< / enumval> <enumval ord="7"> Maintenance< / enumval> <enumval ord="8"> process< / enumval> < / enumtype> <enumtype id="Beh"> <enumval ord="1"> on< / enumval> <enumval ord="2"> Blocked< / enumval> <enumval ord="3"> test< / enumval> <enumval ord="4"> test / blocked< / enumval> <enumval ord="5"> off< / enumval> < / enumtype> <enumtype id="Health"> <enumval ord="1"> Ok< / enumval> <enumval ord="2"> Warning< / enumval> <enumval ord="3"> Alarm< / enumval> < / enumtype> <enumtype id="Mod"> <enumval ord="1"> on< / enumval> <enumval ord="2"> Blocked< / enumval> <enumval ord="3"> test< / enumval> <enumval ord="4"> test / blocked< / enumval> <enumval ord="5"> off< / enumval> < / enumtype> <enumtype id="SIUnit"> <enumval ord="1" / > <enumval ord="2"> m< / enumval> <enumval ord="3"> kg< / enumval> <enumval ord="4"> s< / enumval> <enumval ord="5"> A< / enumval> <enumval ord="6"> K< / enumval> <enumval ord="7"> mol< / enumval> <enumval ord="8"> cd< / enumval> <enumval ord="9"> deg< / enumval> <enumval ord="10"> rad< / enumval> <enumval ord="11"> sr< / enumval> <enumval ord="21"> Gy< / enumval> <enumval ord="22"> q< / enumval> <enumval ord="23"> °C< / enumval> <enumval ord="24"> Sv< / enumval> <enumval ord="25"> F< / enumval> <enumval ord="26"> C< / enumval> <enumval ord="27"> S< / enumval> <enumval ord="28"> H< / enumval> <enumval ord="29"> V< / enumval> <enumval ord="30"> ohm< / enumval> <enumval ord="31"> J< / enumval> <enumval ord="32"> N< / enumval> <enumval ord="33"> Hz< / enumval> <enumval ord="34"> lx< / enumval> <enumval ord="35"> Lm< / enumval> <enumval ord="36"> Wb< / enumval> <enumval ord="37"> T< / enumval> <enumval ord="38"> W< / enumval> <enumval ord="39"> Pa< / enumval> <enumval ord="41"> m2< / enumval> <enumval ord="42"> m3< / enumval> <enumval ord="43"> m / s< / enumval> <enumval ord="44"> m / s2< / enumval> <enumval ord="45"> m3 / s< / enumval> <enumval ord="46"> m / m3< / enumval> <enumval ord="47"> M< / enumval> <enumval ord="48"> kg / m3< / enumval> <enumval ord="49"> m2 / s< / enumval> <enumval ord="50"> W / m K< / enumval> <enumval ord="51"> J / K< / enumval> <enumval ord="52"> ppm< / enumval> <enumval ord="53"> 1 / s< / enumval> <enumval ord="54"> rad / s< / enumval> <enumval ord="61"> VA< / enumval> <enumval ord="62"> Watts< / enumval> <enumval ord="63"> VAr< / enumval> <enumval ord="64"> phi< / enumval> <enumval ord="65"> cos(phi)< / enumval> <enumval ord="66"> Vs< / enumval> <enumval ord="67"> V2< / enumval> <enumval ord="68"> As< / enumval> <enumval ord="69"> A2< / enumval> <enumval ord="70"> A2t< / enumval> <enumval ord="71"> V A< / enumval> <enumval ord="72"> W< / enumval> <enumval ord="73"> V A< / enumval> <enumval ord="74"> V / Hz< / enumval> < / enumtype> <enumtype id="multiplier"> <enumval ord="-24"> y< / enumval> <enumval ord="-21"> z< / enumval> <enumval ord="-18"> a< / enumval> <enumval ord="-15"> f< / enumval> <enumval ord="-12"> p< / enumval> <enumval ord="-9"> n< / enumval> <enumval ord="-6"> μ< / enumval> <enumval ord="-3"> m< / enumval> <enumval ord="-2"> c< / enumval> <enumval ord="-1"> d< / enumval> <enumval ord="0" / > <enumval ord="1"> da< / enumval> <enumval ord="2"> h< / enumval> <enumval ord="3"> k< / enumval> <enumval ord="6"> M< / enumval> <enumval ord="9"> G< / enumval> <enumval ord="12"> T< / enumval> <enumval ord="15"> P< / enumval> <enumval ord="18"> E< / enumval> <enumval ord="21"> Z< / enumval> <enumval ord="24"> Y< / enumval> < / enumtype> < / datatypetemplates> < / scl> In another embodiment, In step S500, the output of the neural network model (i.e., the diagnostic conclusion provided by the neural network model) is further converted into a data format supported by various commonly used power communication protocols such as the IEC61850 protocol and IEC104 through the CIM mode to ensure that the intelligent service provided by the edge device side through the neural network model can be smoothly coupled to the substation integrated automation system. After receiving the data, the integrated automation system parses it to extract the diagnostic conclusion. It can be understood that after receiving the diagnostic conclusion, the integrated automation system can apply it to various aspects of the UHV substation, such as equipment maintenance, fault detection, operation decision-making, etc. The standardized nature of the intelligent service ensures that these diagnostic conclusions can be consistent between different substations and equipment, and simplifies the integration process of the integrated automation system.

[0075] It can be understood that in the present invention, Weighting system: Design and implement a weighting system according to the classification types of partial discharge, chemical detection, micro-water (insulation dryness) detection, vibration noise and physical detection to determine the relative importance of these devices, expert systems and application areas. This is achieved by using predefined rules, empirical data or machine learning methods; Self-learning mechanism: The self-learning mechanism includes machine learning algorithms, so that the intelligent terminal device as an edge device can automatically adjust the weight system and decision-making strategy according to the changing detection values ​​and different diagnostic requirements, and make adjustments to the possible joint diagnosis conclusions and feedback information; Unified interface model: The unified interface model not only unifies field data, diagnostic results, etc. across multiple expert systems, but also becomes a data exchange format that facilitates mutual communication within the substation integrated automation system. It should be noted that the unified interface model is unified to the IEC61970 (DL / T 890.1) standard system using modeling tools, which can ensure interoperability between different programs.

[0076] Therefore, the present invention can not only jointly analyze the diagnostic conclusions of multiple expert systems and give a quantitative evaluation of the degree of credibility, thereby assisting operating personnel in judging indicators and reducing the occurrence of misjudgment, but also the present invention can standardize the smart services of UHV substations to make them consistent in different environments and applications, thereby improving the operating efficiency and maintainability of UHV equipment products. Standardization also helps to reduce training costs because different substations can share similar conclusion application methods.

[0077] In the description of this specification, the description with reference to the terms "one embodiment / method", "some embodiments / methods", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment / method or example are included in at least one embodiment / method or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment / method or example. Moreover, the specific features, structures, materials or characteristics described may be combined in a suitable manner in any one or more embodiments / methods or examples. In addition, those skilled in the art may combine and combine the different embodiments / methods or examples described in this specification and the features of the different embodiments / methods or examples, unless they are contradictory.

[0078] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0079] It should be understood by those skilled in the art that the above embodiments are only for the purpose of clearly illustrating the present invention, and are not intended to limit the scope of the present invention. For those skilled in the art, other changes or modifications may be made based on the above disclosure, and these changes or modifications are still within the scope of the present invention.

Claims

1. A multi-system joint diagnosis method for ultra-high voltage substation equipment status evaluation, characterized in that: The method comprises the following steps: S100, collects the diagnostic results of various expert systems including daily inspection, online monitoring, live detection and offline testing of UHV substation equipment; S200, determining the time section of diagnosis for different time intervals, and establishing a unified interface model that reflects the uniformity of time and data for data from different data sources; S300, using the existing event library of the UHV substation equipment, determining the accuracy and adaptability of the diagnostic results of each expert system for daily inspection, online monitoring, live detection and offline testing of the UHV substation equipment, and constructing a differentiated weight system for different equipment problems, wherein each weight in the differentiated weight system is a weight of the credibility of the diagnostic results of each expert system; S400: Based on the differentiated weight system and unified interface model, a neural network model is trained and obtained by using machine learning technology, and the neural network model is deployed on the edge device side to realize multi-system joint diagnosis of UHV substation equipment status evaluation.

2. The method according to claim 1, characterized in that The UHV substation equipment includes: UHV transformers, UHV reactors, UHV GIS and other main equipment.

3. The method according to claim 1, characterized in that After step S400, the following steps are also included: S500 and the edge device side further utilize the power communication protocol to connect with the substation integrated automation system to become a standard form of UHV substation smart service, so that the edge device side can be accessed at any time on the platform of the substation integrated automation system.

4. The method according to claim 1, characterized in that: Step S100 also includes the following steps: Receive various field data generated by various expert systems, including: data with continuous time series characteristics generated by the online monitoring system, irregular data generated by manual inspection and automatic inspection, and preventive status test data with a long period (more than 1 year).

5. The method according to claim 4, characterized in that Step S100 also includes the following steps: Further collect and manage the on-site equipment information and automatically attribute the various on-site data to the corresponding UHV substation equipment.

6. The method according to claim 5, characterized in that The equipment information includes basic information of major equipment such as UHV transformers, UHV reactors and UHV GIS.

7. The method according to claim 1, characterized in that The diagnostic results of each expert system include but are not limited to: the diagnostic conclusions of dissolved gas in oil (DGA) of oil-immersed equipment, the diagnostic conclusions of fast-reactive hydrogen detection devices, the diagnostic conclusions of comprehensive monitoring devices (partial discharge, ultrasound, vibration), the diagnostic conclusions of GIS gas composition analysis, etc.

8. The method according to claim 1, characterized in that In step S200, the time section of diagnosis is determined based on the equipment characteristics of the UHV substation equipment, the detection items and the credibility of the diagnostic results of the corresponding expert system, so as to provide meaningful, reliable field data and diagnostic results within a specific time section as a training set for the training of the neural network model, wherein the field data is equivalent to the input and the diagnostic result is equivalent to the output, and the input, output and differentiated value system determined in this way are used to train the neural network model.

Images