A multi-source fusion and meta-learning driven power equipment intelligent diagnosis method and system

Abstract

The application relates to the field of artificial intelligence and discloses a power equipment intelligent diagnosis method and system driven by multi-source fusion and meta learning. The method comprises the following steps: collecting vibration, current and temperature and other multi-source operation characteristic data of equipment; performing time-space alignment and normalization processing on the collected data, and extracting a standardized characteristic vector; performing fusion analysis on the multi-source characteristics through a dynamic weight fusion model, combining a fault knowledge base to perform characteristic matching, and realizing intelligent diagnosis output of equipment fault types, predicted downtime and risk levels; when the diagnosis confidence exceeds a threshold value or actual fault confirmation is achieved, triggering a meta learning module to optimize and update the fusion weight; and the diagnosis result is displayed through an edge terminal and fed back to a cloud platform, so that the knowledge base and model performance are further improved. The application has self-adaptability and intelligence, can improve the accuracy and response efficiency of power equipment fault diagnosis, and is suitable for various scenes such as transformer substations and distribution networks.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence, in particular, to a multi-source fusion and meta-learning driven power equipment intelligent diagnosis method and system. BACKGROUND

[0002] With the improvement of the digitalization level of the power system, the fault diagnosis technology of power equipment (such as transformers, circuit breakers, generators, etc.) has increasingly become the key to ensuring the safe and stable operation of the power grid. The traditional power equipment fault diagnosis method mainly relies on single sensor data (such as vibration, current or temperature) and static analysis models based on rules or thresholds, lacking the collaborative fusion of multi-source data. For example, the high-frequency vibration signal (4-8 kHz) of early bearing wear is easily masked by noise, and it is difficult to accurately identify such weak fault features relying only on current or temperature data, resulting in a high fault omission rate

[0003] In recent years, multi-source data fusion and meta-learning technology have gradually attracted attention in the field of industrial fault diagnosis. Multi-source fusion methods can improve the completeness of fault features by integrating heterogeneous data such as vibration, current, temperature, etc.; meta-learning can quickly adapt to new tasks with a small amount of samples, reducing the dependence on labeled data. However, the multi-source data spatio-temporal alignment method is not perfect, and normalization processing can easily lead to the loss of key feature information such as high-frequency vibration; the application of meta-learning in power equipment diagnosis is still limited to simulation environments, and has not been combined with edge computing architecture, making it difficult to meet real-time requirements.

[0004] Based on the above background, the present application provides a multi-source fusion and meta-learning driven power equipment intelligent diagnosis method and system. The present application preserves the physical meaning of multi-source features through spatio-temporal alignment, realizes high-precision diagnosis by combining dynamic weight fusion and fault knowledge base matching, and optimizes model weights online using a meta-learning mechanism. The present application significantly improves the early fault detection rate and diagnosis timeliness while reducing the dependence on labeled data, providing an innovative solution for the intelligent and safe operation of power equipment. SUMMARY

[0005] The purpose of the embodiments of the present application is to provide a multi-source fusion and meta-learning driven power equipment intelligent diagnosis method and system. Through the improved dynamic weight fusion mechanism and meta-learning online optimization, the model can be quickly adapted with only a small amount of samples, significantly reducing the need for labeled data; providing an innovative solution for power equipment intelligent diagnosis with high precision, strong adaptability and low implementation cost.

[0006] To achieve the above purpose, the present application provides the following technical solutions:

[0007] In a first aspect, this invention discloses a multi-source fusion and meta-learning-driven intelligent diagnostic method for power equipment, comprising the following steps: S1, collecting multi-source operational characteristic data such as vibration, current, and temperature of the equipment; S2, preprocessing and spatiotemporally aligning the collected multi-source data to extract standardized feature vectors; S3, performing fusion analysis of multi-source features through a dynamic weight fusion model, and combining it with a fault knowledge base for feature matching to achieve intelligent diagnostic output of equipment fault type, expected downtime, and risk level; S4, when the diagnostic confidence exceeds a threshold or an actual fault is confirmed, triggering the meta-learning module to optimize and update the fusion weights; S5, displaying the diagnostic results through an edge terminal and feeding them back to the cloud platform to further improve the knowledge base and model performance.

[0008] More specifically, the above-described intelligent diagnostic method for power equipment driven by multi-source fusion and meta-learning is characterized in that step S1 includes: installing vibration sensors at key parts of the equipment to acquire three-dimensional ( axis, axis, (Shaft) vibration signal; real-time current signal of the equipment is collected through a current sensor; surface temperature distribution of the equipment is extracted through a thermal imaging camera, and temperature data of key points inside the equipment is obtained in combination with internal temperature sensors.

[0009] More specifically, the above-described method for intelligent diagnosis of power equipment driven by multi-source fusion and meta-learning is characterized in that the multi-source data preprocessing and spatiotemporal alignment in step S2 includes:

[0010] The energy parameters of the 4-8kHz frequency band were extracted from the vibration signal using the wavelet packet energy method.

[0011] The total harmonic distortion (THD) index was extracted from the current data using dq-axis harmonic analysis.

[0012] The thermal imaging image is processed to extract the highest surface temperature.

[0013] Vibration feature vectors are obtained by aligning windows using a unified timestamp. Current eigenvector Temperature eigenvector ;

[0014] Each feature has independent diagnostic significance in the physical system. No normalization is performed during spatiotemporal alignment; the following information is retained respectively:

[0015] Vibration characteristics retain the RMS value and peak acceleration within the window; current characteristics include instantaneous values ​​and harmonic amplitudes; temperature characteristics include the highest surface temperature, the temperature of the internal measuring point, and the temperature difference.

[0016] More specifically, the above-described intelligent diagnostic method for power equipment driven by multi-source fusion and meta-learning is characterized in that, in step S3, the dynamic weight fusion model for fusion analysis of multi-source features is:

[0017] ,

[0018] in , , These are learnable dynamic weights for vibration characteristics, current characteristics, and temperature characteristics, respectively. for Activation function This represents the diagnostic confidence level after fusion.

[0019] Fusion features Perform cosine similarity matching with fault templates in the fault knowledge base:

[0020] ,

[0021] in, Indicates the first Template vectors for each fault mode. Output the fault type with the highest similarity and its similarity. When the similarity exceeds a set similarity threshold, it is diagnosed as the corresponding fault type. The set threshold ranges from 0.85 to 0.95, preferably 0.90.

[0022] According to the fusion output Based on the fault model matching results, the equipment status is divided into different risk levels.

[0023] The risk level classification criteria are as follows:

[0024] Equipment risk levels are classified into the following three categories:

[0025] (Low Risk): Minor anomalies or initial signs of failure have been detected, which will not affect operation for the time being, but continuous monitoring is required;

[0026] (Medium risk): The fault characteristics are obvious but the system is still unstable. It is recommended to handle it during the planned maintenance window. It carries a certain operational risk.

[0027] (High Risk): High probability of failure, expected to fail in a short period of time, requiring immediate shutdown and maintenance.

[0028] More specifically, the multi-source fusion and meta-learning driven power equipment intelligent diagnosis method described above is characterized in that the predicted downtime in step S3 is calculated based on the matching results of the fusion features and the same type of historical cases in the fault knowledge base, the current risk level and the diagnosis confidence.

[0029]

[0030] wherein is the average failure time of the matching type of fault history, is the risk level factor ( ), and is the adjustment coefficient, is the confidence function (such as linear or exponential form).

[0031] More specifically, the multi-source fusion and meta-learning driven power equipment intelligent diagnosis method described above is characterized in that in step S4, the meta-learning optimization module is triggered under the following conditions: when the diagnosis confidence or when an actual verification failure occurs, the meta-learning module is enabled to optimize the fusion weight:

[0032] ,

[0033] wherein the loss function is defined by the mean square error:

[0034] ,

[0035] wherein represents the target confidence corresponding to the fusion feature sample.

[0036] More specifically, the multi-source fusion and meta-learning driven power equipment intelligent diagnosis method described above is characterized in that the diagnosis results in step S5 include: fault type alarm, device risk level display and predicted downtime output, all data are fed back to the cloud platform through the edge terminal for continuous model optimization and historical data archiving.

[0037] In a second aspect, the present application discloses a system for implementing a multi-source fusion and meta-learning driven power equipment intelligent diagnosis method, characterized in that the system comprises: a multi-source data acquisition unit, an edge computing unit, a cloud platform management unit and a human-computer interface unit.

[0038] The multi-source data acquisition unit is used to acquire vibration signals, current signals, internal temperatures and thermal imaging image data during equipment operation, and extract temperature features from the thermal imaging image through an image processing module.

[0039] An edge computing unit connected to the multi-source data acquisition unit, the edge computing unit comprising:

[0040] A space-time alignment and feature extraction module for preprocessing and extracting corresponding features from the collected data;

[0041] An intelligent fusion diagnosis module for calculating diagnosis output according to a fusion model and matching with fault templates stored in a fault knowledge base to realize fault type identification, predicted downtime prediction and risk level division;

[0042] A meta-learning weight updating module for online updating dynamic weights according to actual diagnosis results and a preset loss function.

[0043] A cloud platform management unit for storing a fault knowledge base and realizing data interaction and model issuing between the edge computing unit, while supporting global optimization of the model by the meta-learning engine.

[0044] A man-machine interface unit for displaying and recording diagnosis results, providing fault alarm, device state display, data recording and historical data query functions, and realizing end-to-end real-time monitoring and diagnosis feedback by interacting with the cloud platform.

[0045] Compared with the prior art, the present application has the beneficial effects that: the present application retains key features such as high-frequency vibration, current harmonic, temperature gradient and the like of physical dimension by multi-source data fusion combined with space-time alignment technology, thereby improving fault diagnosis precision; the present application introduces a dynamic weight mechanism of self-learning to improve diagnosis accuracy of the model under different working conditions without manual adjustment of parameters. The present application can significantly improve the accuracy, timeliness and economy of power equipment fault diagnosis, and provides an innovative solution for safe operation and maintenance of smart grids. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments of the present application. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope, and for those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0047] Figure 1 A flowchart of a multi-source fusion and meta-learning driven intelligent diagnosis method of power equipment provided by the present application.

[0048] Figure 2 An architecture diagram of a system of a multi-source fusion and meta-learning driven intelligent diagnosis method of power equipment provided by the present application. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application. It should be noted that similar reference numerals and letters represent similar items in the drawings below, and therefore, once an item is defined in one drawing, it need not be further defined and explained in subsequent drawings.

[0050] The terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises a" does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0051] The terms "first", "second", and the like are merely used to distinguish one entity or operation from another entity or operation, and cannot be understood as indicating or implying relative importance, and cannot be understood as requiring or implying any such actual relationship or order between the entities or operations.

[0052] The basic idea of the present application is to collect multi-source operating characteristic data such as vibration, current and temperature of the equipment; to pre-process and space-time align the collected multi-source data, extract standardized feature vectors; to analyze and fuse the multi-source features through a dynamic weight fusion model, combine with the fault knowledge base for feature matching, realize intelligent diagnosis output of equipment fault type, predicted downtime and risk level; when the diagnosis confidence exceeds the threshold or the actual fault is confirmed, trigger the meta-learning module to optimize and update the fusion weight; display the diagnosis result through the edge terminal and feed back to the cloud platform, further improve the knowledge base and model performance. Through multi-source data fusion, combined with space-time alignment technology, the key features such as high-frequency vibration, current harmonic, temperature gradient of physical dimension are retained, and the fault diagnosis accuracy is improved; by introducing the dynamic weight mechanism of self-learning, the diagnosis accuracy of the model under different working conditions is improved, and manual parameter adjustment is not required.

[0053] As shown in Figure 1 The present application provides a multi-source fusion and meta-learning driven intelligent diagnosis method for power equipment, comprising the following steps:

[0054] S1, collecting multi-source operating characteristic data such as vibration, current and temperature of the equipment;

[0055] Install vibration sensors at key parts of the equipment to obtain three-dimensional (X, Y, Z) vibration data during operation. Shaft, Shaft, The vibration signal of the shaft is collected; the real-time current signal of the equipment is collected by a current sensor; the temperature distribution on the surface of the equipment is extracted by a thermal imaging camera, and the temperature data of key points in the equipment are obtained by combining internal temperature sensors.

[0056] S2, pre-processing and space-time alignment of the collected multi-source data, and extraction of standardized feature vectors;

[0057] The wavelet packet energy method is used to extract the energy parameters of the 4-8 kHz frequency band of the vibration signal;

[0058] The d-q axis harmonic analysis is used to extract the total harmonic distortion (THD) index of the current data;

[0059] The thermal imaging image is processed to extract the highest surface temperature.

[0060] Through unified timestamp window alignment, vibration feature vectors , current feature vectors , and temperature feature vectors are obtained.

[0061] Each type of feature has independent diagnostic significance in the physical system, and no normalization processing is performed during the space-time alignment process, and the following information is retained respectively:

[0062] The vibration feature retains the RMS value and peak acceleration in the window; the current feature includes the instantaneous value and harmonic amplitude; and the temperature feature includes the highest surface temperature, the temperature of the internal temperature measurement point, and the temperature difference.

[0063] S3, multi-source feature fusion analysis is performed by a dynamic weight fusion model, and feature matching is performed in combination with a fault knowledge base to realize intelligent diagnosis output of the equipment fault type, predicted downtime, and risk level;

[0064] 1) Dynamic weight fusion of multi-source feature fusion analysis, dynamic weight fusion model:

[0065] ,

[0066] Among them , , are the learnable dynamic weights of the vibration feature, the current feature, and the temperature feature, is an activation function, represents the fused diagnosis confidence.

[0067] 2) The fused feature is matched with the fault templates in the fault knowledge base in terms of cosine similarity:

[0068] ,

[0069] wherein, is the template vector representing the th failure mode. The output is the failure type with the highest similarity and its similarity, when the similarity exceeds a set similarity threshold, the corresponding failure type is diagnosed; the set threshold value ranges from 0.85 to 0.95, preferably 0.90.

[0070] 3) According to the fusion output and the failure model matching result, the device state is divided into different risk levels,

[0071] The risk level division standard is:

[0072] The device risk level is divided into the following three categories:

[0073] (Low risk): slight abnormalities or initial failure signs are detected, which do not affect operation for the time being and can be continuously monitored;

[0074] (Medium risk): the failure characteristics are obvious but not stable, and it is recommended to be handled in the planned maintenance window, with certain operation risk;

[0075] (High risk): the failure probability is high, and failure is expected to occur in a short time, which requires immediate shutdown for maintenance.

[0076] 4) Based on the matching result of the fusion features and the same type of historical cases in the failure knowledge base, the current risk level and the diagnostic confidence, the predicted downtime is calculated:

[0077]

[0078] wherein is the average failure time of the th matched failure history, is the risk level factor ( ), and are adjustment coefficients, is a confidence function (such as linear or exponential form).

[0079] S4, when the diagnostic confidence exceeds the threshold or the actual failure is confirmed, the meta-learning module is triggered to optimize and update the fusion weight;

[0080] The meta-learning optimization module is triggered under the following conditions: when the diagnostic confidence or the actual verification fails, the meta-learning module is enabled to optimize the fusion weight:

[0081] ,

[0082] wherein the loss function is defined by mean square error:

[0083] ,

[0084] wherein represents the target confidence corresponding to the fusion feature sample.

[0085] S5, the diagnostic result is displayed through the edge terminal and fed back to the cloud platform, and the knowledge base and the model performance are further improved.

[0086] The diagnostic result includes: fault type alarm, device risk level display and predicted downtime output, and all data are fed back to the cloud platform through the edge terminal for continuous model optimization and historical data archiving.

[0087] As shown in Figure 2 , the present application provides a multi-source fusion and meta-learning driven power equipment intelligent diagnosis method system architecture diagram, which comprises: a multi-source data acquisition unit, an edge computing unit, a cloud platform management unit and a man-machine interface unit.

[0088] The multi-source data acquisition unit is used for acquiring vibration signals, current signals, internal temperatures and thermal imaging image data during equipment operation, and extracting temperature features from the thermal imaging image through an image processing module.

[0089] The edge computing unit is connected with the multi-source data acquisition unit, and comprises:

[0090] The time-space alignment and feature extraction module is used for pre-processing the collected data and extracting corresponding features;

[0091] The intelligent fusion diagnosis module is used for calculating diagnostic output according to the fusion model, and matching the stored fault templates in the fault knowledge base, so as to realize fault type recognition, predicted downtime prediction and risk level division;

[0092] The meta-learning weight updating module is used for updating the dynamic weight online according to the actual diagnostic result and the preset loss function.

[0093] The cloud platform management unit is used for storing the fault knowledge base, realizing data interaction and model issuing between the edge computing unit, and supporting global optimization of the model by the meta-learning engine.

[0094] The man-machine interface unit is used for displaying and recording the diagnostic result on site, providing fault alarm, device state display, data recording and historical data query functions, and realizing end-to-end real-time monitoring and diagnostic feedback by interacting with the cloud platform.

[0095] The above merely provides an example of the present application, and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A multi-source fusion and meta-learning driven intelligent diagnosis method for power equipment, characterized in that, Comprise the following steps: S1, collect the vibration, current and temperature of the equipment Multisource operation characteristic data; S2, the collected multisource data is pretreated and spatiotemporal alignment, and the standardized feature vector is extracted; S3, the multisource characteristics are analyzed by a dynamic weight fusion model, the feature matching is carried out in combination with a fault knowledge base, and intelligent diagnosis outputs such as equipment fault type, predicted downtime and risk level are realized; S4, when the diagnosis confidence exceeds the threshold or the actual fault is confirmed, the meta-learning module is triggered to optimize and update the fusion weight; S5, the diagnosis result is displayed through the edge terminal and fed back to the cloud platform, and the knowledge base and model performance are further improved; In step S3, the dynamic weight fusion model for fusion analysis of multi-source features is as follows: ; wherein , , are respectively learnable dynamic weights of the vibration feature, the current feature and the temperature feature, is an activation function, represents the fused diagnosis confidence. Fused features Cosine similarity matching with fault templates in the fault knowledge base, vibration feature vector, current feature vector, temperature feature vector: ; wherein, represents the template vector of the th failure mode; the output is the most similar failure type and its similarity, and when the similarity exceeds the set similarity threshold, the corresponding failure type is diagnosed; the set threshold value ranges from 0.85 to 0.

95. According to the fusion output and the fault model matching result, the device state is divided into different risk levels, The risk level classification criteria are: The device risk level is classified into the following three categories: Indicates low risk: minor abnormalities or initial signs of failure are detected, which do not affect operation at the moment and can be continuously monitored; Medium risk: The failure feature is obvious but not stable, it is recommended to be handled in the planned maintenance window, and has certain operation risk; High risk: high probability of failure, expected to occur in a short time, immediate shutdown for repair; The predicted downtime in the step S3 is calculated based on the matching result of the fusion features and the same type of historical cases in the fault knowledge base, the current risk level and the diagnosis confidence: ; wherein is the average failure time of the matching historical case of the same type of fault, is the average failure time of the matching historical case of the same type of fault, is the risk level factor, and is the adjustment coefficient, is the confidence function, and the confidence function is linear or exponential. In the step S4, the meta-learning optimization module is triggered under the following conditions: when the diagnosis confidence or when the actual verification fails, the meta-learning module is enabled to optimize the fusion weight: ; wherein the loss function is defined by the mean square error: ; wherein represents the target confidence corresponding to the fusion feature sample.

2. The multi-source fusion and meta-learning driven intelligent diagnosis method for power equipment according to claim 1, characterized in that, The step S1 comprises: installing a vibration sensor at a key part of the equipment, acquiring a three-dimensional vibration signal during operation, three-dimensional directions shaft, shaft, shaft; acquiring a real-time current signal of the equipment through a current sensor; extracting a surface temperature distribution of the equipment through a thermal imaging camera, and acquiring temperature data of key points inside the equipment in combination with an internal temperature sensor.

3. The multi-source fusion and meta-learning driven intelligent diagnosis method for power equipment according to claim 1, characterized in that, The multisource data in the step S2 is pretreated and spatiotemporal alignment, and comprises: The wavelet packet energy method is used for extracting the energy parameters of the 4-8 kHz frequency band of the vibration signal; The d-q axis harmonic analysis is used for extracting the total harmonic distortion index of the current data; The image processing is performed on the thermal imaging image, and the surface maximum temperature is extracted; Window alignment by unified timestamp, to get vibration feature vector , current feature vector , temperature feature vector ; Each type of feature has independent diagnostic significance in the physical system, and no normalization processing is performed in the spatiotemporal alignment process, and the following information is respectively reserved: The vibration characteristics reserve the RMS value and peak acceleration in the window; the current characteristics include instantaneous value and harmonic amplitude; the temperature characteristics include surface maximum temperature, internal temperature and temperature difference.

4. The multi-source fusion and meta-learning driven intelligent diagnosis method for power equipment according to claim 1, characterized in that, The diagnosis result in the step S5 comprises: fault type alarm, equipment risk level display and predicted downtime output, all data are fed back to the cloud platform through the edge terminal, and are used for continuous model optimization and historical data archiving.

5. A system for implementing the multi-source fusion and meta-learning driven power equipment intelligent diagnosis method of any one of claims 1-3, characterized in that, The system comprises: a multisource data acquisition unit, an edge computing unit, a cloud platform management unit and a man-machine interface unit; The multisource data acquisition unit is used for acquiring the vibration signal, current signal, internal temperature and thermal imaging image data during equipment operation, and extracting temperature characteristics from the thermal imaging image through an image processing module; The edge computing unit is connected with the multisource data acquisition unit, and comprises: The spatiotemporal alignment and feature extraction module is used for pretreating each data and extracting corresponding features; The intelligent fusion diagnosis module is used for calculating diagnosis output according to the fusion model, and matching the fault templates stored in the fault knowledge base, realizing fault type identification, predicted downtime prediction and risk level division; The meta-learning weight updating module is used for updating the dynamic weight online according to the actual diagnosis result and the preset loss function; The cloud platform management unit is used for storing the fault knowledge base, realizing data interaction and model issuing between the edge computing unit, and supporting global optimization of the model by the meta-learning engine; The man-machine interface unit is used for displaying and recording the diagnosis result on site, providing fault alarm, equipment state display, data recording and historical data query functions, and realizing end-to-end real-time monitoring and diagnosis feedback by interacting with the cloud platform.

Images