Power distribution network equipment multi-modal data defect fault intelligent analysis system

Abstract

The present application relates to the technical field of power grid inspection, and discloses a power distribution network equipment multi-modal data defect fault intelligent analysis system, which comprises: a system receiving multi-modal inspection data and infrared temperature measurement results; when the object is a power distribution transformer with A, B and C three-phase joints, the A-phase joint temperature, the B-phase joint temperature and the C-phase joint temperature at the same collection time are read; a three-phase temperature difference relationship matrix is generated in the order of A, B and C fixed rows and columns; the three-phase average temperature, the relative heat of each phase, the three-phase temperature difference total amount and the hottest phase number are synchronously generated; and the three-phase temperature difference data state is determined through zero diagonal line check sum and anti-symmetry relationship check. The rule classification module uses the three-phase temperature difference data as phase relationship evidence to participate in equipment defect level merging when the data is available, and falls back to the original rule classification process when the data is unavailable or the check fails; and the prompt template assembly module writes the three-phase temperature difference data and limits the phase position output.

Description

Technical Field

[0001] This invention relates to the field of power grid inspection technology, and more specifically, to a multimodal data defect and fault intelligent analysis system for distribution network equipment. Background Technology

[0002] Distribution network inspection is typically completed jointly by on-site inspection instruments, an operation and maintenance platform, a power intranet server, and equipment ledgers. After the inspection instruments collect audible, ultrasonic, ultraviolet, image, and infrared thermometric results, the platform archives them according to task, site, equipment, and collection time. Then, the rule engine and model service provide the equipment status. For three-phase distribution transformer terminals, maintenance personnel are more concerned with the temperature distribution between different phase terminals and the reliability of the phase position, rather than just the highest infrared temperature of a single unit.

[0003] In common implementations, the temperatures of phase A, phase B, and phase C connectors are typically entered into the database as ordinary fields. During subsequent processing, the rule engine needs to construct the phase relationships from field names, measurement point numbers, and equipment logs; the model's prompt template also relies on text descriptions to explain the phase positions. While this approach can handle general over-temperature alarms in engineering applications, when all three phases heat up synchronously with the load, when one phase is locally overheated, or when two phases are both overheated, the system needs to additionally determine the relative relationships between the phases, resulting in a longer processing chain and more potential points of error.

[0004] Furthermore, there is a mapping relationship between infrared temperature measurement points and phase identifiers in the equipment ledger. If the temperature field is entered in the wrong order, the measurement point label is lost, or a certain phase temperature is missing, subsequent rules and models may still continue to generate phase positions. For operation and maintenance scenarios that require tracing report conclusions back to original values, infrared evidence, and rule numbers, such issues will increase review costs and also affect the differentiation between common three-phase heating and localized overheating when merging defect levels. Summary of the Invention

[0005] This invention provides an intelligent analysis system for multimodal data defects and faults in power distribution network equipment, which solves the technical problems mentioned in the background art.

[0006] To achieve the above-mentioned objectives, the present invention provides the following technical solution.

[0007] The intelligent analysis system for multi-modal data defects and faults of power distribution network equipment includes a multi-source data unified access module, a three-phase temperature difference data generation module, a rule classification module, a prompt template assembly module, and a result output module. The multi-source data unified access module is used to receive multimodal inspection data and infrared temperature measurement results; The three-phase temperature difference data generation module is used to read the temperatures of phase A, phase B, and phase C at the same acquisition time when the object is a distribution transformer with three-phase terminals A, B, and C. It generates a three-phase temperature difference relationship matrix that expresses the directional temperature difference relationship between any two phases according to a fixed row and column order of A, B, and C. It generates a phase relationship field that includes the average temperature of the three phases, the relative heat of each phase, the total temperature difference of the three phases, and the number of the hottest phase. Based on the zero diagonal check and antisymmetry check of the three-phase temperature difference relationship matrix, it determines the status of the three-phase temperature difference data. The three-phase temperature difference data includes the three-phase temperature difference relationship matrix, the phase relationship field, and the status of the three-phase temperature difference data. The three-phase temperature difference data generation module sends the three-phase temperature difference data to the rule classification module and the prompt template assembly module; When the three-phase temperature difference data is available, the rule classification module uses the three-phase temperature difference data as evidence of the phase relationship of the infrared channel to participate in the merging of equipment defect levels. When the three-phase temperature difference data is unavailable or the verification fails, it reverts to the original rule classification process. The prompt template assembly module writes the three-phase temperature difference data into a fixed field of the cross-modal model and limits the phase position output of the cross-modal model; The results output module outputs a standardized sequence of defect detection results.

[0008] Compared with the prior art, the present invention has the following substantial features and significant progress: This invention utilizes the A-phase, B-phase, and C-phase junction temperatures collected at the same time to generate a three-phase temperature difference matrix in the power grid service. The average three-phase temperature, relative heat of each phase, total three-phase temperature difference, hottest phase number, and three-phase temperature difference data status are then sent to the rule grading module and the prompt template assembly module. This allows the system to distinguish between three-phase simultaneous heating, single-phase relative overheating, and two-phase simultaneous overheating, while retaining infrared temperature measurement evidence and the original multi-source inspection results. It can also revert to the original rule grading process in case of data loss or verification failure, and reduces instances where cross-modal models deviate from the phase position of the three-phase temperature difference data output. This improves the traceability of the distribution transformer terminal inspection results and the consistency of phase position expression. Attached Figure Description

[0009] Figure 1 This is a system architecture and data flow mechanism diagram of the intelligent analysis system for multimodal data defects and faults in power distribution equipment according to the present invention. Detailed Implementation

[0010] The following description is provided in conjunction with the accompanying drawings. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the invention; those skilled in the art can make equivalent substitutions or combinations for the specific implementations.

[0011] Example 1: This implementation method is applicable to scenarios involving computer-aided defect and fault analysis of equipment such as distribution transformer terminals in distribution network inspection operations. The system runs on an edge computing terminal or a power intranet server, receives multimodal inspection data and infrared temperature measurement results collected by the field inspection instrument, and associates the initial screening results from different channels, infrared temperature evidence, rule triggering information, and model-assisted analysis content into a standardized defect detection result sequence that can be traced back.

[0012] In routine power distribution network inspections, the A, B, and C phase connections of a three-phase distribution transformer have fixed phase assignments. If the temperatures of phase A, phase B, and phase C are stored as independent fields, the rule-based classification module and the cross-modal model need to re-infer the relative temperature relationships between phases based on field names and equipment records during subsequent processing. This can easily lead to inconsistencies in phase position representation when all three phases are heating up together, a single phase is relatively overheating, two phases are both too hot, or the field order is misaligned. This implementation method uses a three-phase temperature difference data generation module to form a three-phase temperature difference relationship matrix during the data access phase, making the directed temperature difference relationship between any two phases machine-readable data before it enters the rule-based classification module.

[0013] like Figure 1 As shown, the intelligent analysis system for multimodal data defects and faults in power distribution network equipment includes a multi-source data unified access module 100, a three-phase temperature difference data generation module 200, a rule classification module 300, a prompt template assembly module 400, and a result output module 600. The three-phase temperature difference data generation module 200 internally includes a three-phase temperature difference relationship matrix generation unit 210 and a three-phase temperature difference data state determination unit 220. The cross-modal model 500, in conjunction with the prompt template assembly module 400 and the result output module 600, is used to receive fixed fields and output auxiliary analysis content.

[0014] The purpose of this implementation is to enable the rule grading module to directly read phase relationship evidence and to enable the prompt template assembly module to limit the phase position output of cross-modal models.

[0015] In this embodiment, the three-phase temperature difference data includes a three-phase temperature difference relationship matrix, a phase relationship field, and the three-phase temperature difference data status. The phase relationship field includes the three-phase average temperature, the relative heat of each phase, the total three-phase temperature difference, and the hottest phase number.

[0016] Example 2: The unified multi-source data access module is used to receive multimodal inspection data and infrared temperature measurement results. Multimodal inspection data includes audible sound waveforms, ultrasonic waveforms, ultraviolet pulse data, and infrared images. In specific engineering implementations, the unified multi-source data access module can also receive the temperature spatial distribution matrix corresponding to the infrared image, acquisition time, device number, site number, task number, file size, and checksum. These additional fields are used for data archiving, indexing, verification, and backtracking.

[0017] After an inspection task begins, the multi-source data unified access module first receives the task number, site number, device number, acquisition time, and channel type. The task number distinguishes different inspection tasks, the site number identifies the site where the inspection target is located, the device number matches it with the distribution transformer record in the equipment ledger, the acquisition time determines whether data from different channels belong to the same equipment inspection, and the channel type distinguishes between audible waveforms, ultrasonic waveforms, ultraviolet pulse data, and infrared images. The multi-source data unified access module establishes an inspection data index using the task number, site number, device number, acquisition time, and channel type, enabling subsequent modules to read corresponding data based on the same device and the same acquisition time.

[0018] In terms of field format, task number, site number, and equipment number are all string fields, acquisition time uses a unified time format field, and channel type uses an enumerated field. The channel type value must include at least audible sound waveforms, ultrasonic waveforms, ultraviolet pulse data, infrared images, and infrared temperature measurement results. When establishing the inspection data index, the multi-source data unified access module uses task number, site number, equipment number, acquisition time, and channel type as joint index fields. When multiple acquisition times exist under the same equipment number, the acquisition time of the infrared temperature measurement result is used as the time reference for the A-phase connector temperature, B-phase connector temperature, and C-phase connector temperature entering the three-phase temperature difference data generation module.

[0019] When the multi-source data unified access module performs file length verification, it considers the received file's byte count to match the file size declared in the data packet as a pass condition. When performing checksum verification, it uses the CRC32 checksum written by the inspection instrument with the data packet as the checksum. The receiving end recalculates the checksum of the received file according to the same CRC32 calculation rules. If the recalculated checksum matches the checksum written with the data packet, the checksum verification is considered successful. If any of the field integrity verification, file length verification, checksum verification, or device number matching verification fails, the multi-source data unified access module marks the corresponding data record as abnormal data and prohibits that data record from entering the three-phase temperature difference data generation module.

[0020] When receiving infrared temperature measurement results, the multi-source data unified access module associates these results with the phase identifiers and infrared temperature measurement point numbers in the equipment ledger. For distribution transformers with three-phase terminals (A, B, and C), the module retains the original values ​​of the A-phase terminal temperature, B-phase terminal temperature, and C-phase terminal temperature, and sends these values ​​to the three-phase temperature difference data generation module. The A-phase terminal temperature is the surface temperature directly output by the inspection instrument at the A-phase terminal temperature measurement point; the B-phase terminal temperature is the surface temperature directly output by the inspection instrument at the B-phase terminal temperature measurement point; and the C-phase terminal temperature is the surface temperature directly output by the inspection instrument at the C-phase terminal temperature measurement point.

[0021] The multi-source data unified access module is also used to receive the results of the audible sound mini-model, ultrasound mini-model, ultraviolet mini-model, and infrared hyperthermia results and send them to the rule classification module. The audible sound mini-model, ultrasound mini-model, and ultraviolet mini-model results can be generated by conventional classification mini-models that have been trained offline, while the infrared hyperthermia results can be obtained by performing a threshold comparison between the infrared image and the temperature spatial distribution matrix.

[0022] The format validation of the multi-source data unified access module includes field integrity validation, file length validation, checksum validation, and device number matching validation. Data with missing fields, corrupted files, or mismatched device numbers is marked as abnormal data and does not directly enter the rule classification module. For data that cannot be grouped into the same acquisition time, the multi-source data unified access module retains the original acquisition time and does not forcibly merge data by modifying the acquisition time. This constraint ensures that the A-phase connector temperature, B-phase connector temperature, and C-phase connector temperature read by the three-phase temperature difference data generation module are based on the same acquisition time.

[0023] Abnormal data retains the task number, site number, device number, acquisition time, channel type, and failure reason for subsequent auditing. If the abnormal data involves the temperature of phase A, phase B, or phase C connectors, the three-phase temperature difference data generation module must not deduce the missing temperature from other phase temperatures, other channel results, or cross-modal model outputs. If the abnormal data only involves the results of audible small model, ultrasonic small model, ultraviolet small model, or infrared over-temperature, the rule classification module records the channel as a missing channel in the preset rule matrix and retains the candidate judgments for other channels.

[0024] Example 3: The three-phase temperature difference data generation module is used to read the temperatures of phase A, phase B, and phase C at the same acquisition time when the object is a distribution transformer with three-phase terminals A, B, and C. It generates a three-phase temperature difference relationship matrix that expresses the directional temperature difference relationship between any two phases according to a fixed row and column order of A, B, and C. It generates phase relationship fields including the average temperature of the three phases, the relative heat of each phase, the total temperature difference of the three phases, and the number of the hottest phase. The module determines the status of the three-phase temperature difference data based on the zero diagonal check and antisymmetry check of the three-phase temperature difference relationship matrix.

[0025] Before execution, the three-phase temperature difference data generation module first determines whether the current object is a distribution transformer with A, B, and C phase connections. This determination is based on the equipment type, phase identifier, and infrared temperature measurement point number in the equipment register. If the current object does not have A, B, and C phase connections, the three-phase temperature difference data generation module does not generate a three-phase temperature difference relationship matrix. If the current object is a distribution transformer with A, B, and C phase connections, the three-phase temperature difference data generation module reads the A-phase connection temperature, B-phase connection temperature, and C-phase connection temperature from the infrared temperature measurement results at the same acquisition time.

[0026] When the three-phase temperature difference data generation module reads the temperatures of phase A, phase B, and phase C connectors, all three temperatures are derived from infrared temperature measurements taken at the same acquisition time. The phase assignment is determined by using existing phase identifiers and infrared temperature measurement point numbers in the equipment log. This phase assignment ensures that the temperatures of phase A, phase B, and phase C connectors correspond to the actual phase A, phase B, and phase C connectors, rather than guessing based on the order in which fields are written or temporary positions in the image.

[0027] When any phase temperature is missing, the three-phase temperature difference data generation module records the three-phase temperature difference data status as unavailable, prohibits the estimation and completion of the missing phase temperature, and causes the prompt template assembly module to limit the cross-modal model to output the phase position as undetermined. This process avoids the computer program having to fill in the missing phase temperature based on other phase temperatures or results from other channels when directly acquired temperatures are unavailable.

[0028] The three-phase temperature difference data status is written to the current device record, and its value can only be either available, unavailable, or verification failed. The order of determining the three-phase temperature difference data status is as follows: First, determine whether the temperatures of phase A, phase B, and phase C connectors all exist; if any phase temperature is missing, the three-phase temperature difference data status is determined to be unavailable; if the temperatures of phase A, phase B, and phase C connectors all exist, then generate the three-phase temperature difference relationship matrix and perform zero diagonal verification and anti-symmetry relationship verification; if either the zero diagonal verification or the anti-symmetry relationship verification fails, the three-phase temperature difference data status is determined to be verification failed; if both the zero diagonal verification and the anti-symmetry relationship verification pass, the three-phase temperature difference data status is determined to be available.

[0029] When the temperatures of phase A, phase B, and phase C are all present, the three-phase temperature difference data generation module first forms a three-phase temperature column according to a fixed row and column order of A, B, and C. The phase A junction temperature is represented as follows: The temperature of phase B junction is expressed as The temperature of the C-phase junction is expressed as: The three-phase temperature series is represented as follows The three-phase temperature column is used to express the fixed arrangement relationship of the A-phase junction temperature, B-phase junction temperature, and C-phase junction temperature.

[0030] In the above formula, This represents a three-phase temperature column, with values ​​derived from the A-phase connector temperature, B-phase connector temperature, and C-phase connector temperature. The dimension of each component in the three-phase temperature column is temperature. This represents the temperature of the A-phase connector, and the value is derived from infrared thermometry results. The dimension is temperature. This indicates the temperature of the B-phase connector, and the value is derived from infrared thermometry results. The dimension is temperature. This represents the temperature of the C-phase junction, and the value is derived from infrared thermometry results. The dimension is temperature. This represents transpose and is a dimensionless operator.

[0031] The three-phase temperature difference relationship matrix generation unit 210 performs the calculation of row phase temperature minus column phase temperature for any ordered phase pair. For phase and phase The directed temperature difference between the row phase and the column phase is expressed as: .when When it is positive, it indicates that the row phase temperature is higher than the column phase temperature; when... When the value is negative, it indicates that the row phase temperature is lower than the column phase temperature; when When the value is zero, it indicates that the row phase temperature is the same as the column phase temperature.

[0032] In the above formula, Indicates row phase Relative to column phase The directional temperature difference is derived from the difference between the row phase connector temperature and the column phase connector temperature at the same acquisition time, and its dimension is temperature. Indicates row phase The corresponding joint temperature is derived from infrared thermometry results and is measured in units of temperature. Indicates column phase The corresponding joint temperature is derived from infrared thermometry results and is measured in units of temperature. express and All values ​​are taken from three phases, A, B, and C, constrained by a dimensionless phase set. This calculation does not introduce any new measurements; it is derived solely from the junction temperatures of phases A, B, and C.

[0033] When generating the three-phase temperature difference data generation module, the row and column labels of the three-phase temperature difference relationship matrix are fixed at A, B, and C, respectively, maintaining a fixed row and column order. Each position in the matrix corresponds to a specific temperature direction relative to the column phase. The three-phase temperature difference relationship matrix is ​​represented as follows: .

[0034] In the above formula, This represents a three-phase temperature difference relationship matrix, with values ​​derived from paired directed temperature differences between the A-phase junction temperature, the B-phase junction temperature, and the C-phase junction temperature. The dimension of each element in the matrix is ​​temperature. It represents the directed temperature difference obtained by subtracting the temperature of phase B from the temperature of phase A joint, and its dimension is temperature unit; It represents the directed temperature difference obtained by subtracting the temperature of phase C from the temperature of phase A joint, and its dimension is temperature unit; It represents the directed temperature difference obtained by subtracting the temperature of phase A from the temperature of phase B joint, and its dimension is temperature unit; It represents the directed temperature difference obtained by subtracting the temperature of phase C from the temperature of phase B joint, and its dimension is temperature unit; It represents the directed temperature difference obtained by subtracting the temperature of the A-phase joint from the temperature of the C-phase joint, with the dimension being temperature units; This represents the directed temperature difference obtained by subtracting the temperature of phase B from the temperature of phase C's junction, with the dimension in temperature units. The main diagonal of the three-phase temperature difference matrix represents the comparison between the same phase and itself. The position of the main diagonal... This is the zero value in temperature units.

[0035] Each position in the three-phase temperature difference matrix corresponds to a specific temperature orientation relative to the column phase. For example, the first row and second column correspond to the directed temperature difference between phase A and phase B, and the second row and first column correspond to the directed temperature difference between phase B and phase A. By fixing the row and column labels, the three-phase temperature difference matrix can directly express the directional relationship between any two phases without requiring the rule-based classification module to re-infer it from multiple independent temperature fields.

[0036] The three-phase temperature difference matrix generation unit 210 generates phase relationship fields simultaneously with the three-phase temperature difference matrix. The phase relationship fields include the three-phase average temperature, the relative heat of each phase, the total three-phase temperature difference, and the hottest phase number. Together with the three-phase temperature difference matrix, the phase relationship fields constitute three-phase temperature difference data that can be directly read by the rule grading module and the prompt template assembly module.

[0037] Three-phase average temperature is used to characterize the overall temperature level of the three-phase junctions. The three-phase average temperature is expressed as follows: It is obtained by the arithmetic mean of the A-phase joint temperature, the B-phase joint temperature, and the C-phase joint temperature.

[0038] In the above formula, It represents the three-phase average temperature, which is derived from the arithmetic mean of the A-phase joint temperature, B-phase joint temperature, and C-phase joint temperature, and is measured in units of temperature. This indicates the temperature of the A-phase junction, with the dimension being temperature units. This indicates the temperature of the B-phase connector, with the dimension being temperature units. This represents the temperature of the C-phase junction, with dimensions in temperature units; the denominator... This represents the number of phases involved in the calculation and is a dimensionless constant. The three-phase average temperature is used to determine whether the three-phase junction as a whole is at a high temperature level. However, the three-phase average temperature itself does not express the directional temperature difference relationship between any two phases. Therefore, it needs to be used in conjunction with the three-phase temperature difference relationship matrix and the total three-phase temperature difference.

[0039] The relative heat per phase is used to characterize the degree of temperature difference between each phase joint and the other phase joints. The relative heat per phase is expressed as follows: Where the relative heat component corresponding to each phase of A is expressed as The corresponding relative heat component for each phase of B is expressed as: The relative heat component corresponding to each phase of C is expressed as: .

[0040] In the above formula, This represents the relative heat component of each phase corresponding to A. Its value is derived from a linear combination of the joint temperatures of phase A, phase B, and phase C, and its dimension is temperature. This represents the relative heat component of each phase corresponding to B, with the dimension being temperature units; This represents the relative heat component of each phase corresponding to C, with the dimension being temperature units; This represents the relative heat of each phase, and the dimension of each component in the vector is in temperature units; Indicates phase The corresponding relative heat component for each phase is in units of temperature. Indicates phase The corresponding joint temperature is measured in units of temperature. This represents the three-phase average temperature, with dimensions in temperature units; coefficient. and All are dimensionless constants. A positive relative heat value for each phase indicates that the corresponding phase joint temperature is higher than the average temperature of the three phases; a negative relative heat value for each phase indicates that the corresponding phase joint temperature is lower than the average temperature of the three phases.

[0041] The total three-phase temperature difference is used to characterize the degree of temperature dispersion between the three phase junctions. The total three-phase temperature difference is expressed as follows: It is obtained by summing the squares of three paired temperature differences. The total three-phase temperature difference does not reflect the degree of overall temperature increase, but rather whether there are significant differences between the A-phase joint temperature, the B-phase joint temperature, and the C-phase joint temperature.

[0042] In the above formula, It represents the total temperature difference of the three phases, and its value is derived from the sum of the squares of three paired temperature differences. Its dimension is the square of the temperature unit. It represents the temperature difference between the A-phase connector temperature and the B-phase connector temperature, with the dimension being temperature units; It represents the temperature difference between the A-phase joint temperature and the C-phase joint temperature, with the dimension being temperature units; This represents the temperature difference between phase B and phase C joint temperatures, expressed in units of temperature. When the temperatures of phase A, phase B, and phase C joints are close, the total three-phase temperature difference is close to zero. When there is a significant temperature difference between one or two phase joints and the other phase joints, the total three-phase temperature difference increases.

[0043] The hottest phase number is used to characterize the phase position. The hottest phase number is represented as follows: It is determined by the location of the maximum value in the relative heat of each phase.

[0044] In the above formula, This indicates the hottest phase number, and its value is derived from the phase position where the maximum relative heat of each phase is located. The value is one of the phase numbers A, B, and C. The hottest phase number is a dimensionless phase number. This indicates the position where the maximum value is obtained, and is a dimensionless positional operation. Indicates phase The corresponding relative heat component for each phase is measured in units of temperature. The hottest phase number is used to indicate the position of the phase with the highest relative heat in the current three-phase temperature difference data in the rule classification module. If the relative heat of each phase for two phases meets the two-phase high criterion, the rule classification module records the two high phases, instead of relying solely on the hottest phase number to output a single phase.

[0045] After the three-phase temperature difference relationship matrix is ​​written, the three-phase temperature difference data status determination unit 220 performs zero diagonal verification and antisymmetry verification on the three-phase temperature difference relationship matrix. Zero diagonal verification checks whether the main diagonal of the three-phase temperature difference relationship matrix is ​​zero. Antisymmetry verification checks whether any two directed temperature differences in opposite directions are opposites.

[0046] In the above formula, This represents the transpose of the three-phase temperature difference relationship matrix. After transpose, the dimension of each element in the matrix is ​​still temperature. This indicates that each element in the three-phase temperature difference matrix is ​​the opposite of its value, and the dimension of each element remains temperature. If the three-phase temperature difference matrix is ​​correctly generated from the same set of A-phase junction temperatures, B-phase junction temperatures, and C-phase junction temperatures, then the three-phase temperature difference matrix should satisfy this antisymmetric relationship. If this relationship does not hold, it indicates that the three-phase temperature difference matrix may have issues such as misaligned fields, missing row labels, missing column labels, or corrupted data transmission.

[0047] When both the zero diagonal verification and the antisymmetry verification pass, the three-phase temperature difference data is marked as usable. If any verification fails, the three-phase temperature difference data generation module retains the temperatures of phase A, phase B, and phase C connectors, marks the three-phase temperature difference data as verification failed, and causes the rule grading module to revert to the original rule grading process. Retaining the original temperatures provides a basis for manual review and result retrospection, but the three-phase temperature difference relationship matrix that failed verification is not used for phase position determination.

[0048] When performing zero diagonal verification, the three-phase temperature difference data status determination unit 220 checks item by item whether the three main diagonal positions of A relative to A phase, B relative to B phase, and C relative to C phase in the three-phase temperature difference relationship matrix are zero values ​​in the temperature unit; when performing antisymmetric relationship verification, it checks item by item... , and Is it zero in the temperature unit? After the above verification is passed, the three-phase temperature difference data status determination unit 220 writes the three-phase temperature difference relationship matrix, phase relationship field and three-phase temperature difference data status into the current device record.

[0049] The three-phase temperature difference data generation module sends the three-phase temperature difference data to the rule classification module and the prompt template assembly module. The three-phase temperature difference data sent to the rule classification module is used to participate in the equipment defect level merging; the three-phase temperature difference data sent to the prompt template assembly module is used to write fixed fields of the cross-modal model and limit the phase position output of the cross-modal model.

[0050] Example 4: The rule-based classification module is used to incorporate three-phase temperature difference data as phase relationship evidence for the infrared channel into equipment defect level merging when the three-phase temperature difference data is available. When the three-phase temperature difference data is unavailable or verification fails, it reverts to the original rule-based classification process. The rule-based classification module receives audible small model results, ultrasonic small model results, ultraviolet small model results, and infrared over-temperature results from the multi-source data unified access module, and also receives three-phase temperature difference data from the three-phase temperature difference data generation module.

[0051] The rule grading module inputs the three-phase temperature difference data along with the results from the audible small model, ultrasonic small model, ultraviolet small model, and infrared overheating tests into a preset rule matrix to generate preset equipment defect levels. These preset defect levels can include normal, general, severe, and emergency defects. The results from each channel retain their respective sources in the preset rule matrix, with the three-phase temperature difference data serving as evidence of the phase relationship in the infrared channel to explain the temperature distribution between the three phase junctions.

[0052] The input fields of the preset rule matrix include the candidate levels for the audible channel corresponding to the audible small model result, the candidate levels for the ultrasonic channel corresponding to the ultrasonic small model result, the candidate levels for the ultraviolet channel corresponding to the ultraviolet small model result, the candidate levels for the infrared channel corresponding to the infrared overheating result, the three-phase temperature difference data status, the three-phase average temperature, the relative heat of each phase, the total three-phase temperature difference, and the hottest phase number. The values ​​for each channel candidate level are normal, general defect, severe defect, or emergency defect. The preset equipment defect level output by the preset rule matrix also takes the values ​​of normal, general defect, severe defect, or emergency defect. The preset equipment defect level is the equipment defect level generated by the rule grading module according to the preset rule matrix.

[0053] Before each channel's candidate level enters the preset rule matrix, the rule grading module processes the input according to a unified grading standard. When the audible small model result, ultrasonic small model result, ultraviolet small model result, or infrared overheating result has already output normal, general defect, severe defect, or emergency defect, the rule grading module directly generates the corresponding channel candidate level. When the audible small model result, ultrasonic small model result, ultraviolet small model result, or infrared overheating result outputs a confidence score, the rule grading module reads the channel level classification threshold corresponding to the channel type from the expert threshold library and converts the confidence score into the corresponding channel candidate level. When the audible small model result, ultrasonic small model result, ultraviolet small model result, or infrared overheating result cannot be identified or is missing, the rule grading module does not generate the corresponding channel candidate level and prevents that channel from participating in this preset rule matrix merging.

[0054] The preset rule matrix generates preset equipment defect levels according to safety priority rules. The safety priority rules are as follows: if any channel candidate level is an urgent defect, the preset equipment defect level is an urgent defect; if no urgent defect exists, and any channel candidate level is a severe defect, the preset equipment defect level is a severe defect; if no urgent or severe defect exists, and any channel candidate level is a general defect, the preset equipment defect level is a general defect; if all channel candidate levels are normal, the preset equipment defect level is normal. Three-phase temperature difference data serves as evidence of the phase relationship of the infrared channels in phase position recording and evidence correlation, but the total three-phase temperature difference is not used to replace the equipment safety threshold alone.

[0055] The trigger rule number consists of a rule version number, channel type, criterion type, and sequence number. The channel type can be audible, ultrasonic, ultraviolet, infrared overheating, or three-phase temperature difference data; the criterion type can be a conventional alarm threshold, a three-phase common heating criterion, a relative overheating criterion, a two-phase high temperature criterion, an unusable rollback, or a verification failure rollback. The rule classification module simultaneously writes the trigger rule number when generating the preset equipment defect level and phase position, enabling the result output module to trace back from the standardized defect detection result sequence to the rule source that triggered the result.

[0056] When the three-phase temperature difference data is available, the rule-based classification module reads the three-phase average temperature, the relative heat of each phase, the total three-phase temperature difference, and the hottest phase number. When the three-phase average temperature and the total three-phase temperature difference meet the three-phase common heating criterion in the expert threshold library, the rule-based classification module records the phase position as common to all three phases. At this time, the system considers the overall temperature level of the three phase junctions to be high, but the dispersion between the three phases is small, therefore the phase position is not recorded as a single phase.

[0057] The expert threshold library stores at least the common heating temperature threshold, temperature difference tolerance, relative heat threshold, two-phase overheating threshold, channel level classification threshold, threshold version number, applicable equipment type, applicable operating conditions, threshold effective time, and threshold activation status. The common heating temperature threshold is in temperature units and is used for comparison with the three-phase average temperature; the temperature difference tolerance is in the square of temperature units and is used for comparison with the total three-phase temperature difference; the relative heat threshold is in temperature units and is used to determine single-phase relative overheating; the two-phase overheating threshold is in temperature units and is used to determine that two phases are both overheated; the channel level classification threshold is used to convert small model results or infrared overheating results in confidence score form into corresponding channel candidate levels. All the above thresholds are configured by the expert threshold library according to the applicable equipment type, applicable operating conditions, and maintenance rules. The rule classification module only reads the currently valid values ​​from the expert threshold library and does not modify the thresholds during the analysis process.

[0058] When the rule classification module reads the expert threshold library, it uses the device type, operating condition, and threshold version number in the current device record as matching conditions to search for threshold records that match the applicable device type, applicable operating condition, and threshold version number. If multiple matching threshold records exist, the rule classification module reads the set of currently valid values ​​where the threshold is enabled and the threshold's effective time is the most recent. If no matching threshold record exists, or if the matching threshold record lacks a common heating temperature threshold, temperature difference tolerance value, relative heat threshold, two-phase high threshold, or channel level classification threshold, the rule classification module records the corresponding criterion as not triggered and saves the reason for the non-match.

[0059] The above formula represents the criterion for simultaneous heating of the three phases. It represents the average temperature of the three phases, and its dimension is temperature unit; This represents the common heating temperature threshold, with the dimension being temperature units, and the value is derived from an expert threshold library. It represents the total temperature difference between the three phases, with the dimension being the square of the temperature unit; This represents the allowable temperature difference value, with the dimension being the square of the temperature unit, and the value is sourced from an expert threshold library. When the average temperature of the three phases is greater than or equal to the common heating temperature threshold, and the total temperature difference of the three phases is less than or equal to the allowable temperature difference value, the rule classification module determines that the three-phase common heating criterion is established and records the phase position as three-phase common.

[0060] The above formula represents the relative overheating criterion. It represents the total temperature difference between the three phases, with the dimension being the square of the temperature unit; It represents the permissible temperature difference, with the dimension being the square of the temperature unit; Indicates phase The corresponding relative heat component for each phase is in units of temperature. This represents the relative heat threshold, measured in temperature units, and its value is derived from an expert threshold database. When the total temperature difference between the three phases exceeds the allowable temperature difference value, and the relative heat of each phase corresponding to a certain phase is greater than or equal to the relative heat threshold, the rule grading module determines that the relative overheating criterion is met and records the phase position as that phase.

[0061] The above formula represents the criterion for two-phase bias. It represents the total temperature difference between the three phases, with the dimension being the square of the temperature unit; It represents the permissible temperature difference, with the dimension being the square of the temperature unit; and These represent the relative thermal components of each phase corresponding to two different phases, with the dimension being temperature units; This represents the relative thermal component of each phase corresponding to the remaining phase, with the dimension being temperature units; This represents the two-phase over-temperature threshold, measured in temperature units, and its value is derived from an expert threshold database. When the total temperature difference between the three phases exceeds the allowable temperature difference value, the relative heat of each phase corresponding to two different phases is greater than or equal to the two-phase over-temperature threshold, and the relative heat of each phase corresponding to the remaining phases is less than zero, the rule grading module determines that the two-phase over-temperature criterion is valid and records the phase positions as the two over-temperature phases.

[0062] When the three-phase temperature difference data is available, the rule classification module processes the phase position according to a fixed priority: first, it determines whether the three-phase common heating criterion is valid; if the three-phase common heating criterion is not valid, it then determines whether the two-phase overheating criterion is valid; if the two-phase overheating criterion is not valid, it then determines whether the relative overheating criterion is valid; if none of the three criteria are valid, the three-phase temperature difference data is only saved as evidence of normal phase relationship, and no definite single-phase relative overheating position is output.

[0063] The rule-based classification module records the phase position as phase A when the total three-phase temperature difference and the relative heat of each phase corresponding to a given phase meet the relative overheating criteria in the expert threshold library. This process is used to identify single-phase relative overheating scenarios. For example, when the relative heat of each phase corresponding to phase A is significantly higher than that of phases B and C, and the total three-phase temperature difference meets the relative overheating criteria, the rule-based classification module records the phase position as phase A.

[0064] The rule-based grading module records the phase positions as two high-temperature phases when the relative heat of each phase corresponding to two phases meets the two-phase high-temperature criterion in the expert threshold library. This process is used to avoid missing the second high-temperature phase by outputting only a single phase based on the hottest phase number. If the relative heat of each phase corresponding to phases A and B are both positive and meet the two-phase high-temperature criterion, while the relative heat of each phase corresponding to phase C is negative, then the rule-based grading module records the phase positions as phases A and B.

[0065] When the three-phase temperature difference data status is unavailable, the rule-based classification module does not use the three-phase temperature difference relationship matrix, phase relationship field, and unavailable data in the three-phase temperature difference data status to determine the phase position. Instead, it reverts to the original rule-based classification process. The original rule-based classification process can still generate preset equipment defect levels based on the audible small model results, ultrasonic small model results, ultraviolet small model results, and infrared overheating results, but the phase position is not determined by the missing three-phase temperature difference data.

[0066] When the three-phase temperature difference data status is "verification failed," the rule classification module also reverts to the original rule classification process. Verification failure indicates that the three-phase temperature difference relationship matrix does not meet the zero diagonal verification or anti-symmetric relationship verification; the rule classification module must not record phase positions based on this three-phase temperature difference relationship matrix. This constraint is used to prevent the impact of field order misalignment, missing row labels, missing column labels, or transmission corruption on the merging of equipment defect levels.

[0067] Example 5: The prompt template assembly module is used to write three-phase temperature difference data into fixed fields of the cross-modal model and limit the phase position output of the cross-modal model. The prompt template assembly module receives three-phase temperature difference data from the three-phase temperature difference data generation module and receives equipment defect level and phase position from the rule grading module. When assembling the fixed fields, the prompt template assembly module does not rewrite the three-phase temperature difference data into free text guesses, but instead retains the A-phase joint temperature, B-phase joint temperature, C-phase joint temperature, three-phase temperature difference relationship matrix, relative heat of each phase, total three-phase temperature difference, hottest phase number, and three-phase temperature difference data status.

[0068] The template assembly module writes fixed fields into the cross-modal model, including A-phase junction temperature, B-phase junction temperature, C-phase junction temperature, three-phase temperature difference relationship matrix, relative heat of each phase, total three-phase temperature difference, hottest phase number, and three-phase temperature difference data status. These fixed fields are consistent with the three-phase temperature difference data source read by the rule grading module; therefore, the cross-modal model 500 can reference the same phase relationship evidence as the rule grading module when generating auxiliary analysis content.

[0069] The prompt template assembly module writes fixed fields into the cross-modal model in a fixed order. This fixed order is as follows: basic equipment information, A-phase connector temperature, B-phase connector temperature, C-phase connector temperature, three-phase temperature difference matrix, phase relationship field, three-phase temperature difference data status, equipment defect level, phase location, and evidence number. The three-phase temperature difference matrix field must include both row and column labels, with both labels fixed in the order of A, B, C. The phase relationship field must include the three-phase average temperature, relative heat of each phase, total three-phase temperature difference, and the hottest phase number.

[0070] When the three-phase temperature difference data status is unavailable, the template assembly module is prompted to still write the existing values ​​from the A-phase joint temperature, B-phase joint temperature, and C-phase joint temperature. Simultaneously, the field corresponding to the missing phase is marked as undetermined, and the fields for the three-phase temperature difference relationship matrix, relative heat of each phase, total three-phase temperature difference, and hottest phase number are also marked as undetermined. When the three-phase temperature difference data status is verification failed, the template assembly module is prompted to write the A-phase joint temperature, B-phase joint temperature, C-phase joint temperature, and the three-phase temperature difference data status. The three-phase temperature difference relationship matrix field is marked as verification failed, and cross-modal models are not allowed to reference this three-phase temperature difference relationship matrix to form a definite phase position. Cross-modal models must reference the three-phase temperature difference data status and evidence number when outputting phase positions.

[0071] When the phase position is recorded as common to all three phases, the template assembly module prompts that the cross-modal model must not output the phase position as a single phase solely based on the highest infrared temperature. This restriction prevents the cross-modal model from ignoring the fact that the total temperature difference between the three phases is relatively small after seeing the highest infrared temperature, thus mistakenly writing the overall temperature rise as a local defect in a single phase. The cross-modal model can indicate in the auxiliary analysis that the overall temperature level of the three phase joints is high, but the phase position should remain common to all three phases.

[0072] When the three-phase temperature difference data is unavailable, the template assembly module indicates that the cross-modal model will output the phase position as undetermined. This unavailable state is typically caused by the absence of temperature for any phase. In this case, the cross-modal model must not guess any phase A, B, or C based on audible small model results, ultrasonic small model results, ultraviolet small model results, or local visual features in infrared images. This processing ensures that the phase position output is based on directly acquired temperature data.

[0073] When the three-phase temperature difference data status is "verification failed," the template assembly module is prompted to retain the three-phase temperature difference data status in a fixed field and to restrict the cross-modal model from using the failed three-phase temperature difference relationship matrix to determine the phase position. The cross-modal model can output data quality prompts, indicating that the A-phase connector temperature, B-phase connector temperature, and C-phase connector temperature and their phase assignments need to be verified, but the failed three-phase temperature difference relationship matrix must not be used as the basis for phase conclusions.

[0074] The cross-modal model 500 receives fixed fields output by the prompt template assembly module 400, performs joint analysis on multimodal inspection data, three-phase temperature difference data and image evidence, and generates auxiliary analysis content.

[0075] Example 6: The results output module outputs a standardized defect detection result sequence. It receives the standardized defect detection result sequence from the rule classification module and auxiliary analysis content from the cross-modal model. The results output module also outputs natural language maintenance suggestions and associates and saves the standardized defect detection result sequence, natural language maintenance suggestions, three-phase temperature difference data, trigger rule number, and evidence number to support backtracking from phase position to three-phase temperature difference data and evidence number.

[0076] The standardized defect detection result sequence includes task number, site number, equipment number, equipment defect level, phase location, defect type, trigger rule number, and evidence number. Natural language maintenance suggestions are used to explain to maintenance personnel the equipment parts, phase locations, and review items that require attention. After the three-phase temperature difference data is associated and saved with the trigger rule number and evidence number, maintenance personnel can view the A-phase connector temperature, B-phase connector temperature, C-phase connector temperature, three-phase temperature difference relationship matrix, relative heat of each phase, total three-phase temperature difference, hottest phase number, and three-phase temperature difference data status from the phase location.

[0077] When the results output module performs associated saving, it uses the task number, device number, and acquisition time as the association keys for the same device's result record. Each standardized defect detection result sequence is associated with at least one trigger rule number and one evidence number; the evidence number points to the infrared thermometry result, three-phase temperature difference data, and corresponding image evidence; the trigger rule number points to the triggered rule version number, channel type, and criterion type in the preset rule matrix. If the phase position is determined by the three-phase common heating criterion, relative overheating criterion, or two-phase high temperature criterion, then the evidence number must be able to locate the A-phase joint temperature, B-phase joint temperature, C-phase joint temperature, three-phase temperature difference relationship matrix, relative heat of each phase, total three-phase temperature difference, hottest phase number, and three-phase temperature difference data status.

[0078] When tracing back from the phase position to the three-phase temperature difference data and evidence number, the result output module first locates the current device result record based on the task number, device number, and acquisition time in the standardized defect detection result sequence. Then, it reads the trigger rule number based on the phase position record, and subsequently reads the corresponding infrared thermometry result and three-phase temperature difference data based on the evidence number. If the three-phase temperature difference data status is unavailable or verification failed, the tracing result retains the reason for unavailability or verification failure and displays the trigger rule number from which the rule classification module reverted to the original rule classification process.

[0079] When the phase position output by the result output module is common to all three phases, the associated and saved three-phase temperature difference data should show that the average temperature of the three phases and the total temperature difference of the three phases meet the three-phase common heating criterion in the expert threshold library. When the phase position output by the result output module is a specific phase, the associated and saved three-phase temperature difference data should show that the total temperature difference of the three phases and the relative heat of each phase corresponding to that phase meet the relative superheat criterion in the expert threshold library. When the phase position output by the result output module is two slightly higher phases, the associated and saved three-phase temperature difference data should show that the relative heat of each phase corresponding to the two phases meets the two slightly higher phase criterion in the expert threshold library.

[0080] When the three-phase temperature difference data is unavailable or verification failed, the result output module can still output the standardized defect detection result sequence and natural language maintenance suggestions generated by the original rule-based grading process. However, the phase position should be consistent with the output defined by the prompt template assembly module. For the unavailable state, the phase position is output as undetermined; for the verification failure state, the result output module saves the data quality prompt corresponding to the verification failure.

[0081] As an optional implementation, the temperatures of phase A, phase B, and phase C connectors are all collected at the same time. The differences between these temperatures are small, and the infrared over-temperature result does not trigger a high-level alarm. The three-phase temperature difference relationship matrix generated by the three-phase temperature difference data generation module 200 satisfies the zero diagonal check and antisymmetry check, and the three-phase temperature difference data status is usable. After reading the three-phase average temperature, the relative heat of each phase, and the total three-phase temperature difference, the rule classification module 300 confirms that the total three-phase temperature difference is small, and combines the results of the audible small model, ultrasonic small model, ultraviolet small model, and infrared over-temperature results to generate a preset equipment defect level of normal or lower level in the preset rule matrix. The result output module 600 saves the three-phase temperature difference data and evidence number, so that the phase relationship under normal steady-state conditions can still be traced back.

[0082] As an optional implementation, during three-phase co-heating, the temperatures of phase A, phase B, and phase C joints are all higher than normal operating levels, but the differences between the three phase joints are small. The three-phase average temperature and the total three-phase temperature difference satisfy the three-phase co-heating criterion in the expert threshold library. The rule classification module 300 records the phase position as three-phase common, instead of recording the phase with the highest infrared temperature as a single phase. The prompt template assembly module 400 writes the three-phase temperature difference relationship matrix, the relative heat of each phase, the total three-phase temperature difference, the hottest phase number, and the three-phase temperature difference data status into fixed fields, and restricts the cross-modal model 500 from outputting the phase position as a single phase solely based on the highest infrared temperature. During three-phase co-heating, the rule classification module 300 compares the three-phase average temperature with the co-heating temperature threshold and the total three-phase temperature difference with the allowable temperature difference value. The three-phase co-heating criterion is established only when the three-phase average temperature is greater than or equal to the co-heating temperature threshold and the total three-phase temperature difference is less than or equal to the allowable temperature difference value. Once the three-phase common heating criterion is established, even if the hottest phase number can give one of phases A, B, and C, the rule classification module 300 will still record the phase position as three-phase common.

[0083] As an optional implementation, when single-phase relative overheating occurs, the temperatures of phase A, phase B, and phase C joints have all been correctly collected, and the three-phase temperature difference data is available. If the relative heat of each phase corresponding to phase A is significantly higher than that of each phase corresponding to phases B and C, and the total three-phase temperature difference and the relative heat of each phase corresponding to phase A meet the relative overheating criteria in the expert threshold library, then the rule grading module 300 records the phase position as phase A. If the corresponding phase is phase B or phase C, the rule grading module 300 records the phase position as phase B or phase C in the same way. This operating condition demonstrates the supporting role of the three-phase temperature difference relationship matrix and the relative heat of each phase for the single-phase relative overheating position. When single-phase relative overheating occurs, the rule grading module 300 compares the total three-phase temperature difference with the allowable temperature difference value and compares the relative heat of each phase with the relative heat threshold. Only when the total three-phase temperature difference is greater than the allowable temperature difference value, and the relative heat of each phase corresponding to a phase is greater than or equal to the relative heat threshold, is the relative overheating criterion established. Once the relative overheating criterion is met, the rule classification module 300 records the phase position as the phase that satisfies the relative overheating criterion and associates the trigger rule number with the phase position and saves it.

[0084] As an optional implementation, when both phases are simultaneously elevated, the relative heat of each phase corresponding to the two phases satisfies the two-phase elevation criterion in the expert threshold library, while the relative heat of each phase corresponding to the other phase is lower than the average temperature of the three phases. The rule classification module 300 records the phase positions as two elevated phases, instead of only using the phase corresponding to the hottest phase number as the unique phase position. The prompt template assembly module 400 writes the two elevated phases together with the three-phase temperature difference data into a fixed field, enabling the cross-modal model 500 to explain the fact that both phases are simultaneously elevated when generating auxiliary analysis content. When both phases are simultaneously elevated, the rule classification module 300 compares the total three-phase temperature difference with the allowable temperature difference value, and compares the relative heat of each phase corresponding to the two different phases with the two-phase elevation threshold respectively. The two-phase elevation criterion is established only when the total three-phase temperature difference is greater than the allowable temperature difference value, the relative heat of each phase corresponding to the two different phases is greater than or equal to the two-phase elevation threshold, and the relative heat of each phase corresponding to the remaining phases is less than zero. Once the two-phase-high criterion is met, the rule classification module 300 records the phase position as two high-phase phases.

[0085] The software in this implementation can be deployed on an edge computing terminal or a power grid intranet server. The edge computing terminal or power grid intranet server can run data access services, file storage services, relational database services, model inference services, rule engine services, prompt template services, and result feedback services.

[0086] The on-site inspection instrument is used to collect audible sound waveforms, ultrasonic waves, ultraviolet pulse data, infrared images, and infrared temperature measurement results. The phase assignment of the A-phase, B-phase, and C-phase connector temperatures in the infrared temperature measurement results is determined by the phase identifier and infrared temperature measurement point number in the equipment log. The equipment log can be maintained by the existing operation and maintenance platform; the system only reads existing phase identifiers and infrared temperature measurement point numbers.

[0087] The multi-source data unified access module, three-phase temperature difference data generation module, rule classification module, prompt template assembly module, and result output module can be implemented through software processes, service interfaces, function modules, or container services. Data transmission between modules can use database records, message queues, file references, or memory objects, but the transmitted objects still retain the aforementioned three-phase temperature difference data, preset equipment defect levels, phase positions, fixed fields, and standardized defect detection result sequences.

[0088] The expert threshold library can be deployed in a relational database, configuration file, or rule engine configuration item. It stores the aforementioned common heating temperature threshold, temperature difference tolerance, relative heat threshold, two-phase overheating threshold, channel level classification threshold, threshold version number, applicable equipment type, applicable operating conditions, threshold effective time, and threshold activation status. The rule classification module still reads rules and uses the expert threshold library according to the aforementioned currently valid values.

[0089] The cross-modal model 500 can be deployed as an intranet visual language model or a large language model service. Its input, output, and phase position constraint rules are executed according to the fixed field constraints of the prompt template assembly module.

[0090] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A multi-modal data defect and fault intelligent analysis system for power distribution network equipment, comprising a multi-source data unified access module, a three-phase temperature difference data generation module, a rule classification module, a prompt template assembly module, and a result output module, characterized in that: The multi-source data unified access module is used to receive multimodal inspection data and infrared temperature measurement results; The three-phase temperature difference data generation module is used to read the temperatures of phase A, phase B, and phase C at the same acquisition time when the object is a distribution transformer with three-phase terminals A, B, and C. It generates a three-phase temperature difference relationship matrix that expresses the directional temperature difference relationship between any two phases according to a fixed row and column order of A, B, and C. It generates a phase relationship field that includes the average temperature of the three phases, the relative heat of each phase, the total temperature difference of the three phases, and the number of the hottest phase. Based on the zero diagonal check and antisymmetry check of the three-phase temperature difference relationship matrix, it determines the status of the three-phase temperature difference data. The three-phase temperature difference data includes the three-phase temperature difference relationship matrix, the phase relationship field, and the status of the three-phase temperature difference data. The three-phase temperature difference data generation module sends the three-phase temperature difference data to the rule classification module and the prompt template assembly module; When the three-phase temperature difference data is available, the rule classification module uses the three-phase temperature difference data as evidence of the phase relationship of the infrared channel to participate in the merging of equipment defect levels. When the three-phase temperature difference data is unavailable or the verification fails, it reverts to the original rule classification process. The prompt template assembly module writes the three-phase temperature difference data into a fixed field of the cross-modal model and limits the phase position output of the cross-modal model; The results output module outputs a standardized sequence of defect detection results.

2. The intelligent analysis system for multi-modal data defects and faults of power distribution network equipment according to claim 1, characterized in that, Multimodal inspection data includes audible sound waveforms, ultrasonic waveforms, ultraviolet pulse data, and infrared images; the multi-source data unified access module establishes an inspection data index based on task number, site number, device number, acquisition time, and channel type.

3. The intelligent analysis system for multi-modal data defects and faults of power distribution network equipment according to claim 1, characterized in that, When the three-phase temperature difference data generation module reads the A-phase connector temperature, B-phase connector temperature, and C-phase connector temperature, all three temperatures are obtained from infrared temperature measurement results at the same acquisition time. The phase assignment is determined by the existing phase identifier and infrared temperature measurement point number in the equipment log. If any phase temperature is missing, the three-phase temperature difference data generation module records the three-phase temperature difference data status as unavailable, prohibits the estimation and completion of the missing phase temperature, and causes the prompt template assembly module to limit the cross-modal model to output the phase position as undetermined.

4. The intelligent analysis system for multi-modal data defects and faults of power distribution network equipment according to claim 1, characterized in that, When the three-phase temperature difference data generation module generates the three-phase temperature difference relationship matrix, it fixes the row and column labels of the three-phase temperature difference relationship matrix to A, B, and C, respectively, in a fixed row and column order, and makes each position in the three-phase temperature difference relationship matrix correspond to a specific row phase relative to the column phase in terms of temperature height.

5. The intelligent analysis system for multi-modal data defects and faults of power distribution network equipment according to claim 1, characterized in that, After writing the three-phase temperature difference relationship matrix, the three-phase temperature difference data generation module performs zero diagonal verification and antisymmetric relationship verification on the three-phase temperature difference relationship matrix. When any verification fails, the A-phase connector temperature, B-phase connector temperature and C-phase connector temperature are retained, the three-phase temperature difference data status is recorded as verification failure, and the rule classification module is rolled back to the original rule classification process.

6. The intelligent analysis system for multi-modal data defects and faults of power distribution network equipment according to claim 1, characterized in that, The phase relationship field includes the three-phase average temperature, which characterizes the overall temperature level of the three phase joints; the relative heat of each phase, which characterizes the degree of temperature difference between each phase joint and the other phase joints; the total three-phase temperature difference, which characterizes the degree of temperature dispersion among the three phase joints; and the hottest phase number, which characterizes the phase position.

7. The intelligent analysis system for multi-modal data defects and faults of power distribution network equipment according to claim 6, characterized in that, When the three-phase average temperature and the total three-phase temperature difference satisfy the three-phase common heating criterion in the expert threshold library, the rule classification module records the phase position as three-phase common; when the total three-phase temperature difference and the relative heat of each phase corresponding to a phase satisfy the relative overheating criterion in the expert threshold library, the phase position is recorded as that phase; when the relative heat of each phase corresponding to two phases satisfies the two-phase higher criterion in the expert threshold library, the phase position is recorded as two higher phases.

8. The intelligent analysis system for multi-modal data defects and faults of power distribution network equipment according to claim 1, characterized in that, The multi-source data unified access module is also used to receive the results of the audible small model, ultrasonic small model, ultraviolet small model and infrared over-temperature and send them to the rule classification module; the rule classification module sends the three-phase temperature difference data together with the results of the audible small model, ultrasonic small model, ultraviolet small model and infrared over-temperature into the preset rule matrix to generate the preset equipment defect level.

9. The intelligent analysis system for multi-modal data defects and faults of power distribution network equipment according to claim 1, characterized in that, The prompt template assembly module writes fixed fields into the cross-modal model, including A-phase junction temperature, B-phase junction temperature, C-phase junction temperature, three-phase temperature difference relationship matrix, relative heat of each phase, total three-phase temperature difference, hottest phase number, and three-phase temperature difference data status. When the phase position is recorded as common to all three phases, the prompt template assembly module restricts the cross-modal model from outputting the phase position as a single phase based solely on the highest infrared temperature. When the three-phase temperature difference data status is unavailable, the prompt template assembly module restricts the cross-modal model from outputting the phase position as undetermined.

10. The intelligent analysis system for multi-modal data defects and faults of power distribution network equipment according to claim 1, characterized in that, The results output module is also used to output natural language maintenance suggestions and associate and save the standardized defect detection result sequence, natural language maintenance suggestions, three-phase temperature difference data, trigger rule number and evidence number to support backtracking from the phase position to the three-phase temperature difference data and evidence number.

Images