A method for partial discharge detection and early warning of transformer-type cable supports

Abstract

This invention discloses a method for partial discharge detection and early warning of transformer-type cable supports, specifically relating to the field of power partial discharge monitoring technology. It addresses the problems of existing transformer-type cable support methods, such as difficulty in accurately locating individual faults in partial discharge, significant influence of structure and operating conditions on monitoring results, and the lack of a unified health assessment and early warning mechanism. By establishing a three-port equivalent model including conductor, insulation, transformer winding, and grounding parameters, and using unified encoding of structural fingerprint, coupling fingerprint, operational slice, and operating condition vector, a partial discharge health index under a benchmark operating condition is constructed by cascading calibration pulse, partial discharge signal filtering and deconvolution, and operating condition sensitivity calculation. This allows for traceable, quantifiable, and horizontally comparable assessment of the partial discharge intensity and repetition rate of a single support under different structural and load environments, solving the problem that traditional partial discharge detection methods struggle to objectively reflect the true insulation state of a single support and that results fluctuate with field conditions.

Description

Technical Field

[0001] This invention relates to the field of power partial discharge monitoring technology, specifically to a method for partial discharge detection and early warning of transformer-type cable supports. Background Technology

[0002] Currently, partial discharge monitoring of medium-voltage switchgear mainly employs offline withstand voltage plus partial discharge tests, online transient ground voltage detection, and ultra-high frequency sensor monitoring to assess the insulation status of the entire cabinet or cable terminals. These methods typically treat the primary conductor, supporting insulation components, and current transformers as a whole, with the measured signal usually being a high-frequency current or voltage at a specific location. Anomalies are then determined using simple amplitude thresholds or empirical rules. For transformer-type cable supports installed in outgoing circuits that combine conductor support and current transformer measurement functions, existing technologies generally rely solely on type tests or routine partial discharge inspections before commissioning. After commissioning, they often depend on periodic power outage tests, infrared thermography, and visual inspections, rarely establishing online partial discharge monitoring models corresponding to the structural characteristics of individual supports.

[0003] On the other hand, partial discharge behavior within switchgear is significantly affected by the primary wiring method, the geometric arrangement between conductors and the cabinet, as well as operating conditions such as load, temperature, and humidity within the electrical cabinet. Most existing online monitoring devices only collect partial discharge signals at a fixed location, lacking differentiated descriptions of different current transformer-type cable supports, making it difficult to establish a one-to-one correspondence between the collected high-frequency signals and specific supports. Furthermore, when assessing the severity of partial discharge, the amplitude and repetition rate of partial discharge under different operating conditions are often not normalized, resulting in poor comparability of monitoring results under different times and load conditions. In addition, existing early warning logic often uses a single threshold or a few statistical indicators to trigger alarms, lacking a health index that reflects the changing trend of support insulation status and its hierarchical management mechanism, resulting in insufficient connection between monitoring results and operation and maintenance decisions.

[0004] In summary, existing technologies for online monitoring of partial discharge on transformer-type cable supports suffer from several drawbacks. These include the inability to establish a stable correspondence between structure and signal for a single support, the difficulty in quantifying and classifying partial discharge levels while considering the impact of operating conditions, and the difficulty in establishing a closed loop between monitoring results and maintenance early warning processes. Therefore, it is necessary to provide a partial discharge detection and risk assessment solution for transformer-type cable supports to improve the relevance and practicality of insulation condition assessment. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for partial discharge detection and early warning of transformer-type cable supports, thereby solving the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for partial discharge detection and early warning of a current transformer type cable post, comprising: S1. Obtain the conductor, insulation, transformer winding and grounding parameters of the current transformer type cable support, establish a three-port equivalent model, generate a structural fingerprint, and bind it to the support identifier. S2. When the power is off, inject a partial discharge calibration pulse with a known charge amount, collect the response waveform in the partial discharge detection winding, and generate a coupled fingerprint by inverting the pulse response from the pulse source to the detection winding based on the structural fingerprint. S3. Configure the operation slice rules under a unified time clock, synchronously collect the partial discharge detection winding signals, current signals, temperature signals and humidity signals of each support, and archive them according to the operation slice. S4. Within each running slice, the signal is filtered and deconvolved using the pulse response corresponding to the coupled fingerprint, partial discharge events are extracted, and the equivalent charge and phase information are calculated to generate the partial discharge event record for that support. S5. Based on the load current, temperature and humidity of the operating slice, form the operating condition vector. According to the operating condition sensitivity model, convert the partial discharge equivalent charge and repetition rate to the reference operating condition, and calculate the partial discharge health index of a single transformer cable support. S6. Compare the partial discharge health index and the risk threshold classification table to determine the risk level. When the risk level reaches the alarm level, generate an early warning record containing the pillar identifier, partial discharge health index and risk level and send it to the operation and maintenance system.

[0007] Furthermore, S1 includes: When establishing a three-port equivalent model, the monitoring device assembles conductor parameters, insulation parameters, transformer winding parameters, and grounding parameters into a structural parameter vector in a preset order. The monitoring device converts each parameter item in the structural parameter vector into a fixed-length decimal string and concatenates them in a preset order to generate a structural fingerprint. The structural fingerprint, along with the support identifier consisting of the switch cabinet number, circuit number, and support column number, and the structural fingerprint version number, is written into a non-volatile storage medium.

[0008] Furthermore, S2 includes: When the switchgear is de-energized and grounded, the monitoring device injects a partial discharge calibration pulse of equivalent charge between the primary conductor and the grounding terminal of the current transformer cable support through a partial discharge calibration pulse generator and a current limiting element. Under the constraint of a unified timing clock, the partial discharge detection winding acquires the high-frequency response of the partial discharge calibration pulse within the high-frequency acquisition window through the high-frequency acquisition channel, generates a time-stamped original waveform record, and performs time alignment, baseline correction and bandpass filtering preprocessing on the original waveform record in sequence.

[0009] Furthermore, the monitoring device calls the structural fingerprint and three-port equivalent model corresponding to the support column identification, treats the equivalent power supply as a known input, convolves the pulse response to be identified with the waveform of the equivalent power supply to obtain the predicted response, and compares it point by point with the preprocessed original waveform record. The pulse response is iteratively adjusted under the constraint that the amplitude of the pulse response is zero at the time start and decays to below the amplitude threshold within a set time window to obtain the calibration pulse response from the equivalent discharge power supply to the partial discharge detection winding. The calibration pulse response is used as a coupling fingerprint and written into the read-only log area together with the support column identifier, structural fingerprint version number and calibration conditions.

[0010] Furthermore, S3 includes: The monitoring device adds time stamps to the partial discharge detection winding signal, current signal, temperature signal and humidity signal based on a unified time clock, and divides the time axis into continuous and non-overlapping operating slices according to the operating slice rules; Within each operating segment, the partial discharge detection winding signal, current signal, representative temperature value, and representative humidity value of each current transformer cable support are synchronously collected. The corresponding signal segments are associated with the operating segment identifier and the support identifier to generate an operating segment record and store it in the local storage medium.

[0011] Furthermore, S4 includes: For transformer-type cable supports, the calibration pulse response in the coupling fingerprint corresponding to the support identification is called to filter and deconvolve the partial discharge detection winding signal based on the calibration pulse response; Constraints are applied during the deconvolution process to ensure that the equivalent discharge current is non-zero only within a limited time sample sequence and zero outside of that time sample sequence. The timing of the partial discharge event is determined based on the amplitude threshold and the minimum time interval. The equivalent charge of the partial discharge event is calculated based on the reconstructed equivalent discharge current and the sampling time interval. The equivalent charge, current phase, running slice identifier, pillar identifier, and coupling fingerprint version number are used to generate a partial discharge event record and write it to the event log.

[0012] Furthermore, S5 includes: After obtaining the partial discharge event records within the operating slice, the monitoring device constructs an operating condition vector based on the representative values ​​of the load current, cabinet temperature, and cabinet relative humidity of the operating slice. The charge gain coefficient and repetition rate gain coefficient are obtained from the working condition vector in the working condition sensitivity model. The equivalent charge and repetition rate of partial discharge are converted to the reference operating condition using the charge gain coefficient and repetition rate gain coefficient.

[0013] Furthermore, the monitoring device performs statistical processing on the partial discharge equivalent charge sequence and partial discharge repetition rate sequence converted to the reference operating condition within the statistical period; This results in indicators such as charge quantity, repetition rate, and partial discharge phase distribution stability. The partial discharge health index is obtained by normalizing the charge quantity index, repetition rate index, and partial discharge phase distribution stability index according to preset weights and then summing them by weight.

[0014] Furthermore, S6 includes: Configure a risk threshold classification table with version number. After calculating the partial discharge health index, map the partial discharge health index to a risk level based on the risk threshold classification table. In the observation window, determine whether the alarm level has been reached based on the partial discharge health index sequence. When the alarm level is reached, an alarm record is generated, which includes device identifier, support identifier, partial discharge health index, risk level and alarm unique identifier. The record is sent to the operation and maintenance system through the field communication channel. The alarm unique identifier is used to control alarm deduplication and resending within the preset resending limit. The alarm generation and sending are also recorded in the operation log.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. By establishing a three-port equivalent model for transformer-type cable supports, including conductors, insulators, transformer windings, and grounding parameters, and using structural fingerprints, coupling fingerprints, operational slices, and operating condition vectors as unified codes, a partial discharge health index oriented towards benchmark operating conditions is constructed by connecting calibration pulse tests, partial discharge event filtering and deconvolution extraction, and operating condition sensitivity conversion in series. This achieves a traceable, quantifiable, and horizontally comparable online evaluation effect on the partial discharge intensity and repetition rate of a single support under different structural differences and load environments, solving the problems in traditional partial discharge detection that make it difficult to quantitatively reflect the true insulation state of a single support and that evaluation results fluctuate with field conditions.

[0016] 2. By combining the partial discharge health index with a risk threshold classification table with version management and a sliding observation window, and by embedding the structural fingerprint version, coupling fingerprint version, operating condition sensitivity model version, and unique warning mark in the warning record, the monitoring device and the operation and maintenance system are linked through a standard communication channel. This achieves the effect of automatically converting continuous partial discharge monitoring results into graded warnings and maintenance suggestions, reducing duplicate alarms and missed reports, and supporting post-event accountability, thereby improving the pertinence and timeliness of medium-voltage switchgear operation and maintenance decisions. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a partial discharge detection and early warning method for a current transformer type cable support according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example: Figure 1 A flowchart illustrating a partial discharge detection and early warning method for a current transformer-type cable post according to the present invention is provided. The method includes: S1. Obtain the conductor, insulation, transformer winding, and grounding parameters of the current transformer type cable support, establish a three-port equivalent model, generate a structural fingerprint, and bind it to the support identifier. The specific implementation is as follows: Before powering on a medium-voltage switchgear, to provide a stable and reliable assessment for each transformer-type cable post during subsequent partial discharge detection and early warning, it is necessary to first establish a structural description and equivalent model in the monitoring device that is consistent with the actual site conditions. In this scheme, a transformer-type cable post refers to a composite component that simultaneously undertakes the functions of primary conductor support and current inductance measurement. Internally, it contains an iron core and transformer windings; externally, it is an integrally cast or assembled insulating support component. Both ends are connected to the primary busbar or cable via conductor clamping structures, and an independent grounding terminal is provided at the bottom or side. Regarding this component, this method uniformly defines several types of structural information related to partial discharge behavior as conductor parameters, insulation parameters, transformer winding parameters, and grounding parameters.

[0020] The preferred conductor parameters include conductor diameter, minimum creepage distance between conductor center and cabinet metal shell, conductor fixed position on support and relative height. Their physical meaning describes the electric field distribution and the geometric boundaries of areas where discharge channels may occur. The preferred insulation parameters include support height, outer diameter, dimensions of different cross-sections, and relative permittivity of the manufacturing material. These parameters determine the insulation dielectric thickness, the degree of local electric field distortion, and the length of the surface discharge path. The preferred transformer winding parameters include primary conductor crossing method, core cross-sectional dimensions, number of turns in the secondary winding, axial position of the winding on the core, and secondary winding lead-out method. Their physical meaning affects the path and amplitude of partial discharge pulse energy coupled to the detection winding. The preferred grounding parameters include the specific installation position of the grounding terminal on the support, grounding wire connection path, distance between the grounding point and other grounding points in the cabinet, and wiring configuration, used to define the high-frequency return path.

[0021] In actual deployment, the monitoring device interfaces with the production line process files, primary system design drawings, and product nameplates. Preferably, the structural dimensions, tolerance ranges, and material information given in the design phase are first imported in batches. Then, after installation, on-site commissioning personnel conduct random checks and verifications of key dimensions. For example, calipers are used to measure the height of the exposed section of the support column, and special gauges are used to measure conductor gaps and creepage distances. The measured values ​​are preferably recorded in millimeters. Rated voltage information is recorded in kilovolts, rated current is recorded in amperes, and environmental factors such as temperature can be recorded in degrees Celsius in the operating condition module later.

[0022] For switchgear with a rated voltage of 10,000 kV, field experience and relevant standards typically recommend a minimum creepage distance of no less than a certain multiple. In this design, the minimum creepage distance of the transformer-type cable support is preferably designed to be around 800 mm, and the support height is designed to be around 400 mm, so as to ensure both insulation margin and suitability for the space inside the cabinet.

[0023] After acquiring or confirming the above parameters, the monitoring device establishes an equivalent model with three ports: the primary conductor end, the transformer winding end, and the grounding end. The three-port equivalent model preferably adopts a lumped parameter network form, and the dominant high-frequency coupling relationship between each port is represented by equivalent capacitance and equivalent impedance. This is used for subsequent calculation of the propagation characteristics of calibration pulses between each port. Preferably, the three-port equivalent model is used to describe the coupling characteristics in the range from tens of kilohertz to several megahertz, which covers the main spectral components of partial discharge monitored by the device.

[0024] To facilitate unified management and subsequent coding, the arrangement order of the structural parameter vectors is preferably fixed in advance as follows: conductor parameters, insulation parameters, transformer winding parameters, and grounding parameters are arranged in sequence. Within the same category, size parameters are recorded first, followed by material parameters and layout parameters. For example, in the conductor parameter section, the conductor diameter is recorded first, followed by the minimum creepage distance between the conductor center and the cabinet, and then the installation height of the conductor on the support column. In the insulation parameter section, the support column height is recorded first, followed by the support column outer diameter, and then the relative permittivity of the insulation material is recorded.

[0025] When generating structural parameter vectors, the monitoring device preferably uses on-site measured values ​​as the primary source. When the difference between the design drawing values ​​and the measured values ​​is within the allowable tolerance range, the measured values ​​are used. When the difference exceeds the allowable tolerance range, it is marked as a structural change, requiring a re-audit of the installation and the establishment of a new structural parameter vector. Allowable tolerances can be pre-set according to assembly specifications as absolute deviations not exceeding a certain number of millimeters or relative deviations not exceeding a certain percentage. Preferably, for medium-voltage switchgear transformer-type cable supports, the allowable tolerance for conductor-related dimensions can be set to no more than five millimeters or no more than five percent of the nominal size. Critical dimensions preferably include conductor diameter, minimum creepage distance, support height, support outer diameter, and installation height directly related to the partial discharge channel. Any change in any of these critical dimensions exceeding the allowable tolerance is considered a substantial change in structural parameters.

[0026] To ensure consistency of structural fingerprints across different devices, the monitoring device, after assembling the geometric and electrical parameters into a structural parameter vector, converts each item into a fixed-length decimal string according to a uniform number of decimal places, and then concatenates them sequentially in the above order to form a character sequence. This character sequence is defined as the structural fingerprint in this method. In scenarios where enhanced security is required, it is preferable to add device identifiers and verification fields before and after the character sequence. As long as the structural fingerprints obtained from the same structural parameter vector under the same encoding rules remain unique, it is considered an equivalent implementation.

[0027] The structural fingerprint is bound to the support column identifier. The support column identifier can be composed of the switch cabinet number, circuit number and support column serial number. When generating the structural fingerprint, the monitoring device also assigns a structural fingerprint version number. The structural parameter vector, structural fingerprint, support column identifier and version number are written together into the local non-volatile storage medium. The storage medium can be the flash memory inside the device or an external memory card. The writing process includes a timestamp and configuration source information to form a chain of evidence record.

[0028] Preferably, when any key dimension in the structural parameter vector changes beyond the aforementioned allowable tolerance range, or when the transformer winding lead-out method or grounding terminal position changes, the monitoring device generates a new structural fingerprint version number for the support column and marks the original version as deactivated, retaining it only during evidence chain log queries and no longer participating in online calculations. If subsequent equipment modifications or support column replacements result in changes to geometric parameters or connection methods, commissioning personnel will remeasure and input the relevant parameters according to the above rules, and complete the construction of a new structural parameter vector and generation of a structural fingerprint in the monitoring device. The old version of the structural fingerprint and the parameter vector at that time will still be stored in the log for retrospective analysis of the correspondence between historical calculation conclusions and the structural state at that time.

[0029] Through the above methods, under actual production line and on-site commissioning conditions, this approach solidifies the key structural features of the current transformer cable support into a set of parameters with a clear sequence, fixed coding rules, and traceability within the monitoring device. This provides a clear basis for subsequent calibration pulse tests, coupling transfer characteristic identification, and partial discharge event equivalent source reconstruction. It enables those skilled in the art, with knowledge of the switchgear voltage level, structural dimensional tolerances, and basic assembly methods, to complete structural modeling and fingerprint generation according to the steps described in this section, obtaining a three-port equivalent model and structural fingerprint consistent with the actual components on site. This leads to stable and consistent partial discharge detection and evaluation results in subsequent stages.

[0030] S2. During a power outage, inject a partial discharge calibration pulse with a known charge amount, collect the response waveform at the partial discharge detection winding, and generate a coupling fingerprint based on the pulse response from the pulse source to the detection winding, specifically as follows: After completing the structural fingerprint registration and arranging the power outage maintenance window, in order to ensure that the partial discharge behavior of each transformer-type cable post corresponds one-to-one with its own structural characteristics during subsequent online monitoring, this method performs a partial discharge calibration pulse test on each transformer-type cable post individually on-site. In this scheme, the partial discharge calibration pulse refers to a type of standard pulse with a pre-set equivalent charge, controlled waveform leading-edge width and duration, and a spectrum covering the main frequency band of partial discharge. Its function is to simulate a single partial discharge process occurring between the primary conductor and ground, so that the equivalent discharge source in the three-port equivalent model has a calculable and traceable input.

[0031] The equivalent charge refers to the total amount of charge injected between the primary conductor end and the ground end after integration over time. The unit is preferably picocoulomb. On-site, a combination of fixed values ​​can be set to cover the measurement range of the monitoring device. For example, a lower level can be selected to test the noise suppression capability, a middle level can be selected to calibrate the normal operating range, and a higher level can be selected to verify the upper limit of the linear range. The spectrum of the calibration pulse preferably covers the frequency band from several tens of kilohertz to several megahertz to match the working frequency band of the aforementioned three-port equivalent model.

[0032] In actual implementation, the switchgear is in a de-energized and reliably grounded state. According to the primary wiring diagram, the operator connects the partial discharge calibration pulse generator between the primary conductor end of the corresponding current transformer cable post and the nearest grounding point. The pulse generator is connected to the primary conductor through a current limiting element. The current limiting element can be a high-resistance element or a RC network, which is used to limit the instantaneous current peak value during injection and control the rising edge slope to avoid adverse effects on the equipment insulation. The grounding terminal is preferably directly connected to the main grounding busbar of the cabinet to ensure a clear and stable circuit path.

[0033] In this method, the partial discharge detection winding refers to a high-frequency detection winding arranged near the transformer core or closely coupled to its magnetic circuit. This detection winding is used to collect high-frequency response signals related to partial discharge during calibration and operation. Its two ends are connected to the high-frequency acquisition channel of the monitoring device through shielded cables. To ensure that the calibration results have sufficient statistical significance and suppress occasional interference, this method preferably injects multiple partial discharge calibration pulses into each support at a fixed rhythm. The rhythm can be set to several times per second, and the time interval between pulses is set to be several times greater than the circuit response decay time, so as to ensure that the electromagnetic state returns to rest before the next injection.

[0034] Before each calibration pulse injection, the monitoring device records the trigger time using a unified timing clock and opens a high-frequency acquisition window on the partial discharge detection winding channel. The high-frequency acquisition window covers a short pre-trigger time before the calibration pulse is triggered to a range of several microseconds after the trigger. The window length is preferably set to be sufficient to cover the time range of the main reflection and attenuation processes in the circuit. The sampling rate is preferably set to one hundred megaseconds per second or a value around that level, so as to obtain sufficient sampling points on the waveform leading edge in the tens of nanosecond range.

[0035] During the data acquisition process, the monitoring device establishes an original waveform record for each calibration pulse. The original waveform record includes at least the sampling point sequence, the unified timing time, the support marker, and the corresponding calibration charge level information. After a calibration batch is completed, the monitoring device performs unified preprocessing on this set of original waveforms: aligning each pulse on the time axis according to the timing time, so that the main peak of the waveform is located near the unified sample index; performing baseline correction based on the average value and fluctuation range of the pulse-free segment, adjusting the average value of the signal-free segment to near zero; and using bandpass filtering to weaken components significantly lower or higher than the partial discharge frequency band, so that the subsequent identification process mainly focuses on the response within the target frequency band.

[0036] In this method, the equivalent discharge source refers to the ideal current source or charge source located between the primary conductor end and the ground end in the three-port equivalent model. It is used to abstractly represent the current injection behavior of a single partial discharge. Its time variation can be regarded as a standard waveform determined by the set calibrated charge quantity, the nominal waveform of the pulse generator, and the loop impedance under calibration conditions.

[0037] After completing the preprocessing of the original waveform, the monitoring device calls the structural fingerprint and three-port equivalent model parameters corresponding to the support column identification, and regards the calibration test of the support column as a system response identification process between the known input and the measured output: the aforementioned equivalent discharge power supply is used as the input in the model, the assumed pulse response is convolved with the equivalent discharge power supply waveform to obtain the predicted response of the partial discharge detection winding end, and compared with the measured response given by the high-frequency acquisition channel point by point to calculate the difference between the two.

[0038] The monitoring device can initially set the impulse response sample sequence to a single-peak decay type or a small amount of oscillatory decay type. Then, under the premise of meeting the physical constraints, the amplitude and shape of the sample are gradually adjusted. The physical constraints preferably include the impulse response starting from zero near the beginning of the time axis, monotonically or oscillatingly decaying to near zero before the end of the predetermined time window, and there is no long period of continuous growth.

[0039] The above prediction and comparison can be repeated several times within the identification period. When the difference between the predicted response and the measured response is within a preset allowable range, the pulse response sample sequence at this time is identified as the calibration pulse response from the equivalent discharge power source to the partial discharge detection winding. The allowable range of difference is preferably measured as a certain percentage of the measured response peak value, for example, it can be set to not exceed 10% of the measured peak value.

[0040] To improve the stability of the calibration results, this method preferably performs statistical averaging and deviation removal on the pulse responses obtained from multiple injections. For example, the median or weighted average of multiple test results is calculated for each sampling point, and individual results that significantly deviate from the median are regarded as interference points and removed. The degree of deviation can be set to deviate from the median by a certain number of times, such as no more than three times the noise fluctuation. The noise fluctuation can be obtained statistically from the amplitude distribution of the waveform in the pulseless section.

[0041] After the above processing, a representative impulse response curve is obtained and stored in this method as a discrete time series of a certain length. This time series, together with the calibrated charge level, the identification time period length, the preprocessing filter bandwidth, and the number of experiments involved in the statistics, constitutes a set of impulse response parameters. This scheme defines this set of parameters as a coupled fingerprint.

[0042] The coupling fingerprint reflects the energy coupling characteristics and waveform characteristics between the equivalent discharge source and the partial discharge detection winding under the current structural fingerprint conditions of the current transformer cable post. It serves as the basis for subsequently calculating the equivalent discharge charge and equivalent waveform based on the partial discharge detection winding signal under online operating conditions. After completing the coupling fingerprint identification of a single post, the monitoring device assigns a coupling fingerprint version number to this set of parameters. The coupling fingerprint, coupling fingerprint version number, corresponding post identifier, structural fingerprint version number, calibration date, calibration charge level, and the number of tests included in the statistics are all written to the read-only log area. The log area is preferably a non-volatile storage space that only allows append-only writing, used to form the initial record of the evidence chain.

[0043] For scenarios where multiple supports of the same model have already undergone coupling fingerprint calibration during the factory type testing phase, the consistency of key sampling points can be verified by comparing the structural fingerprint and selecting a small number of partial discharge calibration pulse tests during on-site device commissioning to confirm whether the factory-calibrated coupling fingerprint should be used. For scenarios where the structural fingerprint version changes due to replacing supports or adjusting wiring on-site, the partial discharge calibration pulse test and pulse response identification process should be re-executed according to the above steps to generate a new coupling fingerprint version number.

[0044] Preferably, in a switchgear with a rated voltage of 10,000 kV, the equivalent charge of the partial discharge calibration pulse can be set to 100 picocoulombs, the sampling rate of the high-frequency acquisition channel can be set to 100 megaseconds per second, the identification time period can be set to the order of several microseconds, the allowable difference between the predicted response and the measured response can be set to no more than 10% of the measured peak value, and results that deviate from the median value by more than three times the noise fluctuation in multiple tests can be judged as interference and eliminated. Those skilled in the art can obtain a coupling fingerprint that matches the current transformer cable support structure and wiring method by performing field wiring, calibration pulse injection, high-frequency waveform acquisition, pulse response identification and coupling fingerprint registration according to the above steps, thus providing a stable and consistent physical basis for the reconstruction of partial discharge events during subsequent operation.

[0045] S3. Configure the operation slice rules under a unified time synchronization clock, synchronously collect the partial discharge detection winding signals, current signals, temperature signals, and humidity signals of each support, and archive them according to the operation slice. The specific implementation is as follows: After the monitoring device completes the establishment and operation of structural fingerprints and coupling fingerprints, to ensure the consistency of time between the partial discharge signals, current conditions, and environmental information of each current transformer cable support, this method uses a unified timing clock to align the time of all relevant channels throughout the entire cabinet. In this scheme, the unified timing clock refers to a unified time reference provided by an internal or external timing source of the monitoring device. Preferably, it can be combined with an internal high-precision crystal oscillator working in conjunction with an external timing signal. The time resolution can be set to the microsecond level, and the current time is periodically broadcast via a fieldbus or independent timing line, ensuring that each partial discharge acquisition unit and current measurement unit maintains the same time scale in hardware. Each high-frequency sampling point and each power frequency current sampling point has a time stamp based on the unified timing clock, allowing measurement results from different locations and branches to be arranged along the same time axis.

[0046] Based on unified time synchronization, this method configures operational slicing rules in the monitoring device, dividing the time axis into a series of continuous and non-overlapping operational slices. In this scheme, an operational slice refers to a fixed time window consisting of several power frequency cycles. The window length is preferably set to an integer number of power frequency cycles to ensure complete coverage of several power frequency voltage and current changes within each window. For example, in a power grid with a rated frequency of fifty cycles per second, one power frequency cycle is twenty milliseconds, and the operational slice length can be set to twenty power frequency cycles, corresponding to a time length of approximately 0.4 seconds. When configuring the operational slicing rules, the monitoring device aligns the start and end times of the slices with the unified time synchronization clock, ensuring that each slice starts from a fixed power frequency phase position, facilitating subsequent cross-slice comparisons of partial discharge phase distribution. Once the device enters operational mode, within each operational slice, the monitoring device synchronously acquires the partial discharge detection winding signals of each current transformer cable support within the cabinet, the current signals of the corresponding circuits, and the ambient temperature and humidity signals within the cabinet.

[0047] In this method, the partial discharge detection winding signal refers to the high-frequency voltage or current signal transmitted from the partial discharge detection winding of each transformer-type cable post. These signals are continuously acquired through a high-frequency acquisition channel at a fixed sampling rate, and each sampling point has a unified timing time. The current signal is provided by the power frequency current transformer of each circuit, which is continuously acquired at a lower sampling rate and also has a unified timing time.

[0048] In order to obtain the current phase in subsequent steps, this method identifies the zero point or peak position of the power frequency waveform in the current signal of each circuit according to a unified timing time, converts the time length within one power frequency cycle into a standard angle, and calculates the power frequency phase angle corresponding to any moment within the operating slice. The power frequency phase corresponding to the moment when the partial discharge event occurs is regarded as the current phase of that event.

[0049] The temperature and humidity signals inside the cabinet are provided by temperature and humidity sensors installed in appropriate locations inside the switch cabinet. The sensor output signals are sampled at fixed time intervals through the acquisition channel in the monitoring device. The temperature unit is preferably degrees Celsius, and the humidity unit is preferably relative humidity percentage.

[0050] To control the amount of data and highlight the characteristics of the operating conditions, this method selects representative values ​​for the temperature and humidity sequences within each operating slice. The representative value can preferably be the average value, median value, or weighted representative value after excluding obvious outliers within the slice. The determination of outliers can refer to the aforementioned noise fluctuation range or set a threshold based on the normal rate of change of the sensor. The representative value is then associated with the operating slice identifier and written into the operating condition record.

[0051] In this scheme, the operating condition record refers to a record item that summarizes the environmental and load conditions of a certain current transformer cable support within a certain time window, with the operating slice as the granularity. It includes at least the representative values ​​of the cabinet temperature and humidity in the operating slice, as well as the representative value of the load current of the corresponding circuit in the slice. The representative value of the load current can preferably be the effective value or average value of the current in the slice, calculated after simple filtering out obvious abnormal samples.

[0052] At the end of each operating slice, the monitoring device archives the time segments of the partial discharge detection winding signals, the time segments of the current signals, and the operating condition records of each support within the slice according to the operating slice identifier and the support identifier, forming an operating slice record with a time tag and a support tag. The operating slice record can be stored in local storage medium in chronological order, and the unified timing clock configuration and operating slice rule version are indicated at the top of the record.

[0053] Since all records rely on a unified time clock for time stamping and slicing, this method can use the running slice and unified time as clues to match partial discharge events, power frequency current phase, and temperature and humidity conditions during subsequent partial discharge event identification and operating condition vector construction. This ensures that the amplitude and phase information of partial discharge events, as well as the corresponding load and environmental conditions, match in time.

[0054] Preferably, in a switchgear with a rated voltage of 10,000 kilovolts and a rated frequency of 50 times per second, the time resolution of the unified timing clock can be set to one microsecond, the length of the operating slice can be set to twenty power frequency cycles, corresponding to approximately 0.4 seconds, the current signal sampling rate can be set to one thousand times per second, and the temperature and humidity sampling period can be set to the order of several seconds. Those skilled in the art can configure the monitoring device for timing, slice rules, and acquisition rhythm according to the above rules, so as to form a time-aligned, orderly slice and clearly representative operating condition value operating slice record in actual operation, providing a public, sufficient and reproducible time basis for subsequent partial discharge event identification, operating condition vector construction and health index calculation.

[0055] S4. Within each running slice, the signal is filtered and deconvolved using the pulse response corresponding to the coupled fingerprint to extract partial discharge events and calculate the equivalent charge and phase information, generating a partial discharge event record for that support pillar. Specifically, the implementation is as follows: After the operational slices are archived, in order to extract discrete events corresponding to the equivalent discharge behavior from the continuous partial discharge detection winding signal within each operational slice, this method calls the coupling fingerprint corresponding to the identifier of each current transformer cable support. The calibration pulse response stored in the coupling fingerprint is used as the feature kernel to filter and deconvolve the partial discharge detection winding signal of that support within that operational slice. In this scheme, the pulse response in the coupling fingerprint refers to the discrete response sequence at the partial discharge detection winding end when the equivalent discharge power source generates a unit pulse under calibration conditions. This sequence has been fixed through the aforementioned calibration steps and multiple experimental statistics, reflecting the energy coupling path and waveform morphology from the equivalent discharge power source to the partial discharge detection winding under the current structural fingerprint conditions. The partial discharge detection winding signal within the operational slice in this scheme refers to the voltage or current waveform continuously acquired at a high-frequency sampling rhythm within the time range of that operational slice, with a unified time stamp.

[0056] In the filtering process, the monitoring device preferably uses the pulse response in the coupled fingerprint as a reference, slides the pulse response and the segment of the partial discharge detection winding signal in the time domain to perform similarity calculation, or constructs a bandpass weight that matches the spectrum of the pulse response in the frequency domain, amplifies the frequency components that match the pulse response shape, and suppresses broadband noise and switching operation interference that are significantly different from the pulse response, thereby obtaining a filtered and enhanced partial discharge candidate waveform.

[0057] Based on this, this method uses the impulse response in the coupled fingerprint as a feature of the equivalent system to perform deconvolution operation on the filtered partial discharge detection winding signal. In this scheme, deconvolution means treating the partial discharge detection winding signal as the result of linear superposition of an unknown equivalent discharge current sequence through a known impulse response. By gradually adjusting the sample values ​​of the equivalent discharge current sequence, the predicted waveform obtained by superimposing the current sequence and the impulse response is made to approximate the filtered measured waveform within the allowable error range.

[0058] When the monitoring device performs deconvolution operation, it preferably selects a certain length of time period around the candidate partial discharge peak as the deconvolution interval on the time axis. Within this interval, it is initially assumed that the equivalent discharge current is a series of non-negative pulse samples concentrated in a short time. Combining the three-port equivalent model and the pulse response stored in the coupling fingerprint, the predicted waveform of the partial discharge detection winding end is obtained by superposition calculation in each round of adjustment. The predicted waveform is compared with the filtered waveform point by point. When the local error exceeds the preset allowable range, the amplitude and distribution of the equivalent discharge current sample at the corresponding time point are adjusted so that the predicted waveform gradually approaches the measured waveform.

[0059] To conform to the physical meaning, the deconvolution operation preferably applies several constraints, such as the equivalent discharge current being mainly concentrated in one or a few adjacent sample points in time, with no long-term continuous current segments in the same direction, and the equivalent discharge current time series obtained by deconvolution decaying to near zero after a certain period of time; the monitoring device can correct or reject deconvolution results that do not meet these constraints during multiple rounds of adjustment.

[0060] Through the above deconvolution process, a set of equivalent discharge current sample sequences is obtained in each candidate time interval. This scheme takes the region of concentrated amplitude in the sequence as the time position of a single partial discharge event, and multiplies the sum of the values ​​of each sample in the sequence by the sampling time interval as the equivalent charge of the partial discharge event, which corresponds to the picocoulomb unit used in the calibration stage in terms of dimensions.

[0061] To distinguish adjacent partial discharge events from continuous waveforms within a running slice, this method employs two constraints during deconvolution: an amplitude threshold and a minimum time interval. The amplitude threshold is preferably determined based on the noise level of the non-partial discharge section within the running slice and can be set as a certain multiple of the peak-to-peak value of the noise waveform. When the peak value of the equivalent discharge current obtained after deconvolution is lower than this threshold, the candidate event is considered a noise fluctuation and not registered. The minimum time interval constraint is preferably determined based on the equipment response time and the expected partial discharge repetition rate. For example, it is stipulated that the time interval between two effective partial discharge events must not be less than a certain number of microseconds or a certain number of sampling points. When multiple local peaks appear within this interval, similar peaks in the equivalent current sequence can be merged into one event to avoid splitting the same physical discharge process into multiple events.

[0062] For several time positions within the running slice that satisfy the amplitude threshold and the minimum time interval constraint between them, this method treats them as the time of independent partial discharge events, and uses a unified timing clock and the aforementioned power frequency phase calculation rules to determine the phase value of each partial discharge event in the corresponding power frequency cycle, thereby obtaining the power frequency phase information of the event.

[0063] After completing the calculation of equivalent charge and power frequency phase, the monitoring device combines the equivalent charge, power frequency phase, unified timing of occurrence, and support identifier, operating slice identifier, and coupling fingerprint version number of each partial discharge event to form a partial discharge event record for that support within the operating slice. The partial discharge event record may also preferably include the amplitude threshold value, minimum time interval setting, and deconvolution error assessment results used in the deconvolution process, so as to be traced in subsequent health index calculation and maintenance analysis.

[0064] All partial discharge event records are written to the event log in the order of the running slice identifier and the pillar identifier. The event log can be stored in the non-volatile storage medium of the monitoring device and includes version information of the current deconvolution rule and threshold parameters.

[0065] For a current transformer cable support with insulation defects, under the above filtering and deconvolution steps, preferably, the equivalent charge value obtained by deconvolution can be observed to be significantly higher than that of other supports in the same cabinet within multiple consecutive operating slices. For example, it can reach the order of hundreds of picocoulombs under the reference operating conditions, and the number of partial discharge events registered within the same time length is significantly more than that of other supports. This difference will be reflected in the event log as an increase in the number of event records corresponding to this support and an overall rightward shift in the distribution of equivalent charge, providing a direct basis for subsequent statistical calculation of the partial discharge health index and determination of risk level based on multiple operating slices.

[0066] Preferably, in a switchgear with a rated voltage of 10,000 kV, the allowable range of deconvolution error can be set to a certain percentage of the peak value of the filtered waveform, for example, no more than 10%. The amplitude threshold can be set to three times the peak-to-peak value of the noise within the operating slice. The minimum time interval between partial discharge events can be set to a certain number of microseconds, corresponding to a certain number of high-frequency sampling points. Based on the above rules and the aforementioned coupling fingerprint calibration results, those skilled in the art can realize robust identification of partial discharge events and repeatable calculation of equivalent charge on actual operating data, providing a fully disclosed and reproducible event extraction basis for the entire set of partial discharge detection and early warning methods.

[0067] S5. Based on the load current, temperature, and humidity of the operating segment, a working condition vector is formed. Using the working condition sensitivity model, the equivalent charge and repetition rate of partial discharge are converted to the baseline working condition. The partial discharge health index of a single transformer-type cable support is then calculated. The specific implementation is as follows: After obtaining partial discharge event records for several operating segments, in order to compare the partial discharge behavior under different loads and environments, this method constructs a condition vector for each operating segment based on the representative values ​​of load current, temperature, and humidity of the corresponding transformer-type cable support. In this scheme, the condition vector refers to an ordered parameter set composed of a finite number of quantities that can characterize the electrical load and environmental state. It includes at least the representative values ​​of load current, cabinet temperature, and cabinet relative humidity within the operating segment. The representative value of load current can be selected according to a unified rule based on the average, effective, or maximum value of the power frequency current waveform within the operating segment, and is expressed in amperes. The representative value of temperature can be determined based on the average or median value of the temperature measurement sequence within the operating segment after removing obvious anomalies, and is expressed in degrees Celsius. The representative value of humidity can be determined based on the average or median value of the cabinet relative humidity measurement sequence within the operating segment after removing anomalies, and is expressed as a percentage.

[0068] When constructing the operating condition vector, the monitoring device caches the vector along with the operating slice identifier and the support identifier for later use in conversion and statistics. In this scheme, the operating condition sensitivity model refers to a set of rules or lookup tables formulated before the device is put into field use, based on partial discharge event records, load conditions, and environmental condition data accumulated during the type testing and trial operation phases of the same type of current transformer cable support. This model has an independent version number and is used to describe the overall trend of the equivalent charge of partial discharge and the partial discharge repetition rate as a function of load current, temperature, and humidity.

[0069] In this scheme, the partial discharge repetition rate refers to the ratio of the number of partial discharge events registered within a certain operating slice length to that time length, reflecting the frequency of partial discharge occurrence under that operating condition. Preferably, the operating condition sensitivity model can be implemented using a partitioned lookup table method. The load current, the temperature inside the electrical cabinet, and the relative humidity are divided into several ranges, and these three are combined to form a three-dimensional operating condition grid cell. Within each grid cell, the average gain coefficient of the equivalent charge of partial discharge relative to the reference operating condition and the average gain coefficient of the partial discharge repetition rate relative to the reference operating condition are obtained based on type tests and historical records. These two coefficients, along with the grid boundaries, are stored in the operating condition sensitivity model. During online operation, when the operating condition vector of a certain operating slice falls into one or more adjacent grid cells, the monitoring device can directly look up the gain coefficient in the corresponding grid or interpolate the gain coefficients of adjacent grids to obtain the operating condition coefficients used to calculate the equivalent charge of partial discharge and the repetition rate under that operating condition. The operating condition sensitivity model preferably locks in a stable "relationship between operating condition and partial discharge intensity" through the above-mentioned three-dimensional mesh division and statistical coefficients, and retains the old version and its applicable scope when the model version is updated, so as to verify the difference in conversion results under different versions in the future.

[0070] This method, when the device is running online, calculates the equivalent charge of each partial discharge event and the partial discharge repetition rate within the current operating slice as equivalent values ​​under a preset reference operating condition based on the operating condition vector and operating condition sensitivity model of the current operating slice. In this scheme, the reference operating condition refers to a set of standard loads and environmental conditions used for comparison. Preferably, it can be set as a certain percentage of the rated current, a certain degree Celsius of ambient temperature, and a certain percentage of relative humidity; for example, it can be set as 50% of the rated current, 25 degrees Celsius of ambient temperature, and 50% of relative humidity.

[0071] The underlying logic of the conversion process is as follows: First, using the charge gain coefficient of the corresponding grid cell in the operating condition sensitivity model, the partial discharge equivalent charge observed under a certain operating condition vector is proportionally amplified or reduced to correspond to the equivalent charge under the reference operating condition. Then, using the repetition rate gain coefficient of the same grid cell, a similar conversion is performed on the partial discharge repetition rate of the operating slice, thereby obtaining a directly comparable equivalent charge sequence and repetition rate sequence under the reference operating condition. Through this conversion, the differences in load current and environmental conditions among different operating slices can be eliminated, enabling horizontal comparison of partial discharge intensity from different time periods and different operating states on a unified benchmark.

[0072] Preferably, within a statistical period, the monitoring device comprehensively evaluates the equivalent charge sequence and repetition rate sequence of all operating segments of the single current transformer cable support after conversion to the baseline operating condition. On the one hand, it can calculate the average value, percentile value, and proportion of events above a certain threshold of the converted equivalent charge, reflecting the overall level and degree of abnormality of the partial discharge amplitude. On the other hand, it can calculate the average value, peak value, and proportion of time spent in the high repetition rate range after conversion, reflecting the frequency of partial discharge occurrence. Simultaneously, it can also statistically analyze the concentration of the partial discharge phase distribution based on the aforementioned unwinding power frequency phase information, for example, by assessing whether the proportion of events in a certain phase interval is significantly high, thereby reflecting the stability of the partial discharge phase distribution. In this scheme, the statistical period is preferably defined as a time period containing several consecutive operating segments, for example, it can be set to contain ten operating segments, corresponding to a time length of several seconds to tens of seconds, and is fixed in the device configuration with a version number.

[0073] In this scheme, the partial discharge health index refers to a dimensionless value obtained by normalizing and weighting the aforementioned multidimensional statistics according to pre-set weights. It is used to summarize the overall partial discharge risk level of a single current transformer cable support under baseline operating conditions using a single indicator. The weights can be determined before the device leaves the factory based on test experience and operation and maintenance requirements. For example, the influence weight of the converted equivalent charge can be greater than that of the repetition rate, and the influence weight of the repetition rate can be greater than that of the phase distribution stability. Preferably, the weight of the equivalent charge quantum index can be set to 0.5, the weight of the repetition rate sub-index can be set to 0.3, and the weight of the partial discharge phase distribution stability sub-index can be set to 0.2. This set of weights is fixed in the operating condition sensitivity model and health index calculation rules in the form of version numbers.

[0074] When the monitoring device calculates the partial discharge health index, it can first map the converted equivalent charge, converted repetition rate, and partial discharge phase distribution stability to sub-indicators between zero and several according to their respective characteristic intervals. For example, the converted equivalent charge can be mapped to near zero in the low value interval and to near several in the high value interval, and the repetition rate can be mapped to near zero in the low frequency region and to near several in the high frequency region. Then, the sub-indicators are added together according to a predetermined weight to obtain an overall index value.

[0075] To facilitate understanding and use by maintenance personnel, this method preferably divides the partial discharge health index into several levels according to the numerical range. For example, zero to one can be classified as normal level, one to two as attention level, two to three as warning level, and three and above as severe level. These range divisions and their corresponding meanings are fixed along with the version number of the operating condition sensitivity model.

[0076] When the partial discharge health index of a single current transformer cable support, after being converted to the baseline operating condition within a statistical period, remains at the warning or severe level for an extended period, or shows a continuous upward trend over multiple consecutive statistical periods, the monitoring device records the partial discharge health index value, corresponding level, and trend of change of the support in the health assessment log, and can trigger subsequent warning steps.

[0077] Preferably, in a switchgear with a rated voltage of 10,000 kilovolts and a rated current of several amperes, the reference operating conditions can be set to 50% of the rated current, an ambient temperature of 25 degrees Celsius, and a relative humidity of 50%. The partial discharge health index is divided into levels of 0 to 1, 1 to 2, 2 to 3, and above 3. For a support column with obvious insulation defects, the partial discharge health index may reach about 3.5 after conversion under the above reference operating conditions and be classified as severe.

[0078] Those skilled in the art can implement the above-mentioned working condition vector construction method, the gridded formulation method of the working condition sensitivity model, the conversion rules, and the health index calculation process to reproduce the partial discharge health assessment process of this method under actual production line operating conditions, and obtain a quantitative and clear partial discharge health index that can be used for operation and maintenance decisions.

[0079] S6. Compare the partial discharge health index with the risk threshold classification table to determine the risk level. When the risk level reaches the alarm level, generate an early warning record containing the support identifier, partial discharge health index, and risk level, and send it to the operation and maintenance system. The specific implementation is as follows: After the partial discharge health index is calculated, in order to reliably transform the quantitative assessment results into actionable operation and maintenance signals, this method configures a risk threshold classification table with a version number in the monitoring device, and uses this classification table as a benchmark to classify and track the trend of the partial discharge health index of each transformer-type cable support.

[0080] In this scheme, the risk threshold classification table refers to the risk level boundaries and their meanings defined in advance according to the range of health index values. It includes at least four ranges: normal, attention, warning, and severe. For example, zero to one can be classified as normal, one to two as attention, two to three as warning, and three and above as severe. The classification table is managed with an independent version number. When the classification boundaries or level meanings are adjusted, a new version is generated and updated in the device.

[0081] After each health index calculation, the monitoring device first maps the numerical value to the corresponding risk level based on the currently effective risk threshold classification table and caches it locally. Then, it introduces an observation window on the time axis to perform smoothing analysis on the changes in the health index. In this scheme, the observation window refers to a sliding time period covering several consecutive health index sampling points. The health index sampling points can correspond to a statistical period consisting of several operating slices. The length of the observation window can preferably be set to several consecutive statistical periods or the set of operating slices it covers, for example, it can be set to a sliding window containing ten adjacent statistical periods.

[0082] The monitoring device reads the health index sequence corresponding to the pillar within each observation window, calculates the maximum value, average value, and slope of change of the sequence within the window, where the slope can be obtained by linearly fitting the relationship between the health index and time within the window, and is used to measure the overall upward or downward trend of the health index. When the maximum value or average value of the health index within the window has fallen into the warning level or the severe level, this method marks the risk level of the pillar within the observation window as the alarm level. When the slope exceeds a preset rise threshold, even if the current health index is still at the attention level, the trend can be judged as unfavorable and the attention intensity can be increased accordingly. The rise threshold can be set as a certain limit of the increase in health index per unit time based on experience from type tests and historical operating data. For example, within the time range corresponding to the length of an observation window, an increase in health index of not less than 0.5 is considered to have a significant upward trend.

[0083] To avoid frequent alarms caused by short-term fluctuations, this method requires that a pillar be deemed alarm-level only if it meets the above conditions within at least one complete observation window. Furthermore, if the same pillar maintains a high health index or its slope is consistently positive and its absolute value exceeds a preset threshold across multiple consecutive observation windows, the pillar is considered to have a continuously deteriorating trend. When the alarm conditions are met, the monitoring device generates an early warning record and writes it to local storage. In this scheme, the early warning record refers to a structured record used to trigger an operation and maintenance response, which includes at least the pillar identifier, structural fingerprint version number, coupling fingerprint version number, operating condition sensitivity model version number, risk threshold classification table version number, current partial discharge health index value, corresponding risk level, maximum and average health index values ​​within the observation window, slope of change, early warning generation time, and a unique locally generated identifier.

[0084] The unique identifier is preferably coded by combining the device identifier, support identifier, observation window start time, and current version number. It is used for idempotent control and deduplication of the same warning in both the local system and the maintenance system, preventing duplicate work orders due to retransmission or repeated reporting. After generating a warning record, the monitoring device sends the record to the maintenance system through a configured field communication channel. The communication method can be a local Ethernet network or a fieldbus; the specific method used depends on the structure of the station's automation system, but it must at least ensure that all fields in the warning record can be parsed by the maintenance system.

[0085] The receiving interface of the operation and maintenance system preferably supports parsing the device identifier, support identifier, partial discharge health index, risk level, warning unique marker and various version number fields, and returns a response message containing a status code and confirmation time. When the status code is zero, it means that the operation and maintenance system has successfully received and stored the warning record. When the status code is non-zero, it means that the reception failed or manual intervention is required. For example, it can be agreed that status code one indicates that the request format is incorrect, status code two indicates that authentication failed, and status code three indicates that the version is mismatched.

[0086] To prevent duplicate alarms and control communication load, the monitoring device sends the highest-level warning only once for the same pillar within a single observation window. If the health index of this pillar fluctuates further within the window but the risk level does not increase, no new warning record will be sent again. For duplicate warning requests, the device compares the unique marker of the newly generated warning with the marker already recorded locally. If the markers are found to be the same, it is considered a duplicate and will not be sent again. At the same time, the deduplication behavior is recorded in the operation log for post-event verification.

[0087] When a communication channel times out or fails to receive a confirmation message from the maintenance system within a preset time, this method allows automatic retransmission of the warning record within a limited number of times. The number of retransmissions is preferably capped at three. Each retransmission records the retransmission time, number of retransmissions, and the most recent status code in the operation log. If successful confirmation is still not received after reaching the cap, the monitoring device stops automatic retransmission, marks the warning record as "pending manual processing," and displays it as a reminder on the local interface or in the upper-level system. In this solution, the operation log refers to a log file or log entry that records key operations such as warning generation, warning transmission, communication retry, and status code return in chronological order. This log is used by maintenance personnel to trace the generation and transmission process of partial discharge warnings during fault analysis and responsibility allocation.

[0088] For the aforementioned transformer-type cable support with a partial discharge health index of approximately 3.5 under baseline operating conditions, when it is first determined to be at the severe level within a certain observation window, the monitoring device immediately generates an early warning record based on the currently effective risk threshold classification table. A unique identifier and version number are written into the early warning record and sent to the operation and maintenance system via Ethernet. After successful receipt, the operation and maintenance system includes the support in the recent power outage maintenance plan or arranges on-site verification measures such as infrared thermography and partial discharge retesting based on the risk level and health index value. This forms a closed-loop connection between online partial discharge monitoring results and actual operation and maintenance decisions, enabling this method to not only complete partial discharge detection and risk assessment at the signal processing level, but also achieve an operable and traceable early warning and handling process at the workflow level.

[0089] In the operating scenario shown in this embodiment: In an indoor substation with a rated voltage of 10,000 kV, a newly commissioned medium-voltage switchgear supplies power to three cable feeders. Each feeder circuit is equipped with a current transformer-type cable support, which serves to support the corresponding primary conductor and measure current. Before powering on the equipment, the commissioning personnel first established structural descriptions for the three current transformer-type cable supports in the monitoring device based on the primary system design drawings and the manufacturer's supply information. For the first support, the commissioning personnel measured the conductor diameter at the site using vernier calipers to be approximately 40 mm, the minimum creepage distance between the conductor center and the metal casing of the switchgear to be approximately 800 mm, the exposed height of the support to be approximately 400 mm, the outer diameter to be approximately 150 mm, and the insulation material to be epoxy resin, with a relative permittivity that can be approximated as 4. According to the nameplate and drawings, the cross-sectional dimensions of the current transformer core, the number of secondary turns, the winding links, and the installation position of the grounding terminal were also recorded in the monitoring device. The monitoring device arranges the geometric and electrical parameters related to conductors, insulators, transformer windings, and grounding in a pre-fixed order to form a structural parameter vector. Each parameter is uniformly stored to the decimal place of millimeters or corresponding engineering units, converted into a consistent-length decimal string, and concatenated into a character sequence as the structural fingerprint of that support column. This fingerprint is then bound to the column identifier "First support column of a certain circuit in a certain cabinet," generating a structural fingerprint version number 1, which is written to local non-volatile storage. Simultaneously, a timestamp and the configuration source of "factory delivery + on-site verification" are added. For the second and third supports, the commissioning personnel complete the structural parameter measurement and entry process in the same way. For the third support column, due to the special structure designed to accommodate cable heads, its conductor installation height is slightly different. During entry, the monitoring device identifies this difference as a critical dimensional deviation, but still within the allowable tolerance. Therefore, it generates another set of structural fingerprints based on the actual measured values ​​and marks it as a different fingerprint version for different supports within the same cabinet.

[0090] After determining the structural fingerprint, the station scheduled a power outage calibration during the off-peak load period at night. Once the power outage maintenance began, all relevant circuits were reliably grounded. The commissioning personnel connected a partial discharge calibration pulse generator between the primary conductor end of the first support column and the main grounding busbar of the cabinet. By controlling the rising edge and amplitude of the injected current through a series high-impedance element, the calibration charge was optimally set to one hundred picocoulombs, and dozens of calibration pulses were continuously injected at a rhythm of one to two per second. Under a unified timing clock, the monitoring device opened a high-frequency acquisition window on the corresponding partial discharge detection winding channel at a sampling rate of one hundred mega-times per second, acquiring the complete waveform each time from several hundred nanoseconds before triggering to several microseconds after triggering. After calibration, the monitoring device performs baseline correction and bandpass filtering on this batch of waveforms, attenuating components significantly below tens of kilohertz or above several megahertz in the frequency domain, aligning the waveforms to near the pulse peak, and calling the three-port equivalent model and structural fingerprint parameters of the support pillar. Multiple rounds of calculations are performed in the model using the set equivalent discharge power supply waveform to deduce the calibration pulse response from the equivalent discharge power supply to the partial discharge detection winding. In multiple test results, the monitoring device performs median statistics on the response at each time sampling point, considering samples deviating from the median by more than three times the noise fluctuation in the no-signal area as interference and discarding them. Finally, a representative pulse response sequence is obtained, starting at zero within a few microseconds, rising rapidly, and then oscillating and decaying to near zero. This sequence, along with the calibration charge, 100 megahertz sampling rate, identification time window length, and the number of trials involved in the statistics, is registered as the coupling fingerprint of the support pillar. A coupling fingerprint version number (1) is assigned and bound to the structural fingerprint version number (1) and the support pillar identifier, and written to the read-only log area. The second and third pillars were calibrated in the same way. The third pillar had a slightly different pulse response waveform attenuation tail due to a slight difference in conductor arrangement. The coupling fingerprint recorded this difference, which provided a physical basis for distinguishing the partial discharge behavior of different pillars in subsequent online operation.

[0091] After completing the structural fingerprint and coupling fingerprint registration for all supports, the switchgear was officially put into operation. The monitoring device activated the unified timing clock, periodically broadcasting the timing information to each high-frequency acquisition unit and power frequency current acquisition unit at a resolution of one microsecond. The time axis was divided into continuous operating slices every twenty power frequency cycles, with each slice approximately 0.4 seconds long and the initial phase aligned to a fixed power frequency phase position. During operation, the circuit to which the first support belongs gradually carried a load of approximately 50% of the rated current. The temperature given by the temperature sensor inside the cabinet fluctuated slowly around 25 degrees Celsius, and the humidity recorded by the relative humidity sensor was around 50%. Within each operating slice, the monitoring device continuously acquired the partial discharge detection winding signals of the three supports and the corresponding power frequency current waveforms of the circuits. It also sampled the temperature and humidity signals at intervals of several seconds. At the end of the slice, it calculated the average values ​​of temperature and humidity within that slice, removed any significant abrupt changes, and recorded them as representative values ​​for temperature and humidity for that slice. Simultaneously, based on the correspondence between the zero point or peak position in the power frequency current waveform and the unified timing information, it calculated the power frequency phase at any given moment. The monitoring device archives the high-frequency waveform segments, current waveform segments, and representative values ​​of temperature and humidity and load current collected within this operating slice according to the operating slice identifier and support column identifier, forming a series of operating slice records with consistent structure. The record header indicates the current time synchronization configuration version and the operating slice rule version.

[0092] After a period of operation, the insulation of the third support showed early signs of aging near a cable head, leading to frequent partial discharge activity under combined high load and high humidity conditions. Within the corresponding support's operating slice, the monitoring device retrieved the support's coupling fingerprint, using the impulse response as a feature to perform matched filtering and deconvolution operations on the high-frequency signal. Within several operating slices, dozens of partial discharge events meeting the amplitude threshold requirements were identified. The equivalent charge, converted to the representative value of the calibrated charge range, fell within the hundreds of picocoulombs. Meanwhile, the other two supports in the same cabinet only exhibited a small number of low-amplitude events at occasional intervals under similar operating conditions. By using a minimum time interval constraint, the monitoring device merged multiple peak values ​​with intervals less than a few microseconds into a single event. For each event, the power frequency phase was calculated based on a unified timing, generating a partial discharge event record containing the equivalent charge, power frequency phase, occurrence time, support identifier, operating slice identifier, and coupling fingerprint version number, which was then appended to the event log. As summer temperatures and humidity rise, the number of partial discharge events for the third pillar in multiple continuously operating slices is significantly higher than that for the other two pillars. This is reflected in the event log as a denser number of event entries for this pillar, and an overall rightward shift in the distribution of equivalent charge.

[0093] After running for a certain observation period, the monitoring device performs condition-based conversion on the partial discharge event records of each support column within that period, according to a pre-configured statistical period, such as several operating slices as one period. For the third support column, at the end of each operating slice, the monitoring device reads the representative values ​​of load current, temperature, and humidity from the operating slice records to form the operating condition vector for that slice. It then looks up the charge gain coefficient and repetition rate gain coefficient for the corresponding operating condition interval in the operating condition sensitivity model. The equivalent charge and repetition rate corresponding to the number of partial discharge events in that slice are converted to the baseline operating condition according to the operating condition coefficients, i.e., the equivalent values ​​under the conditions of 50% rated current, 25 degrees Celsius, and 50% relative humidity. Within a statistical period, the monitoring device calculates the average value, percentile value, and percentage of events exceeding a certain reference threshold for the equivalent charge and repetition rate after conversion for all operating slices. Simultaneously, it statistically analyzes the proportion of events clustered within a specific power frequency phase interval, obtaining statistical quantities for three dimensions: converted partial discharge amplitude, frequency, and phase distribution stability. Based on pre-defined weights (e.g., 0.5, 0.3, and 0.2), these three dimensions are mapped to sub-indicators and weighted to generate the partial discharge health index for the third pillar within that statistical period. Because the third pillar exhibits high charge, high repetition rate, and phase concentration within a specific phase interval under multiple high-load and high-humidity conditions, its converted statistical results are generally higher. The health index gradually rises from nearly two to approximately three and a half over several consecutive statistical periods, while the health indices of the other two pillars remain below one for an extended period.

[0094] At the end of each statistical period, the monitoring device, based on the current operating condition sensitivity model version and health index calculation rule version, calls the risk threshold classification table with version numbers, assigning the third pillar's health index of 3.5 to the severe level. Simultaneously, using a sliding observation window encompassing ten statistical periods on the time axis, it calculates the maximum value, average value, and time-fitted slope of the pillar's health index sequence within the window. At a certain moment, when the maximum value and average value of the third pillar's health index within the observation window are both between the warning and severe levels, and the net increase in the slope within one window length exceeds 0.5, the monitoring device determines that the pillar has reached the alarm level within that observation window and generates an alarm record. This record includes the pillar identifier, structural fingerprint version number 1, coupling fingerprint version number 1, current operating condition sensitivity model version number, risk threshold classification table version number, the maximum value, average value, and slope of the health index within the observation window, the current health index value, and the alarm generation time. The monitoring device uses a unique identifier formed by combining the device identifier, pillar identifier, observation window start time, and the aforementioned version numbers, writes this identifier into the alarm record, and sends the alarm record to the operation and maintenance system via the station's Ethernet network. After verifying the format, authentication, and version information, the operation and maintenance system's receiving interface returns a status code of zero and a confirmation time, indicating that the warning record has been successfully entered into the database and has entered the work order management process. Based on this, the monitoring device marks the warning as confirmed locally and will not send the same level of warning to the same pillar again in the same observation window. Instead, it writes the health index and warning status to the health assessment log for subsequent operation and maintenance analysis.

[0095] When the operation and maintenance system aggregates early warning records from multiple switchgears, it includes the third support pillar in the upcoming power outage maintenance plan based on the fact that the health index of the third support pillar has been consistently at the severe level and the trend of change is significantly upward. The system schedules a power outage for this circuit during the next planned maintenance window to conduct on-site partial discharge retesting, insulation inspection, and cable head treatment. Through this process, the station has achieved a complete set of practical operational scenarios, from structural fingerprint modeling, calibration of coupled fingerprints, operational slice acquisition, event extraction, operating condition conversion and health index calculation, to risk classification, early warning generation, and operation and maintenance closed loop. Those skilled in the art can understand the causal relationships and data transmission methods between the steps, and reproduce this method under similar voltage levels and equipment conditions to achieve online monitoring and early warning of partial discharge status of transformer-type cable supports.

[0096] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0097] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0098] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center containing one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0099] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0100] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0101] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0102] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0103] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0104] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0105] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for partial discharge detection and early warning of transformer-type cable supports, characterized in that, include: S1. Obtain the conductor, insulation, transformer winding and grounding parameters of the current transformer type cable support, establish a three-port equivalent model, generate a structural fingerprint, and bind it to the support identifier. S2. When the power is off, inject a partial discharge calibration pulse with a known charge amount, collect the response waveform in the partial discharge detection winding, and generate a coupled fingerprint by inverting the pulse response from the pulse source to the detection winding based on the structural fingerprint. S3. Configure the operation slice rules under a unified time clock, synchronously collect the partial discharge detection winding signals, current signals, temperature signals and humidity signals of each support, and archive them according to the operation slice. S4. Within each running slice, the signal is filtered and deconvolved using the pulse response corresponding to the coupled fingerprint, partial discharge events are extracted, and the equivalent charge and phase information are calculated to generate the partial discharge event record for that support. S5. Based on the load current, temperature and humidity of the operating slice, form the operating condition vector. According to the operating condition sensitivity model, convert the partial discharge equivalent charge and repetition rate to the reference operating condition, and calculate the partial discharge health index of a single transformer cable support. S6. Compare the partial discharge health index and the risk threshold classification table to determine the risk level. When the risk level reaches the alarm level, generate an early warning record containing the pillar identifier, partial discharge health index and risk level and send it to the operation and maintenance system.

2. The method for partial discharge detection and early warning of a transformer-type cable support according to claim 1, characterized in that, S1 includes: When establishing a three-port equivalent model, the monitoring device assembles conductor parameters, insulation parameters, transformer winding parameters, and grounding parameters into a structural parameter vector in a preset order. The monitoring device converts each parameter item in the structural parameter vector into a fixed-length decimal string and concatenates them in a preset order to generate a structural fingerprint. The structural fingerprint, along with the support identifier consisting of the switch cabinet number, circuit number, and support column number, and the structural fingerprint version number, is written into a non-volatile storage medium.

3. The method for partial discharge detection and early warning of a transformer-type cable support according to claim 1, characterized in that, S2 include: When the switchgear is de-energized and grounded, the monitoring device injects a partial discharge calibration pulse of equivalent charge between the primary conductor and the grounding terminal of the current transformer cable support through a partial discharge calibration pulse generator and a current limiting element. Under the constraint of a unified timing clock, the partial discharge detection winding acquires the high-frequency response of the partial discharge calibration pulse within the high-frequency acquisition window through the high-frequency acquisition channel, generates a time-stamped original waveform record, and performs time alignment, baseline correction and bandpass filtering preprocessing on the original waveform record in sequence.

4. The method for partial discharge detection and early warning of a transformer-type cable support according to claim 3, characterized in that: The monitoring device calls the structural fingerprint and three-port equivalent model corresponding to the support column identification, treats the equivalent power supply as a known input, convolves the pulse response to be identified with the waveform of the equivalent power supply to obtain the predicted response, and compares it point by point with the preprocessed original waveform record. The pulse response is iteratively adjusted under the constraint that the amplitude of the pulse response is zero at the time start and decays to below the amplitude threshold within a set time window to obtain the calibration pulse response from the equivalent discharge power supply to the partial discharge detection winding. The calibration pulse response is used as a coupling fingerprint and written into the read-only log area together with the support column identifier, structural fingerprint version number and calibration conditions.

5. The method for partial discharge detection and early warning of a transformer-type cable support according to claim 1, characterized in that, S3 include: The monitoring device adds time stamps to the partial discharge detection winding signal, current signal, temperature signal and humidity signal based on a unified time clock, and divides the time axis into continuous and non-overlapping operating slices according to the operating slice rules; Within each operating segment, the partial discharge detection winding signal, current signal, representative temperature value, and representative humidity value of each current transformer cable support are synchronously collected. The corresponding signal segments are associated with the operating segment identifier and the support identifier to generate an operating segment record and store it in the local storage medium.

6. The method for partial discharge detection and early warning of a transformer-type cable support according to claim 1, characterized in that, S4 include: For transformer-type cable supports, the calibration pulse response in the coupling fingerprint corresponding to the support identification is called to filter and deconvolve the partial discharge detection winding signal based on the calibration pulse response; Constraints are applied during the deconvolution process to ensure that the equivalent discharge current is non-zero only within a limited time sample sequence and zero outside of that time sample sequence. The timing of the partial discharge event is determined based on the amplitude threshold and the minimum time interval. The equivalent charge of the partial discharge event is calculated based on the reconstructed equivalent discharge current and the sampling time interval. The equivalent charge, current phase, running slice identifier, pillar identifier, and coupling fingerprint version number are used to generate a partial discharge event record and write it to the event log.

7. The method for partial discharge detection and early warning of a transformer-type cable support according to claim 1, characterized in that, S5 include: After obtaining the partial discharge event records within the operating slice, the monitoring device constructs an operating condition vector based on the representative values ​​of the load current, cabinet temperature, and cabinet relative humidity of the operating slice. The charge gain coefficient and repetition rate gain coefficient are obtained from the working condition vector in the working condition sensitivity model. The equivalent charge and repetition rate of partial discharge are converted to the reference operating condition using the charge gain coefficient and repetition rate gain coefficient.

8. The method for partial discharge detection and early warning of a transformer-type cable support according to claim 7, characterized in that: The monitoring device performs statistical processing on the partial discharge equivalent charge sequence and partial discharge repetition rate sequence converted to the reference operating condition within the statistical period; This results in indicators such as charge quantity, repetition rate, and partial discharge phase distribution stability. The partial discharge health index is obtained by normalizing the charge quantity index, repetition rate index, and partial discharge phase distribution stability index according to preset weights and then summing them by weight.

9. The method for partial discharge detection and early warning of a transformer-type cable support according to claim 1, characterized in that, S6 include: Configure a risk threshold classification table with version number. After calculating the partial discharge health index, map the partial discharge health index to a risk level based on the risk threshold classification table. In the observation window, determine whether the alarm level has been reached based on the partial discharge health index sequence. When the alarm level is reached, an alarm record is generated, which includes device identifier, support identifier, partial discharge health index, risk level and alarm unique identifier. The record is sent to the operation and maintenance system through the field communication channel. The alarm unique identifier is used to control alarm deduplication and resending within the preset resending limit. The alarm generation and sending are also recorded in the operation log.

Images