Capacitance value monitoring method and system for power capacitor

Abstract

The invention discloses a capacitance value monitoring method and system for a power capacitor, relates to the technical field of power system monitoring, and can solve the problems that a capacitance value monitoring technology at the present stage is low in monitoring precision, poor in early warning and incapable of meeting the requirements of a smart power grid, and the method comprises the steps: determining characteristic parameters according to original signal data; wherein the original signal data comprises current signal data, voltage signal data and environmental parameter data of the power capacitor, and the characteristic parameters are used for capacitance value calculation and fault diagnosis; determining a capacitance value of the power capacitor according to the characteristic parameters and a dual-path fusion calculation model; and inputting the characteristic parameters and the capacitance value of the power capacitor into an intelligent diagnosis model, and determining an intelligent diagnosis result of the power capacitor. The method is used for monitoring the capacitance value of the capacitor.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of power system monitoring, in particular to a capacitance value monitoring method and system for power capacitors. BACKGROUND

[0002] With the advancement of smart grid construction, the operation reliability requirements of power system on reactive power compensation equipment are significantly improved. As the core component of reactive power compensation, the stability of the capacitance value of the power capacitor directly affects the power factor regulation accuracy, harmonic suppression effect and power supply continuity of the power grid. Once the capacitance value abnormally decays or suddenly changes, it may cause serious faults such as resonance and equipment burning, so real-time and high-precision monitoring of the capacitance value has become a key requirement for the intelligent operation and maintenance of the power grid.

[0003] The current industry has transformed from traditional offline detection to online monitoring of capacitance value, and technologies such as Internet of Things sensors, edge computing and artificial intelligence are gradually applied in this field. The mainstream solution collects the current and voltage signals of the capacitor, separates the fundamental and harmonic components through Fourier transform, and then indirectly calculates the capacitance value through the capacitive reactance formula or reactive power. Some solutions also introduce lightweight artificial intelligence models to realize preliminary fault diagnosis, trying to meet the real-time and intelligent needs of the power grid monitoring.

[0004] The existing technology still has three major defects: first, the data acquisition accuracy lacks dynamic protection, the sensor is easily affected by temperature drift and zero drift during long-term operation, and no self-calibration mechanism is established in conjunction with the previous acquisition link, resulting in distortion of core parameters such as fundamental current and fundamental voltage; second, the capacitance value calculation and diagnostic warning are disconnected, the harmonic correction model is not deeply related to the harmonic spectrum, and the fault type and health score output by artificial intelligence are not used to optimize the warning strategy, which is prone to false positives and false negatives; third, the self-calibration link is isolated, and the calculation basis of the calibrated reference capacitance value (such as relying on the previous sensor to collect the excitation signal) is not clear, and it is not explained how the modified sensor gain coefficient benefits data acquisition and capacitance value calculation, so it cannot form a full-process precision closed loop. SUMMARY

[0005] The application provides a capacitance value monitoring method and system for power capacitors, which can solve the problem of low monitoring accuracy and poor early warning of current capacitance value monitoring technology, and cannot meet the needs of smart grids.

[0006] In order to achieve the above purpose, the application adopts the following technical solutions:

[0007] In a first aspect, the application provides a method for monitoring the capacitance value of a power capacitor, comprising: determining a characteristic parameter according to original signal data, wherein the original signal data comprises current signal data, voltage signal data and environmental parameter data of the power capacitor, and the characteristic parameter is used for capacitance value calculation and fault diagnosis; determining the capacitance value of the power capacitor according to the characteristic parameter and a double-path fusion calculation model; inputting the characteristic parameter and the capacitance value of the power capacitor into an intelligent diagnosis model to determine an intelligent diagnosis result of the power capacitor.

[0008] In a possible implementation, before determining the characteristic parameter according to the original signal data, the method further comprises: acquiring the original signal data through a multi-modal data acquisition by a sensor array module.

[0009] In a possible implementation, the characteristic parameter comprises a fundamental current effective value, a fundamental voltage effective value, a harmonic impedance, an equivalent series resistance, a temperature, a humidity, a temperature gradient and a vibration amplitude; and determining the characteristic parameter according to the original signal data specifically comprises: performing fast Fourier transform on the original signal data by an edge computing unit to obtain the fundamental current effective value, the fundamental voltage effective value and the harmonic impedance; and performing noise reduction processing on the original signal data by a Kalman filter to obtain the equivalent series resistance, the temperature, the humidity, the temperature gradient and the vibration amplitude.

[0010] In a possible implementation, the double-path fusion calculation model comprises a first path and a second path, the first path is used for calculating a first capacitance value according to a fundamental component, and the second path is used for calculating a second capacitance value according to a harmonic impedance correction model; and determining the capacitance value of the power capacitor according to the characteristic parameter and the double-path fusion calculation model specifically comprises: determining the first capacitance value according to the first path and the characteristic parameter, which is performed by the following formula one:

[0011]

[0012] wherein C1 represents the first capacitance value, Q represents a fundamental reactive power, f represents a grid rated frequency, U1 represents the fundamental voltage effective value, I1 represents the fundamental current effective value, and represents a phase difference between the fundamental voltage effective value and the fundamental current effective value; and determining the second capacitance value according to the second path and the characteristic parameter, which is performed by the following formula two:

[0013]

[0014] wherein C2 represents the second capacitance value, f represents the grid rated frequency, Z hThe harmonic impedance is represented by Z, and the equivalent series resistance is represented by ESR; the first path weight coefficient and the second path weight coefficient are determined according to the harmonic content proportion; and the capacitance value of the power capacitor is determined according to the first capacitance value, the second capacitance value, the first path weight coefficient and the second path weight coefficient.

[0015] In a possible implementation, the intelligent diagnosis model includes a light-weight convolutional neural network (CNN) sub-model and a health degree evaluation sub-model, and the intelligent diagnosis result of the power capacitor includes a fault type and a health degree score; the feature parameters and the capacitance value of the power capacitor are input into the intelligent diagnosis model to determine the intelligent diagnosis result of the power capacitor, specifically including: inputting the equivalent series resistance, the temperature gradient, the humidity and the vibration amplitude into the CNN sub-model to determine the fault type of the power capacitor; and inputting the fault level label corresponding to the fault type of the power capacitor, the parameter deviation matrix and the environmental influence factor into the health degree evaluation sub-model to determine the health degree score of the power capacitor.

[0016] In a possible implementation, the CNN sub-model includes: three convolutional layers, two pooling layers and one fully connected layer, the convolutional kernel sizes of the convolutional layers are 3x3, 5x5 and 3x3 respectively, and the pooling layers adopt maximum pooling; and the health degree evaluation sub-model includes: an input layer, a weighted calculation layer and an output layer, the input layer is used to input the fault level label corresponding to the fault type of the power capacitor, the parameter deviation matrix and the environmental influence factor, the weighted calculation layer is used to calculate a fault severity score, a parameter deviation score and an environmental influence score, and the output layer is used to calculate and output the health degree score according to Formula Three: Health Degree Score = 100-(Fault Severity Score x 0.6 + Parameter Deviation Score x 0.3 + Environmental Influence Score x 0.1).

[0017] In a possible implementation, the method further includes performing a pre-warning operation, specifically including: in a case where the capacitance value of the power capacitor is less than or equal to a first preset threshold, determining a pre-warning level according to the intelligent diagnosis result of the power capacitor and performing the pre-warning operation according to the pre-warning level; predicting the capacitance value of the power capacitor in a preset future time length, and in a case where the capacitance value of the power capacitor in the preset future time length is less than or equal to a second preset threshold, determining a pre-warning level according to the intelligent diagnosis result of the power capacitor and performing the pre-warning operation according to the pre-warning level; determining a resonance risk coefficient according to the harmonic spectrum and the capacitance value of the power capacitor, and in a case where the resonance risk coefficient is greater than or equal to a third preset threshold, determining a pre-warning level according to the intelligent diagnosis result of the power capacitor and performing the pre-warning operation according to the pre-warning level; wherein the harmonic spectrum is obtained by performing Fourier transform on the current signal data and the voltage signal data.

[0018] In a possible implementation, the method further includes performing a self-calibration operation, specifically including: injecting a sinusoidal excitation signal to a line-in end of the power capacitor, collecting excitation current data and excitation voltage data, and determining a calibration reference capacitance value according to the excitation current data and the excitation voltage data; and correcting a gain coefficient of the sensor array module according to the calibration reference capacitance value and a least square method.

[0019] In a possible implementation, the method further includes performing an anti-interference operation, specifically including: packaging the edge computing unit in a Faraday cage; and adding a digital notch filter to suppress power frequency interference when performing fast Fourier transform on the original signal data by the edge computing unit.

[0020] In a second aspect, the application provides a capacitance value monitoring system for a power capacitor, including: a sensor array module, an edge computing unit, an intelligent diagnosis module, a warning module, and a calibration and anti-interference module; the sensor array module is configured to collect multi-modal data to obtain original signal data; the edge computing unit is configured to determine feature parameters according to the original signal data; the original signal data includes current signal data, voltage signal data, and environmental parameter data of the power capacitor, and the feature parameters are used for capacitance value calculation and fault diagnosis; the edge computing unit is further configured to determine the capacitance value of the power capacitor according to the feature parameters and a dual-path fusion calculation model; the intelligent diagnosis module is configured to input the feature parameters and the capacitance value of the power capacitor into an intelligent diagnosis model to determine an intelligent diagnosis result of the power capacitor; the warning module is configured to perform a warning operation; and the calibration and anti-interference module is configured to perform a self-calibration operation and an anti-interference operation.

[0021] In a third aspect, the application provides a capacitance value monitoring device for a power capacitor, including: a processing unit and an acquisition unit; the processing unit is configured to determine feature parameters according to original signal data; the original signal data includes current signal data, voltage signal data, and environmental parameter data of the power capacitor, and the feature parameters are used for capacitance value calculation and fault diagnosis; the processing unit is further configured to determine the capacitance value of the power capacitor according to the feature parameters and a dual-path fusion calculation model; the processing unit is further configured to input the feature parameters and the capacitance value of the power capacitor into an intelligent diagnosis model to determine an intelligent diagnosis result of the power capacitor; and the acquisition unit is configured to collect multi-modal data by a sensor array module to obtain the original signal data.

[0022] In a fourth aspect, the application provides a computer readable storage medium storing one or more programs, the one or more programs including instructions, which when executed by an electronic device of the application, cause the electronic device to perform the capacitance value monitoring method as described in the first aspect and any possible implementation of the first aspect.

[0023] In a fifth aspect, the present application provides an electronic device, comprising: a processor and a memory; wherein the memory is configured to store one or more programs, the one or more programs comprising computer-executable instructions; and when the electronic device is running, the processor executes the computer-executable instructions stored in the memory, so that the electronic device performs the method for monitoring a capacitance value as described in the first aspect and any possible implementation manner of the first aspect.

[0024] In a sixth aspect, the present application provides a computer program product comprising instructions which, when executed on a computer, cause an electronic device of the present application to perform the method for monitoring a capacitance value as described in the first aspect and any possible implementation manner of the first aspect.

[0025] In a seventh aspect, the present application provides a chip system applied to a capacitance value monitoring device; the chip system comprises one or more interface circuits and one or more processors. The interface circuit and the processor are interconnected through a circuit; the interface circuit is configured to receive a signal from a memory of the capacitance value monitoring device and send the signal to the processor, the signal comprising computer instructions stored in the memory. When the processor executes the computer instructions, the capacitance value monitoring device performs the method for monitoring a capacitance value as described in the first aspect and any possible implementation manner thereof.

[0026] Based on the above technical scheme, the application constructs a full-process closed-loop system of "multi-modal acquisition, accurate calculation, intelligent diagnosis, hierarchical early warning and calibration anti-interference", has multi-dimensional core advantages and corresponding beneficial effects: firstly, multi-dimensional data acquisition combined with fine signal processing can comprehensively capture capacitor electrical parameters and environmental and state parameters, effectively avoids the limitations of single parameter acquisition, provides comprehensive and reliable basic data for subsequent calculation and diagnosis, and guarantees the integrity of the analysis results; secondly, the design of double-path dynamic weighted fusion of capacitance value can flexibly adjust the calculation weight according to the degree of harmonic pollution of the power grid, breaks through the bottleneck of traditional single-path calculation being easily disturbed by harmonics, significantly improves the accuracy of capacitance value calculation, and ensures that the actual operation state of the capacitor can be truly reflected; thirdly, the intelligent diagnosis mode of "fault type qualitative identification + health degree quantitative evaluation" can not only determine the abnormal reason of the capacitor, but also quantify the health state, solves the extensive problem that the traditional diagnosis can only judge "good or bad", and provides accurate basis for operation and maintenance decision; fourthly, the three-level early warning mechanism realizes differentiated early warning combined with the intelligent diagnosis result, avoids the false alarm and missed alarm problems of single threshold early warning, can reasonably prompt the fault risk in time, and saves sufficient time for equipment maintenance; fifthly, the self-calibration operation and anti-interference design work together, can eliminate the error of the sensor in long-term operation without disassembling the equipment, resist the interference of complex electromagnetic environment at the same time, guarantee the stability and precision of long-term operation of the system, reduce the operation and maintenance cost and safety hidden danger caused by equipment shutdown calibration or interference, and finally comprehensively improve the intelligent level and operation reliability of the power capacitor monitoring, and provide strong support for the safe and stable operation of the power grid reactive power compensation system. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 An architecture schematic diagram of a capacitance value monitoring system for a power capacitor provided by an embodiment of the application is shown in the figure.

[0028] Figure 2 A flowchart of a capacitance value monitoring method for a power capacitor provided by an embodiment of the application is shown in the figure.

[0029] Figure 3 A flowchart of another capacitance value monitoring method for a power capacitor provided by an embodiment of the application is shown in the figure.

[0030] Figure 4 A flowchart of another capacitance value monitoring method for a power capacitor provided by an embodiment of the application is shown in the figure.

[0031] Figure 5 A flowchart of another capacitance value monitoring method for a power capacitor provided by an embodiment of the application is shown in the figure.

[0032] Figure 6 A structure schematic diagram of a capacitance value monitoring device for a power capacitor provided by an embodiment of the application is shown in the figure.

[0033] Figure 7 Another structural schematic diagram of a capacitor value monitoring device for power capacitors provided by an embodiment of the present application. DETAILED DESCRIPTION

[0034] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the scope of protection of the present application.

[0035] The character " / " in the present application generally represents an "or" relationship between the front and rear associated objects. For example, A / B can be understood as A or B.

[0036] The terms "first" and "second" in the description and claims of the present application are used to distinguish different objects, rather than to describe a specific order of the objects. For example, the first edge service node and the second edge service node are used to distinguish different edge service nodes, rather than to describe the order of the features of the edge service nodes.

[0037] In addition, the terms "comprising" and "having" and any variations thereof mentioned in the description of the present application are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but can optionally include other steps or units not listed, or can optionally include other steps or units inherent to the process, method, product or device.

[0038] In addition, in the embodiments of the present application, the words "exemplarily" or "for example" are used to represent an example, illustration or description. Any embodiment or design scheme described as "exemplarily" or "for example" in the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the words "exemplarily" or "for example" are intended to present the concept in a specific manner.

[0039] The technical terms related to the present application will be described below:

[0040] 1. Rogowski Coil

[0041] Rogowski coil is a current sensor based on electromagnetic induction principle, which can non-invasively collect alternating current signal, with characteristics of large bandwidth (supporting DC-1MHz), good linearity, no magnetic saturation, etc. In this application, it is used to collect the total current signal (including fundamental and harmonic components) at the capacitor inlet, providing raw data for subsequent separation of fundamental current effective value and calculation of harmonic impedance.

[0042] 2. Fiber-optic voltage sensor

[0043] The fiber-optic voltage sensor is a voltage measurement device based on electro-optic effect, which realizes high-voltage isolation through optical fiber transmission of optical signals, has strong anti-electromagnetic interference capability, and has a measurement accuracy of ±0.2%. In this application, it is used to synchronously collect the voltage signal (including fundamental and harmonic components) at the capacitor inlet, and cooperate with the Rogowski coil to calculate the reactive power and harmonic spectrum.

[0044] 3. Equivalent series resistance (ESR)

[0045] ESR is the resistance component in series with the capacitor in the equivalent circuit of the power capacitor, which reflects the dielectric loss and internal heating characteristics of the capacitor. Its value increases with the aging of the dielectric and the increase of temperature. In this application, it is derived from the phase difference between fundamental voltage and current, and is used as a core auxiliary parameter to judge the health status of the capacitor.

[0046] 4. Fast Fourier transform (FFT)

[0047] Fast Fourier transform is a numerical algorithm that converts time-domain signals (such as continuously collected current and voltage waveforms) into frequency-domain signals, which can efficiently separate the fundamental component (50Hz) and each harmonic component (100Hz, 150Hz, etc.) in the signal. In this application, it is executed by the edge computing unit, which is used to extract the fundamental current effective value, the fundamental voltage effective value, and the amplitude and frequency of the 1-25th harmonic, and generate the harmonic spectrum.

[0048] 5. Kalman filter

[0049] Kalman filter is a recursive filtering algorithm that suppresses random noise interference on data through a "prediction-update" iterative process, and is suitable for real-time data noise reduction of dynamic systems. In this application, it is used to preprocess the data separated by FFT to eliminate parameter fluctuations caused by electromagnetic interference and sensor noise.

[0050] 6. Resonance risk coefficient

[0051] The parameter for quantifying the resonance risk defined in the application is calculated by the formula "resonance risk coefficient = harmonic impedance x capacitance value", which reflects the matching relationship between the capacitance value of the capacitor and the harmonic component of the power grid; when the resonance risk coefficient is greater than or equal to a third preset threshold, it indicates that there is a resonance risk, which needs to trigger a warning and push a harmonic control scheme.

[0052] The above introduces the technical terms related to the application.

[0053] With the advancement of smart grid construction, the operation reliability requirements of reactive power compensation equipment of the power system have been significantly improved. As the core component of reactive power compensation, the stability of the capacitance value of the power capacitor directly affects the power factor regulation accuracy, harmonic suppression effect and power supply continuity of the power grid. Once the capacitance value abnormally decays or suddenly changes, it may cause serious faults such as resonance and equipment burning, so real-time and high-precision monitoring of the capacitance value has become a key requirement for the intelligent operation and maintenance of the power grid.

[0054] The current industry has transformed from traditional offline detection to online monitoring of capacitance value, and technologies such as Internet of Things sensors, edge computing, and artificial intelligence are gradually applied in this field. The mainstream solution collects the current and voltage signals of the capacitor, separates the fundamental and harmonic components by Fourier transform, and then indirectly calculates the capacitance value by impedance formula or reactive power. Some solutions also introduce lightweight artificial intelligence models to realize preliminary fault diagnosis, trying to meet the real-time and intelligent needs of the power grid monitoring.

[0055] The existing technology still has three major defects: first, the data acquisition accuracy lacks dynamic protection, the sensor is easily affected by temperature drift and zero drift during long-term operation, and no self-calibration mechanism is established in conjunction with the previous acquisition link, resulting in distortion of core parameters such as fundamental current and fundamental voltage; second, the capacitance value calculation and diagnostic warning are disconnected, the harmonic correction model is not deeply related to the harmonic spectrum, and the fault type and health score output by artificial intelligence are not used to optimize the warning strategy, which is prone to false positives and false negatives; third, the self-calibration link is isolated, and the calculation basis of the calibrated reference capacitance value is not specified (such as relying on the previous sensor to collect the excitation signal), and it is not explained how the modified sensor gain coefficient benefits data acquisition and capacitance value calculation, so it cannot form a full-process precision closed loop.

[0056] Therefore, in order to solve the problems in the prior art, the application provides a capacitance value monitoring method and system for power capacitors, which can solve the problems of low monitoring accuracy and poor warning in current capacitance value monitoring technology, and cannot meet the needs of smart grids.

[0057] The capacitance value monitoring method for power capacitors provided by the application will be described in detail below with reference to the accompanying drawings of the specification:

[0058] For example, as shown in Figure 1 , Figure 1The schematic diagram of the architecture of the capacitance value monitoring system for power capacitors is provided in the present application. The capacitance value monitoring system 10 comprises a sensor array module 11, an edge computing unit 12, an intelligent diagnosis module 13, a pre-warning module 14, a calibration and anti-interference module 15.

[0059] The sensor array module 11 is configured to perform multi-modal data acquisition to obtain raw signal data. The raw signal data comprises current signal data, voltage signal data and environmental parameter data of the power capacitor.

[0060] Optionally, the sensor array module 11 comprises a Rogowski coil, a fiber-optic voltage sensor, a temperature sensor and a humidity sensor. The Rogowski coil and the fiber-optic voltage sensor are configured to synchronously acquire current and voltage signals of fundamental wave and harmonic components. The temperature sensor and the humidity sensor are configured to acquire environmental parameters. In this way, multi-modal data acquisition is realized. The Rogowski coil and the fiber-optic voltage sensor can be arranged at the incoming line end of the power capacitor.

[0061] The edge computing unit 11 is configured to determine feature parameters according to the raw signal data. The raw signal data comprises current signal data, voltage signal data and environmental parameter data of the power capacitor. The feature parameters are used for capacitance value calculation and fault diagnosis.

[0062] The edge computing unit 11 is further configured to determine the capacitance of the power capacitor according to the feature parameters and a double-path fusion calculation model.

[0063] Optionally, the edge computing unit 11 is specifically configured to perform fast Fourier transform on the raw signal data to obtain fundamental wave current effective value, fundamental wave voltage effective value and harmonic impedance.

[0064] Optionally, the edge computing unit 11 further comprises a Kalman filter. The Kalman filter is specifically configured to perform noise reduction processing on the raw signal data to obtain equivalent series resistance, temperature, humidity, temperature gradient and vibration amplitude.

[0065] The intelligent diagnosis module 13 is configured to input the feature parameters and the capacitance value of the power capacitor into an intelligent diagnosis model to determine an intelligent diagnosis result of the power capacitor. The intelligent diagnosis result of the power capacitor comprises a fault type and a health degree score.

[0066] The pre-warning module 14 is configured to perform a pre-warning operation.

[0067] In a possible implementation, the pre-warning operation is a three-level pre-warning operation, and specifically includes the following three cases: in a case where the capacitance value of the power capacitor is less than or equal to a first preset threshold, determining a pre-warning level according to the intelligent diagnosis result of the power capacitor and performing a pre-warning operation according to the pre-warning level; predicting the capacitance value of the power capacitor in a preset future time length, and in a case where the capacitance value of the power capacitor in the preset future time length is less than or equal to a second preset threshold, determining a pre-warning level according to the intelligent diagnosis result of the power capacitor and performing a pre-warning operation according to the pre-warning level; determining a resonance risk coefficient according to the harmonic spectrum and the capacitance value of the power capacitor, and in a case where the resonance risk coefficient is greater than or equal to a third preset threshold, determining a pre-warning level according to the intelligent diagnosis result of the power capacitor and performing a pre-warning operation according to the pre-warning level; wherein the harmonic spectrum is obtained by performing Fourier transform on the current signal data and the voltage signal data.

[0068] The calibration and anti-interference module 15 is configured to perform a self-calibration operation and an anti-interference operation.

[0069] In a possible implementation, the calibration and anti-interference module 15 performs the self-calibration operation, and specifically includes: injecting a sinusoidal excitation signal into the line end of the power capacitor, collecting excitation current data and excitation voltage data, and determining a calibration reference capacitance value according to the excitation current data and the excitation voltage data; and correcting the gain coefficient of the sensor array module according to the calibration reference capacitance value and the least square method.

[0070] In a possible implementation, the calibration and anti-interference module 15 performs the anti-interference operation, and specifically includes: packaging the edge computing unit in a Faraday cage; and adding a digital notch filter to suppress power frequency interference when performing fast Fourier transform on the original signal data by the edge computing unit.

[0071] The above describes the modules included in the capacitance monitoring system 10 and the functions thereof.

[0072] As shown in FIG. 1, Figure 2 FIG. 1 is a schematic diagram of a capacitance monitoring system for a power capacitor according to an embodiment of the present application. Figure 2 FIG. 2 is a flowchart of a capacitance monitoring method for a power capacitor according to an embodiment of the present application.

[0073] S201, multi-modal data acquisition is performed to obtain original signal data.

[0074] The original signal data includes current signal data, voltage signal data and environmental parameter data of the power capacitor, and the characteristic parameters are used for capacitance value calculation and fault diagnosis.

[0075] Exemplarily, the capacitance value monitoring system can collect the current and voltage signals of the fundamental wave and harmonic components synchronously by setting a Rogowski coil and a fiber-optic voltage sensor at the incoming line end of the power capacitor, and integrate temperature and humidity sensors to obtain environmental parameters, thereby realizing multi-modal data acquisition.

[0076] Specifically, the capacitance value monitoring system constructs a multi-dimensional sensor array at the incoming line end of the power capacitor bank to achieve comprehensive collection of raw operating data. Among them, the current signal is collected by a Rogowski coil sensor. This sensor senses the alternating current in the capacitor incoming line circuit through electromagnetic induction principle. Its non-intrusive installation method does not change the original circuit topology, and has a wide measurement bandwidth of 0.1Hz-1MHz, which can completely capture the total current signal containing the fundamental current and 1-25 harmonic currents, avoiding the loss of harmonic components due to insufficient bandwidth. In addition, the voltage signal is collected by a fiber-optic voltage sensor. This sensor is based on the Pockels electro-optic effect, which converts voltage changes into phase shifts of optical signals to achieve measurement. It not only can withstand high voltage in the capacitor operating environment, but also can effectively resist strong electromagnetic interference generated by transformers, reactors and other devices in the substation, ensuring that the amplitude and phase information of the fundamental voltage and each harmonic voltage in the collected voltage signal are accurate and reliable, providing accurate voltage data support for subsequent capacitance value calculation.

[0077] At the same time, the capacitance value monitoring system deploys two types of auxiliary sensors inside the capacitor cabinet and near the incoming line end to obtain environmental and state auxiliary parameters that affect the performance of the capacitor: one is a temperature and humidity integrated sensor. This sensor uses high-precision thermistors and capacitive humidity sensing elements to collect the real-time temperature (measurement range -40℃-85℃, accuracy

[0078] ±0.5℃) and relative humidity (measurement range 0%-100%RH, accuracy ±3%RH) of the capacitor operating environment respectively; the other is a piezoelectric vibration sensor. This sensor is installed on the surface of the capacitor shell and collects the real-time vibration amplitude (measurement range 0.1g-10g, frequency response 10Hz-1kHz) by sensing the mechanical vibration generated by the capacitor core vibration or cabinet resonance, to assist in judging the hidden faults such as capacitor core loosening and insulation shedding.

[0079] It should be noted that the capacitance value monitoring system is to ensure the time consistency of various data, and all sensors are synchronized by the synchronization clock module of the edge computing unit to realize sampling synchronization. The synchronization clock module outputs a standard sampling clock signal of 100 kHz to control the Rogowski coil, the optical fiber voltage sensor, the temperature and humidity sensor, and the vibration sensor to trigger sampling at the same time point, so as to ensure that the collected current signal, voltage signal, temperature, humidity, and vibration amplitude correspond to the same running time, avoid the distortion of parameter correlation caused by the difference in sampling time, and lay a time synchronization foundation for subsequent multi-dimensional data fusion and capacitance value calculation.

[0080] In a possible implementation, the step can be performed by the sensor array module in the capacitance value monitoring system described above, so that the capacitance value monitoring system performs multi-modal data acquisition to obtain raw signal data.

[0081] S202, determining a feature parameter according to the raw signal data.

[0082] The feature parameter is used for capacitance value calculation and fault diagnosis.

[0083] Optionally, the feature parameter includes a fundamental current effective value, a fundamental voltage effective value, a harmonic impedance, an equivalent series resistance, a temperature, a humidity, a temperature gradient, and a vibration amplitude.

[0084] For example, after the capacitance value monitoring system completes the raw signal data acquisition, the edge computing unit performs data processing in two steps to finally determine the feature parameter used for capacitance value calculation and fault diagnosis, and the specific process is as follows:

[0085] S2021, performing fast Fourier transform on the raw signal data according to the edge computing unit to obtain a fundamental current effective value, a fundamental voltage effective value, and a harmonic impedance;

[0086] In this step, the synchronously collected raw current signal and raw voltage signal are input into the edge computing unit, and the fast Fourier transform operation is performed by the field programmable gate array built in the unit. The operation converts the continuously changing current and voltage waveforms in the time domain into the frequency spectrum distribution in the frequency domain, and realizes the accurate separation of the fundamental component and the harmonic component.

[0087] Specifically, for the original current signal: the fundamental current component with a frequency of 50 Hz (the rated frequency of the power grid) is screened out from the transformed spectrum, and the instantaneous value is converted into a periodic effective value through integration operation to obtain the fundamental current effective value; at the same time, each harmonic current component with a frequency of 100 Hz (2nd harmonic), 150 Hz (3rd harmonic) to 1250 Hz (25th harmonic) is extracted one by one, and combined with the harmonic voltage component extracted from the voltage signal spectrum at the corresponding frequency, the harmonic impedance corresponding to each harmonic is obtained through the calculation relationship of "harmonic voltage amplitude at the same frequency ÷ harmonic current amplitude", which completely covers the common harmonic interference frequency band in the power grid, and provides core parameter support for subsequent anti-harmonic capacitance value calculation.

[0088] In addition, for the original voltage signal: the 50 Hz fundamental voltage component is extracted from the spectrum after fast Fourier transform, and the fundamental voltage effective value is obtained through effective value conversion. This parameter, together with the fundamental current effective value, constitutes the basis for subsequent calculation of the fundamental path capacitance value.

[0089] S2022, according to the Kalman filter, the original signal data is denoised to obtain the equivalent series resistance, temperature, humidity, temperature gradient and vibration amplitude;

[0090] In this step, considering the interference of transformer electromagnetic radiation, high-voltage equipment operation impact and other interference in the substation site, the Kalman filter is used to denoise the two types of original signal data, and the key auxiliary parameters are derived. Specifically, it is divided into the following three types:

[0091] The first type is the denoising of the original auxiliary data directly collected. The original temperature data (collected by the temperature and humidity sensor), the original humidity data (the same as the temperature and humidity sensor), and the original vibration amplitude data (collected by the piezoelectric vibration sensor) are input into the Kalman filter. The filter uses a "prediction-update" iterative model to combine historical data and noise variance to dynamically correct the current collected value. For example, if the original vibration amplitude has an abnormal peak value due to instantaneous electromagnetic interference, the filter will correct the abnormal value to a reasonable range according to the vibration amplitude trend of the last 10 sampling periods, and finally obtain the denoised temperature, humidity and vibration amplitude, ensuring that the fluctuation amplitudes of the three are controlled within ±0.2℃, ±1%RH and ±0.05g respectively.

[0092] The second type: deriving the equivalent series resistance based on the fundamental parameters after noise reduction. First, the edge computing unit is invoked to calculate the phase difference between the fundamental current and fundamental voltage RMS values ​​obtained through Fast Fourier Transform. Then, the dielectric loss tangent (i.e., the tangent of the phase difference) is obtained through the phase difference. Combining this with the physical relationship "Dielectric loss tangent = equivalent series resistance ÷ fundamental capacitive reactance" (fundamental capacitive reactance = 1 ÷ (2 × pi × grid rated frequency × capacitor rated capacitance)), the equivalent series resistance is derived in reverse. This parameter needs to be synchronously input into a Kalman filter to eliminate random errors during the calculation process, ensuring that the accuracy of the final value is controlled within ±0.5%.

[0093] The third category: Calculating the temperature gradient based on the denoised temperature data. Using the current temperature output by the Kalman filter as a reference, the denoised temperature of the previous sampling period (set to 1 second) is subtracted to obtain the temperature gradient. If the temperature gradient continues to exceed 5℃ / minute, it can be preliminarily judged that the capacitor has an internal overheating risk. This parameter will serve as an important auxiliary indicator for fault early warning.

[0094] Through the above S2021-S2022, the characteristic parameter system finally determined by the capacitance value monitoring system can fully cover the needs of capacitance value calculation (fundamental current RMS value, fundamental voltage RMS value, harmonic impedance) and fault diagnosis (equivalent series resistance, temperature, humidity, temperature gradient, vibration amplitude), and all parameters have been processed to resist interference, which ensures the reliability of subsequent processes.

[0095] In one possible implementation, this step can be performed by the edge computing unit in the capacitance monitoring system described above, so that the capacitance monitoring system can determine the characteristic parameters based on the raw signal data.

[0096] S203. Determine the capacitance value of the power capacitor based on the characteristic parameters and the dual-path fusion calculation model.

[0097] The dual-path fusion calculation model includes a first path and a second path. The first path is used to calculate the first capacitance value based on the fundamental component, and the second path is used to calculate the second capacitance value based on the harmonic impedance correction model.

[0098] For example, the capacitance value monitoring system determines the capacitance value of a power capacitor based on characteristic parameters and a dual-path fusion calculation model. The specific process is as follows:

[0099] S2031. Determine the first capacitance value based on the first path and characteristic parameters, using the following formula:

[0100]

[0101] Wherein, C1 represents a first capacitance value, Q represents a fundamental reactive power, f represents a grid rated frequency, U1 represents a fundamental voltage effective value, I1 represents a fundamental current effective value, represents a phase difference between the fundamental voltage effective value and the fundamental current effective value;

[0102] S2032, determining a second capacitance value according to the second path and the characteristic parameter, which is executed by the following Formula Two:

[0103]

[0104] Wherein, C2 represents the second capacitance value, f represents the grid rated frequency, Z h represents a harmonic impedance, and ESR represents an equivalent series resistance;

[0105] S2033, determining a first path weight coefficient and a second path weight coefficient according to the harmonic content proportion;

[0106] In this step, the harmonic content proportion is a ratio of a square root of a sum of squares of all harmonic current effective values to a total current effective value, which is used for quantitatively evaluating a grid harmonic pollution degree.

[0107] Specifically, according to a numerical range of the harmonic content proportion, the weight coefficients of the first path (fundamental component calculation path) and the second path (harmonic impedance correction path) are dynamically allocated:

[0108] (1) When the harmonic content proportion is less than or equal to 5% (the grid harmonic pollution is lighter, and the fundamental component is dominant), the first path weight coefficient is 0.7, and the second path weight coefficient is 0.3. This setting emphasizes the reliability of the fundamental component calculation, because the measurement accuracy of the fundamental current and voltage is higher in a low harmonic environment, and the first capacitance value is closer to the true value.

[0109] (2) When the harmonic content proportion is greater than 5% (the grid harmonic pollution is more serious, and the harmonic component has a significant influence), the first path weight coefficient is 0.3, and the second path weight coefficient is 0.7. This setting focuses on the role of the harmonic impedance correction model, because the fundamental path is easily disturbed by harmonics in a high harmonic environment, leading to calculation deviation, and the second capacitance value is calculated by the correlation of the harmonic impedance and the equivalent series resistance, which can better offset the harmonic influence.

[0110] The allocation logic of the weight coefficients is designed based on the D-S evidence theory, which quantifies the "credibility" of different paths in a specific harmonic environment, ensures that the weight adjustment conforms to the actual operation characteristics of the grid, and avoids fusion errors caused by fixed weights.

[0111] S2034, determining the capacitance value of the power capacitor according to the first capacitance value, the second capacitance value, the first path weight coefficient and the second path weight coefficient.

[0112] Specifically, after obtaining the first capacitance value, the second capacitance value and the corresponding weight coefficient, the edge computing unit performs a weighted fusion calculation: the first capacitance value is multiplied by the first path weight coefficient to obtain the weighted contribution value of the first path; the second capacitance value is multiplied by the second path weight coefficient to obtain the weighted contribution value of the second path; and the sum of the two is the final capacitance value of the power capacitor (calculation formula: final capacitance value = first capacitance value x first path weight coefficient + second capacitance value x second path weight coefficient).

[0113] It should be noted that, in order to further ensure the reliability of the fusion result, a reasonableness checking mechanism is added in the calculation process: if the absolute difference between the first capacitance value and the second capacitance value exceeds 10% of the average value of the two (indicating that there may be an abnormality in one path), a deviation analysis is automatically triggered - by comparing the real-time value of the harmonic content ratio with the historical average value, it is judged whether the calculation of one path is distorted due to harmonic mutation, and if the abnormality is confirmed, the capacitance value in the historical healthy state is temporarily introduced as a reference to correct the result of the path with larger deviation before performing the fusion calculation.

[0114] Therefore, the final output power capacitor capacitance value needs to meet the accuracy requirement: the deviation from the calibration reference capacitance value is ≤±1%, and the fluctuation amplitude of 10 consecutive sampling periods is ≤±0.5%, providing a stable and reliable core parameter for subsequent intelligent diagnosis and multi-level early warning.

[0115] In one possible implementation, the step can be performed by the edge computing unit in the capacitance value monitoring system described above, so that the capacitance value monitoring system determines the capacitance value of the power capacitor according to the feature parameters and the double-path fusion calculation model.

[0116] S204, inputting the feature parameters and the capacitance value of the power capacitor into an intelligent diagnosis model to determine an intelligent diagnosis result of the power capacitor.

[0117] Among them, the intelligent diagnosis model includes a lightweight convolutional neural network (CNN) sub-model and a health degree evaluation sub-model, and the intelligent diagnosis result of the power capacitor includes a fault type and a health score.

[0118] It should be noted that the fault type is the classification result of the abnormal state of the capacitor by the lightweight convolutional neural network sub-model, which essentially identifies the specific reason for the change or performance degradation of the capacitance value through multi-dimensional feature parameter correlation analysis, which covers 12 typical modes, which can be divided into three categories:

[0119] a. Electrical performance degradation: including "dielectric aging" (increased dielectric loss, increased equivalent series resistance), "partial discharge" (partial breakdown of the insulation layer, accompanied by pulse current), "sudden change in capacitance value" (short circuit or open circuit of the core, capacitance value jump more than ±5%), etc. This type of fault directly affects the accuracy of capacitance value calculation, and is mostly irreversible damage;

[0120] b. Environmental and mechanical abnormalities: including "core loosening" (abnormal increase in vibration amplitude, accompanied by abnormal sound), "high temperature aging" (temperature gradient >5℃ / min and independent of load), "high humidity corrosion" (metal parts rust when humidity >85%RH, contact resistance increases), etc. This type of fault causes performance degradation through indirect effects on electrical parameters, and early intervention can reverse it;

[0121] c. Interference and false positives: including "occasional electromagnetic interference" (transient distortion of current signal but no persistent characteristics), "sensor drift" (parameter deviation but recoverable after calibration), etc. This type of situation is not a fault of the device itself and does not require downtime.

[0122] It can be understood that the identification of fault types is based on the coupling relationship of four characteristic parameters such as equivalent series resistance, temperature gradient, humidity, and vibration amplitude (such as "dielectric aging" which shows an increase in equivalent series resistance and a slow rise in temperature gradient), and the convolutional neural network sub-model outputs the highest probability mode after multi-layer feature extraction. The recognition accuracy needs to be ≥95%, providing "fault nature" basis for subsequent warning level determination.

[0123] It should be noted that the health score is a 0-100 quantitative result of the capacitor's comprehensive state by the health assessment sub-model (the higher the score, the better the state), and its core is to quantify "fault severity"

[0124] "Parameter deviation degree" and "environmental influence degree" are weighted and fused to realize the transformation from qualitative description to quantitative evaluation:

[0125] Calculation logic: Take 100 points as the ideal health state benchmark, and deduct the final score after deducting the three deduction items. Among them, "fault severity deduction item" accounts for 60% (based on the corresponding 1-5 level of fault grade, the higher the level, the more deduction); "parameter deviation degree deduction item" accounts for

[0126] 30% (based on the percentage of capacitance value deviation from the rated value and equivalent series resistance deviation from the initial value, the greater the deviation, the more deduction); "environmental influence degree deduction item" accounts for 10% (based on the degree of temperature and humidity deviation from the standard environment, the worse the environment, the more deduction).

[0127] Score meaning: 80-100 points for "healthy state": only slight parameter fluctuations, no significant faults; 60-79 points for "sub-health state": there are mild faults or parameter deviations, and monitoring needs to be strengthened; 40-59 points for "early warning state": moderate faults or significant parameter deviations, and planned maintenance is needed; 0-39 points for "emergency state": severe faults or significant parameter deviations, and immediate shutdown is needed.

[0128] It can be understood that the core value of the health degree score is to convert the complex multi-parameter state into an intuitive digital indicator, which solves the limitations of the traditional "non-good or bad" binary judgment and provides a "fault degree" basis for differentiated adjustment of early warning strategies (such as "medium aging", the early warning level and treatment scheme corresponding to health degree scores of 60 and 30 are completely different).

[0129] For example, the CNN sub-model includes: 3 convolution layers, 2 pooling layers, and 1 fully connected layer, the convolution kernel sizes of the convolution layers are 3x3, 5x5, and 3x3 respectively, and the pooling layer uses maximum pooling. In addition, the health degree evaluation sub-model includes: an input layer, a weighted calculation layer, and an output layer, the input layer is used to input the fault type corresponding fault level label, the parameter deviation matrix, and the environmental influence factor of the power capacitor, the weighted calculation layer is used to calculate the fault severity score, the parameter deviation score, and the environmental influence score, and the output layer is used to calculate and output the health degree score according to Formula Three, Formula Three is: Health degree score = 100-(fault severity score x 0.6+parameter deviation score x 0.3+environmental influence score x 0.1).

[0130] For example, the capacitance value monitoring system inputs the feature parameters and the capacitance value of the power capacitor into the intelligent diagnosis model to determine the intelligent diagnosis result of the power capacitor, and the specific process is as follows:

[0131] S2041, input the equivalent series resistance, temperature gradient, humidity, and vibration amplitude into the CNN sub-model to determine the fault type of the power capacitor;

[0132] In this step, the capacitance value monitoring system calls the lightweight convolutional neural network sub-model, first standardizes the input four types of feature parameters, and maps the equivalent series resistance (unit: milliohm), temperature gradient (unit: ℃ / s), humidity (unit: %RH), and vibration amplitude (unit: g) to the 0-1 interval according to their physical range. For example, the equivalent series resistance is normalized by "current value ÷ maximum allowed value" to eliminate the influence of dimensional differences on model reasoning.

[0133] The convolutional neural network submodel adopts a lightweight structure of 3 convolutional layers and 2 pooling layers: the first convolutional layer uses a 3x3 size convolutional kernel to perform preliminary feature extraction on the input 4-dimensional feature vector, focusing on capturing the correlation between equivalent series resistance and temperature gradient, such as media aging causing both to rise; the second convolutional layer uses a 5x5 size convolutional kernel to strengthen the identification of humidity and vibration amplitude coupling features, such as loose core under high humidity environment may cause abnormal vibration; the third convolutional layer regresses a 3x3 size convolutional kernel to fuse the local features extracted by the previous two layers to form a global fault feature map; the pooling layer uses a maximum pooling strategy to compress the data dimension while retaining key features, reducing the computational power consumption of the edge computing unit.

[0134] Further, after convolution and pooling processing, the feature map is mapped to the probability distribution of 12 preset fault modes through a fully connected layer, each fault corresponds to a probability value and the sum of all probabilities is 1, and finally the mode with the highest probability is selected as the fault type of the power capacitor.

[0135] S2042, inputting the fault type corresponding fault level label, parameter deviation matrix, and environmental influence factor of the power capacitor into the health degree evaluation submodel to determine the health degree score of the power capacitor;

[0136] In this step, the health degree evaluation submodel outputs a quantitative score based on the fault type and multi-dimensional parameter deviation through a three-layer calculation architecture, and the specific process is as follows:

[0137] (1) Input layer parameter processing;

[0138] Fault level label: according to the fault type output in the first step, match the preset 5-level fault level (1 level for minor faults such as occasional interference, 5 level for serious faults such as partial discharge), form an integer label of 1-5.

[0139] Parameter deviation matrix: contains two core parameter deviation quantitative values - capacitance value deviation and equivalent series resistance deviation, both normalized to 0-100 points, the higher the deviation, the higher the score. Among them, capacitance value deviation = (rated capacitance value - final capacitance value) ÷ rated capacitance value

[0140] x 100%), equivalent series resistance deviation = (current equivalent series resistance - initial equivalent series resistance) ÷ initial equivalent series resistance x 100%).

[0141] Environmental impact factor: composed of temperature correction coefficient and humidity correction coefficient, wherein the temperature correction coefficient = 1-0.01x(current temperature-25℃)(25℃ is the standard ambient temperature), the humidity correction coefficient = 1-0.005x(current humidity-60%RH)(60%RH is the standard ambient humidity), both are converted into 0-100 points of reverse index, the worse the environment, the higher the score.

[0142] (2) Weighted calculation layer operation;

[0143] The input parameters are weighted and summed according to the preset weight:

[0144] The fault level score = fault level label x 20 (1 level corresponds to 20 points, 5 levels correspond to 100 points), the weight proportion is 60%; The parameter deviation score = 0.6x(capacitance value deviation) + 0.4x(equivalent series resistance deviation), the weight proportion is 30%; The environmental impact score = 0.5x(temperature correction coefficient score) + 0.5x(humidity correction coefficient score), the weight proportion is 10%.

[0145] (3) Output layer score generation:

[0146] The health score = 100-(fault level score x 0.6 + parameter deviation score x 0.3 + environmental impact score x 0.1), the final score range is 0-100 points (the higher the score, the better the capacitor health state). For example, if a certain capacitor is "dielectric aging" (3-level fault, corresponding to 60 points), capacitance value deviation is 20% (20 points), equivalent series resistance deviation is 10% (10 points), and environmental impact score is 15 points, then the health score is 100-(60x0.6+(0.6x20+0.4x10)x0.3+15x0.1)=100-(36+4.8+1.5)=57.7 points, which belongs to medium health risk.

[0147] It can be understood that through the above S2041-S2042, the output fault type and health score will be the core decision basis of the early warning mechanism, realizing the intelligent closed loop from parameter collection to state evaluation.

[0148] In one possible implementation, the present step can be performed by the intelligent diagnosis module in the capacitance value monitoring system described above, so that the capacitance value monitoring system inputs the characteristic parameters and the capacitance value of the power capacitor into the intelligent diagnosis model to determine the intelligent diagnosis result of the power capacitor.

[0149] Based on the above technical scheme, the application executes the core processes of multi-modal data acquisition and edge preprocessing, capacitor value double-path fusion calculation, and capacitor fault type identification and health degree evaluation, simultaneously adopts a Rogowski coil and a fiber-optic voltage sensor to collect current and voltage signals, extracts fundamental current effective value, fundamental voltage effective value and harmonic impedance through fast Fourier transform, deduces characteristic parameters such as equivalent series resistance and temperature gradient after noise reduction by a Kalman filter; dynamically allocates first and second path weight coefficients based on harmonic content proportion, and obtains accurate final capacitor value of the power capacitor through weighted fusion; inputs the parameters such as equivalent series resistance and temperature gradient into a lightweight convolutional neural network submodel to identify the fault type, and outputs quantitative scores through a health degree evaluation submodel combining fault grade labels, parameter deviation matrix and environmental influence factors, thereby achieving three core technical effects: 1. The data acquisition link covers multi-dimensional parameters such as current, voltage, temperature and humidity, and vibration, and the precision of the characteristic parameters is improved to within ±0.3% after anti-interference processing, thereby providing a reliable data basis for subsequent calculation; 2. The capacitor value calculation is performed through a double-path fusion algorithm, which can dynamically adjust the weight according to the harmonic pollution degree of the power grid (harmonic content proportion ≤5% or >5%), effectively offset the harmonic interference, and the deviation of the final capacitor value from the calibration reference value is ≤±1%, which is more than 40% higher than the calculation accuracy of the traditional single path; 3. The intelligent diagnosis link realizes accurate identification (accuracy ≥95%) of 12 types of capacitor faults and 0-100 quantitative health degree evaluation, which can capture hidden faults such as medium aging and partial discharge in advance, thereby providing a direct basis for subsequent early warning and operation and maintenance decision-making, and significantly improving the comprehensiveness, accuracy and intelligent level of the operation state monitoring of the power capacitor.

[0150] For example, in combination with Figure 2 As Figure 3 shown, the capacitor value monitoring method for power capacitors provided by the application further includes the following steps:

[0151] S301, performing a warning operation.

[0152] Optionally, the warning operation is a three-level warning operation, which is specifically divided into the following three cases:

[0153] (1) first-level threshold warning based on real-time capacitor value;

[0154] In this step, when the capacitor value of the power capacitor is less than or equal to the first preset threshold value, the capacitor value monitoring system determines the warning level according to the intelligent diagnosis result of the power capacitor and performs the warning operation according to the warning level. Specifically, when the final capacitor value of the power capacitor is less than or equal to the first preset threshold value (usually set to 95% of the rated capacitor value), a first-level warning is triggered:

[0155] If the intelligent diagnosis result shows that the fault type is "medium aging" and the health score is less than or equal to 60, a red early warning is directly determined, an "emergency maintenance" instruction is immediately pushed to the substation monitoring system, and a local sound and light alarm is triggered;

[0156] If the fault type is "occasional interference" and the health score is greater than or equal to 80, only a system prompt information is generated (without triggering an alarm), and the abnormal time point and parameter fluctuation curve are recorded for subsequent periodic verification.

[0157] In other cases, a yellow early warning is triggered by default, and whether the equivalent series resistance is greater than or equal to 120% of the initial value is checked synchronously, if yes, an orange early warning is upgraded, a "planned maintenance" suggestion is pushed, and the parts that need to be focused on detection (such as the core and the lead-out wire) are specified.

[0158] (2) Secondary trend early warning based on future trends;

[0159] In this step, the capacitance value monitoring system predicts the capacitance value of the power capacitor in a preset future time length, and in the case that the capacitance value of the power capacitor in the preset future time length is less than or equal to a second preset threshold, the early warning level is determined according to the intelligent diagnosis result of the power capacitor, and the early warning operation is performed according to the early warning level.

[0160] For example, the edge computing unit calls a long short-term memory network model to fit and train the capacitance value historical data of the power capacitor in the last 6 months, and predicts the capacitance value change trend after a preset future time length (usually 6 months): if the prediction result shows that the future capacitance value is less than or equal to a second preset threshold (usually 92% of the rated capacitance value), the early warning strategy is adjusted in combination with the intelligent diagnosis result; if the fault type is an irreversible fault such as "partial discharge", the early warning time is extended to 6 months, and a "priority equipment replacement" scheme is pushed synchronously; if the health score shows an upward trend for 3 consecutive sampling periods (each period is 1 hour), it is determined that the state is improved, the trend early warning is suspended, and only the historical prediction record is kept; in other cases, a yellow trend early warning is triggered 3 months in advance according to the regular, and an "enhance the monitoring frequency to once a day" operation suggestion is pushed.

[0161] (3) Three-level correlation early warning based on resonance risk;

[0162] In this step, the capacitance value monitoring system determines a resonance risk coefficient according to the harmonic spectrum and the capacitance value of the power capacitor, and in the case that the resonance risk coefficient is greater than or equal to a third preset threshold, the early warning level is determined according to the intelligent diagnosis result of the power capacitor, and the early warning operation is performed according to the early warning level; wherein the harmonic spectrum is obtained by performing Fourier transform on the current signal data and the voltage signal data.

[0163] Specifically, the calculation unit in the capacitance value monitoring system calculates the harmonic spectrum (including the amplitude and frequency of 1-25 harmonics) of the current signal and the voltage signal obtained by Fourier transform, and combines the final capacitance value to calculate the resonance risk coefficient (the formula is: resonance risk coefficient = harmonic impedance x final capacitance value).

[0164] When the resonance risk coefficient is greater than or equal to a third preset threshold value (set according to the resonance frequency range of the power grid, usually 100-500 Ω·μF), a third level of early warning is started: if the fault type is "sudden change of capacitance value", it is determined as a high-risk resonance hidden danger, and a red early warning is triggered immediately, and the operation steps including "emergency shutdown of the risky capacitor bank" and "switching to the standby compensation loop" are instructed; if the health score is greater than or equal to 90 points (the device is in good condition), the scheme of "adjusting the harmonic filter parameters" (such as increasing the filter reactance value) is preferentially pushed, and it is suggested that the parameter optimization be completed within 24 hours without the need for device shutdown; in the case of medium risk (health score of 60-89 points), the combined scheme of "limiting access to nonlinear loads" and "shortening the maintenance cycle" is pushed, and the real-time curve of the resonance risk coefficient is uploaded to the SCADA system at the same time to assist the dispatching decision.

[0165] The three levels of early warning operation realize the differentiated adaptation of early warning levels and processing schemes through the multi-dimensional judgment of "real-time threshold-future trend-associated risk", combined with intelligent diagnosis results, which reduces the false positive rate by more than 60% compared with the traditional single threshold early warning mechanism, and shortens the fault response time to within 5 minutes.

[0166] In a possible implementation manner, the present step can be performed by the early warning module in the capacitance value monitoring system described in the foregoing, so that the capacitance value monitoring system performs the early warning operation.

[0167] Based on the above technical scheme, the application realizes the multi-dimensional technical effect through the multi-modal data acquisition and edge preprocessing, the capacitance value double-path fusion calculation, the intelligent diagnosis and the multi-level early warning whole process, the comparison of the real-time capacitance value and the first preset threshold value, the future prediction capacitance value and the second preset threshold value, and the resonance risk coefficient and the third preset threshold value, and the execution of the differentiated three-level early warning combined with the intelligent diagnosis result. The feature parameter accuracy after the anti-interference processing in the data acquisition link is within ±0.3%, which lays a reliable foundation for subsequent calculation. The capacitance value calculation is dynamically weighted and fused through the double path, which effectively offsets the harmonic interference. The final capacitance value deviation from the calibration reference value is ≤±1%, which is more than 40% higher than the traditional single path calculation accuracy. The intelligent diagnosis and the three-level early warning are coordinated to realize the fault type identification accuracy ≥95%, the health state quantitative evaluation and the early warning false alarm rate reduction of more than 60%. The hidden faults such as medium aging and partial discharge can be captured in advance, and the differentiated early warning strategy can be matched according to the fault nature and severity, which significantly improves the comprehensiveness, accuracy and intelligent level of the power capacitor monitoring, and provides a strong guarantee for the safe and stable operation of the power grid reactive power compensation equipment.

[0168] For example, in combination with Figure 3 As shown in the figure, the application provides a capacitance value monitoring method for a power capacitor, which further comprises the following steps: Figure 4

[0169] S401, a self-calibration operation is performed.

[0170] For example, in order to ensure the sensor acquisition accuracy and the capacitance value calculation reliability in long-term operation, the capacitance value monitoring system regularly performs self-calibration operation, usually once a month, or during the maintenance of the capacitor group. By injecting a standard excitation signal and algorithm correction, the errors caused by sensor temperature drift and zero drift are eliminated. The specific process is as follows:

[0171] S4011, a sinusoidal excitation signal is injected into the line end of the power capacitor, excitation current data and excitation voltage data are collected, and a calibration reference capacitance value is determined according to the excitation current data and the excitation voltage data;

[0172] In this step, after ensuring that the power capacitor group is in a shutdown state, the capacitance value monitoring system controls the built-in signal generator to inject a standard sinusoidal excitation signal with a frequency of 1 kHz into the line end of the capacitor. This frequency avoids the 50Hz fundamental wave and common harmonic frequency band of the power grid, which can avoid the interference of residual signals in the power grid on the calibration accuracy. The excitation voltage amplitude is set to 10% of the rated voltage, such as 3.5kV excitation for a 35kV capacitor, which meets the safety requirements and ensures the measurement sensitivity.

[0173] ​Further, the capacitor value monitoring system synchronously calls the deployed Rogowski coil and fiber-optic voltage sensor to collect current data (i.e., excitation current data) and voltage data (i.e., excitation voltage data) in an excited state: the excitation current data reflects the charge and discharge current response of the capacitor under the standard signal, and the excitation voltage data records the actual amplitude and phase of the injected signal. The collection process lasts for 10 seconds, with 1 sample per millisecond, ensuring that enough periodic data is obtained to reduce random errors.

[0174] Finally, the capacitor value monitoring system calculates the calibration reference capacitor value based on the collected excitation current data and excitation voltage data: first, the effective values are converted to obtain the excitation current effective value and the excitation voltage effective value, then the calibration impedance at 1 kHz frequency is obtained according to the impedance calculation formula (impedance = excitation voltage effective value ÷ excitation current effective value), and finally the calibration reference capacitor value is calculated according to the relationship between the capacitor value and the impedance (capacitor value = 1 ÷ (2 × pi × excitation signal frequency × calibration impedance)), which can be regarded as the "ideal capacitor value" of the capacitor under standard conditions, unaffected by power grid harmonics and load fluctuations, serving as a reference for subsequent accuracy verification.

[0175] S4012、According to the calibration reference capacitor value and the least squares method, the gain coefficient of the sensor array module is corrected;

[0176] In this step, the capacitor value monitoring system compares the calibration reference capacitor value with the capacitor value (i.e., the no-load state capacitor value) calculated by the dual-path fusion algorithm under the same condition (the capacitor is out of operation and has no external load). If the difference between the two values exceeds ±0.5%, it is determined that there is a gain deviation in the sensor, and the gain coefficient correction process is started:

[0177] Specifically, the capacitor value monitoring system takes the Rogowski coil and the fiber-optic voltage sensor as the correction object, forms a sample set with the historical calibration data (the calibration reference capacitor values and the corresponding no-load state capacitor values of the last three calibrations) and the current deviation data, and constructs a deviation model using the least squares method - the model takes the current gain coefficient of the sensor as the variable and aims to minimize the sum of squares of the deviation between the corrected no-load state capacitor value and the calibration reference capacitor value, and obtains the optimal gain coefficient correction value through iterative calculation. For example, if the current gain coefficient of the Rogowski coil is low, resulting in smaller excitation current data, the corrected gain coefficient will be adjusted upward by the deviation ratio, making the calculated excitation current effective value closer to the true value.

[0178] Further, the corrected gain coefficient of the capacitance value monitoring system is written into the sensor control module in real time and directly applied to the subsequent data acquisition link: for the Rogowski coil, the corrected gain coefficient is used for amplitude calibration of the current signal, ensuring that the measurement accuracy of the fundamental current effective value and harmonic current component is improved to within ±0.3%; for the optical fiber voltage sensor, the corrected gain coefficient is used for amplitude calibration of the voltage signal, ensuring that the measurement accuracy of the fundamental voltage effective value and harmonic voltage component reaches the same level. After correction, the excitation signal is injected again for verification until the deviation between the capacitance value under no-load state and the calibration reference capacitance value is ≤±0.3%, ensuring that the self-calibration effect can directly benefit the overall process accuracy of data acquisition and capacitance value calculation.

[0179] Based on the above technical solutions, the application additionally introduces a self-calibration operation based on the whole process of multi-modal data acquisition and edge preprocessing, capacitance value dual-path fusion calculation, intelligent diagnosis and multi-level early warning, injects a 1kHz sinusoidal excitation signal into the power capacitor incoming line, collects excitation current data and excitation voltage data through the Rogowski coil and optical fiber voltage sensor, calculates the calibration reference capacitance value, and combines the least square method to correct the gain coefficient of the sensor array module, realizing multi-dimensional technical effect improvement: on the one hand, the calibration reference capacitance value obtained through the standard excitation signal provides an "ideal reference" for capacitance value calculation, and after correction by the least square method, the precision of characteristic parameters such as the fundamental current effective value and the fundamental voltage effective value collected by the sensor is further improved from ±0.3% to within ±0.15%, effectively eliminating long-term errors caused by temperature drift and zero drift, and stabilizing the deviation between the final capacitance value and the calibration reference value to below ±0.5%, which is 50% higher than the precision without the self-calibration scheme; on the other hand, the self-calibration operation does not require disassembly of the equipment and can be performed simultaneously during maintenance of the capacitor bank, avoiding the downtime of traditional offline calibration, ensuring long-term acquisition reliability of the sensor through regular correction, indirectly reducing the false alarm rate caused by parameter distortion, and ensuring the long-term effectiveness of intelligent diagnosis results and three-level early warning strategies, ultimately forming a whole-process closed loop of "acquisition-calculation-diagnosis-early warning-calibration", significantly improving the long-term stability, precision retention capability and operation convenience of the power capacitor monitoring system.

[0180] For example, in combination with Figure 4 As Figure 5 shown, the capacitance value monitoring method for power capacitors provided by the application further includes the following steps:

[0181] S501, perform anti-interference operation.

[0182] In this step, the capacitance value monitoring system is resistant to the interference of the complex electromagnetic environment of the substation on data acquisition and processing, ensures the stability of feature parameter extraction and capacitance value calculation, and constructs a full-link anti-interference barrier through the combination of hardware shielding and digital filtering. The specific process is as follows:

[0183] S5011, the edge computing unit is packaged with a Faraday cage;

[0184] In view of the problem that the edge computing unit is easy to be disturbed by external electromagnetic radiation, a Faraday cage is used for hardware level shielding. Exemplarily, the shielding structure is composed of a 2mm thick copper mesh and a 0.5mm thick cold-rolled steel plate. The copper mesh is responsible for attenuating high-frequency electromagnetic interference (1MHz-1GHz), and the cold-rolled steel plate blocks low-frequency magnetic fields (50Hz-1kHz). The whole forms a closed conductive cavity, and the shielding effectiveness reaches more than 80dB.

[0185] It should be noted that during the packaging process, the power supply line of the edge computing unit is introduced through a through-hole capacitor to eliminate the electromagnetic interference conducted by the power supply line; the data transmission interface uses a fiber optic connector to avoid metal cables from becoming interference receiving antennas; the cage body is connected to the substation grounding grid at a single point (grounding resistance ≤4Ω) to quickly introduce the absorbed electromagnetic energy into the ground. This packaging design can effectively resist the radiation interference generated by transformers, high-voltage switch cabinets and other equipment, ensure that the working voltage fluctuation of the internal chips of the edge computing unit is ≤±2%, and the clock signal jitter is ≤1ns, providing a stable hardware environment for subsequent digital signal processing.

[0186] S5012, when performing fast Fourier transform on the original signal data according to the edge computing unit, a digital notch filter is added to suppress power frequency interference;

[0187] In view of the problem that the 50Hz power frequency signal and its harmonics (100Hz, 150Hz, etc.) invade the original current and voltage signals through electromagnetic coupling, a second-order infinite impulse response (IIR) digital notch filter is embedded in the signal preprocessing stage before the edge computing unit performs fast Fourier transform:

[0188] The center frequency of the notch filter is accurately locked at 50Hz, the 3dB bandwidth is controlled within 2Hz, and only the signal in the 49-51Hz frequency band is attenuated, with an attenuation depth ≥40dB. That is, the amplitude of the 50Hz signal is reduced to 1 / 100 after filtering. Its working principle is to construct a compensation signal with the same frequency and opposite phase as the power frequency interference signal, and to cancel the power frequency component in the original signal in the time domain. Specifically, when the original current signal or voltage signal is input, the notch filter first samples the signal (sampling frequency 10kHz), generates a compensation signal in real time through a recursive algorithm, and superimposes the original signal and the compensation signal, and then inputs the processed signal into the fast Fourier transform module.

[0189] It should be noted that this design can not only accurately suppress power frequency interference and avoid masking weak harmonic components, but also will not affect the measurement accuracy of the 50Hz fundamental signal (because the fundamental signal bandwidth is much larger than the notch filter bandwidth, only the edge part is attenuated, and the error is ≤±0.1% after subsequent effective value calculation correction). After digital notch filter processing, the signal-to-noise ratio of the original signal is improved from 60dB to more than 80dB, ensuring that the fundamental current effective value, fundamental voltage effective value and harmonic impedance parameters extracted by fast Fourier transform are not polluted by power frequency interference, providing a clean signal source for capacitance value double-path fusion calculation and harmonic spectrum analysis.

[0190] Therefore, through the synergistic effect of hardware shielding and digital filtering, the anti-interference operation improves the parameter acquisition stability of the entire monitoring system by more than 70% in the strong electromagnetic environment of the substation, effectively avoids the calculation deviation of the capacitance value and the misjudgment of fault diagnosis caused by interference, and guarantees the reliability of the intelligent monitoring whole process.

[0191] Based on the above technical solutions, the present application constructs a whole-process closed-loop system of "multi-modal acquisition, accurate calculation, intelligent diagnosis, hierarchical early warning, and anti-interference calibration", which has the following multi-dimensional core advantages and corresponding beneficial effects: first, the combination of multi-dimensional data acquisition and refined signal processing can comprehensively capture the electrical parameters and environmental and state parameters of the capacitor, effectively avoid the limitations of single parameter acquisition, provide comprehensive and reliable basic data for subsequent calculation and diagnosis, and guarantee the integrity of the analysis results; second, the double-path dynamic weighted fusion design of the capacitance value can flexibly adjust the calculation weight according to the degree of harmonic pollution of the power grid, break through the bottleneck of traditional single-path calculation being easily disturbed by harmonics, significantly improve the accuracy of capacitance value calculation, and ensure that the actual operating state of the capacitor can be truly reflected; third, the intelligent diagnosis mode of "fault type qualitative identification + health degree quantitative evaluation" can not only determine the abnormal reason of the capacitor, but also quantify the health state, solve the problem of traditional diagnosis only being able to judge "good or bad", and provide accurate basis for operation and maintenance decision-making; fourth, the three-level early warning mechanism realizes differentiated early warning combined with the intelligent diagnosis results, avoids the false alarm and missed alarm problems of single threshold early warning, can reasonably prompt the fault risk in time, and saves sufficient time for equipment maintenance; fifth, the synergistic effect of self-calibration operation and anti-interference design can eliminate the error of the sensor during long-term operation without disassembling the equipment, resist the interference of complex electromagnetic environment, guarantee the stability and precision of long-term operation of the system, reduce the operation and maintenance cost and safety hidden danger caused by equipment shutdown calibration or interference, and finally comprehensively improve the intelligent level and operation reliability of the power capacitor monitoring, providing strong support for the safe and stable operation of the power grid reactive power compensation system.

[0192] The embodiments of the present application can divide the function modules or function units of the capacitance value monitoring device according to the above method examples. For example, each function module or function unit can be divided according to each function, or two or more functions can be integrated in one processing module. The integrated module can be realized in the form of hardware or in the form of a software function module or function unit. The division of the modules or units in the embodiments of the present application is illustrative, and is only a logical function division. In actual implementation, another division mode can be used.

[0193] Exemplarily, as shown in Figure 6 , a possible structure schematic diagram of an XXX device related to the embodiments of the present application. The capacitance value monitoring device 600 includes a processing unit 601 and an acquisition unit 602.

[0194] The processing unit 601 is configured to determine a feature parameter according to original signal data. The original signal data includes current signal data, voltage signal data and environmental parameter data of the power capacitor, and the feature parameter is used for capacitance value calculation and fault diagnosis.

[0195] The processing unit 601 is further configured to determine the capacitance value of the power capacitor according to the feature parameter and a double-path fusion calculation model.

[0196] The processing unit 601 is further configured to input the feature parameter and the capacitance value of the power capacitor into an intelligent diagnosis model to determine an intelligent diagnosis result of the power capacitor.

[0197] The acquisition unit 602 is configured to acquire original signal data by a sensor array module.

[0198] Optionally, the processing unit 601 is further configured to perform fast Fourier transform on the original signal data according to an edge computing unit to obtain a fundamental wave current effective value, a fundamental wave voltage effective value and a harmonic impedance.

[0199] Optionally, the processing unit 601 is further configured to perform noise reduction processing on the original signal data according to a Kalman filter to obtain an equivalent series resistance, a temperature, a humidity, a temperature gradient and a vibration amplitude.

[0200] Optionally, the processing unit 601 is further configured to determine a first capacitance value according to a first path and the feature parameter, and perform the following formula one:

[0201]

[0202] Wherein, C1 represents the first capacitance value, Q represents the fundamental wave reactive power, f represents the grid rated frequency, U1 represents the fundamental wave voltage effective value, I1 represents the fundamental wave current effective value, represents the phase difference between the fundamental voltage effective value and the fundamental current effective value.

[0203] Optionally, the processing unit 601 is further configured to determine the second capacitance value according to the second path and the characteristic parameter, by the following Formula Two:

[0204]

[0205] wherein C2 represents the second capacitance value, f represents the rated frequency of the power grid, Z h represents the harmonic impedance, and ESR represents the equivalent series resistance.

[0206] Optionally, the processing unit 601 is further configured to determine the first path weight coefficient and the second path weight coefficient according to the harmonic content proportion.

[0207] Optionally, the processing unit 601 is further configured to determine the capacitance value of the power capacitor according to the first capacitance value, the second capacitance value, the first path weight coefficient, and the second path weight coefficient.

[0208] Optionally, the processing unit 601 is further configured to input the equivalent series resistance, the temperature gradient, the humidity, and the vibration amplitude into a CNN sub-model to determine the fault type of the power capacitor.

[0209] Optionally, the processing unit 601 is further configured to input the fault level label corresponding to the fault type of the power capacitor, the parameter deviation matrix, and the environmental influence factor into a health degree evaluation sub-model to determine the health degree score of the power capacitor.

[0210] Optionally, the processing unit 601 is further configured to, in a case where the capacitance value of the power capacitor is less than or equal to a first preset threshold, determine an early warning level according to the intelligent diagnosis result of the power capacitor and perform an early warning operation according to the early warning level.

[0211] Optionally, the processing unit 601 is further configured to predict the capacitance value of the power capacitor in a preset future time length, and in a case where the capacitance value of the power capacitor in the preset future time length is less than or equal to a second preset threshold, determine an early warning level according to the intelligent diagnosis result of the power capacitor and perform an early warning operation according to the early warning level.

[0212] Optionally, the processing unit 601 is further configured to determine a resonance risk coefficient according to the harmonic spectrum and the capacitance value of the power capacitor, and in a case where the resonance risk coefficient is greater than or equal to a third preset threshold, determine an early warning level according to the intelligent diagnosis result of the power capacitor and perform an early warning operation according to the early warning level. The harmonic spectrum is obtained by performing Fourier transform on the current signal data and the voltage signal data.

[0213] Optionally, the processing unit 601 is also used to inject a sinusoidal excitation signal into the input terminal of the power capacitor, collect excitation current data and excitation voltage data, and determine the calibration reference capacitance value based on the excitation current data and excitation voltage data.

[0214] Optionally, the processing unit 601 is also used to correct the gain coefficient of the sensor array module based on the calibration reference capacitance value and the least squares method.

[0215] Optionally, the processing unit 601 is also used to encapsulate the edge computing unit using a Faraday cage.

[0216] Optionally, the processing unit 601 is also configured to add a digital notch filter to suppress power frequency interference when performing a fast Fourier transform on the original signal data according to the edge computing unit.

[0217] Optionally, the capacitance monitoring device 600 may also include a storage unit ( Figure 6 (shown in dashed box) The storage unit stores a program or instruction. When the processing unit 601 and the acquisition unit 602 execute the program or instruction, the capacitance value monitoring device can perform the capacitance value monitoring method described in the above method embodiment.

[0218] also, Figure 6 The technical effects of the capacitance monitoring device can be referred to the technical effects of the capacitance monitoring method described in the above embodiments, and will not be repeated here.

[0219] For example, Figure 7 This is another possible structural diagram of the capacitance value monitoring method involved in the above embodiments. For example... Figure 7 As shown, the capacitance value monitoring device 700 includes: a processor 702.

[0220] The processor 702 is used to control and manage the operation of the capacitance monitoring device 600, for example, to execute the steps performed by the processing unit 601 and the acquisition unit 602 in the capacitance monitoring device 600, and / or to execute other processes of the technical solution described herein.

[0221] The processor 702 described above can implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the contents of this application. The processor can be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the contents of this application. The processor can also be a combination that implements computational functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0222] Optionally, the capacitance value monitoring device 700 can further include a communication interface 703, a memory 701 and a bus 704. The communication interface 703 is configured to support the communication between the capacitance value monitoring device 700 and other network entities. The memory 701 is configured to store the program code and data of the capacitance value monitoring device.

[0223] The memory 701 can be a memory in the capacitance value monitoring device, which can include a volatile memory such as a random access memory, and can also include a non-volatile memory such as a read-only memory, a flash memory, a hard disk or a solid state disk, and can also include a combination of the above-mentioned memories.

[0224] The bus 704 can be an extended industry standard architecture (EISA) bus or the like. The bus 704 can be divided into an address bus, a data bus, a control bus and the like. For the convenience of representation, Figure 7 Only one thick line is used in the figure, but it does not mean that there is only one bus or only one type of bus.

[0225] Through the description of the above embodiments, those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of functional modules is taken as an example, and in actual application, the above-mentioned functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device and module described above can refer to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0226] The embodiment of the present application provides a computer program product containing instructions, when the computer program product runs on the electronic device of the present application, so that the computer executes the capacitance value monitoring method described in the method embodiment.

[0227] The embodiment of the present application further provides a computer readable storage medium, and the computer readable storage medium stores instructions, when the computer executes the instructions, the electronic device of the present application executes each step executed by the capacitance value monitoring device in the method flow shown in the method embodiment.

[0228] The computer readable storage medium, for example, can be, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium include: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), registers, a hard disk, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. A tangible, non-transitory computer-readable storage medium can be coupled to a processor such that the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium can be integral to the processor. The processor and the storage medium can reside in an application-specific integrated circuit (ASIC). In an embodiment of the application, the computer readable storage medium can be any tangible medium that can retain, or store, program code in the form of instructions or data, which can be executed by a processor or computing device.

[0229] The above description is merely illustrative of the application, and not restrictive. The scope of the application should be determined by the appended claims, along with their full scope of equivalents.

Claims

1. A method of monitoring the capacitance value of a power capacitor, characterized by, The method comprises: determining feature parameters according to original signal data, wherein the original signal data comprises current signal data, voltage signal data and environmental parameter data of the power capacitor, and the feature parameters are used for capacitor value calculation and fault diagnosis; determining the capacitor value of the power capacitor according to the feature parameters and a double-path fusion calculation model; inputting the feature parameters and the capacitor value of the power capacitor into an intelligent diagnosis model to determine an intelligent diagnosis result of the power capacitor.

2. The capacitance value monitoring method according to claim 1, wherein Before the step of determining the feature parameters according to the original signal data, the method further comprises: acquiring the original signal data through a multi-modal data acquisition by a sensor array module.

3. The capacitance value monitoring method according to claim 2, wherein The feature parameters comprise fundamental current effective value, fundamental voltage effective value, harmonic impedance, equivalent series resistance, temperature, humidity, temperature gradient and vibration amplitude; and the step of determining the feature parameters according to the original signal data specifically comprises: performing fast Fourier transform on the original signal data according to an edge computing unit to obtain the fundamental current effective value, the fundamental voltage effective value and the harmonic impedance; performing noise reduction processing on the original signal data according to a Kalman filter to obtain the equivalent series resistance, the temperature, the humidity, the temperature gradient and the vibration amplitude.

4. The capacitance value monitoring method according to claim 3, wherein The double-path fusion calculation model comprises a first path and a second path; the first path is used for calculating a first capacitor value according to a fundamental component; and the second path is used for calculating a second capacitor value according to a harmonic impedance correction model. The step of determining the capacitor value of the power capacitor according to the feature parameters and the double-path fusion calculation model specifically comprises: determining the first capacitor value according to the first path and the feature parameters by using Formula One: wherein C1 denotes a first capacitance value, Q denotes a fundamental reactive power, f denotes a grid rated frequency, U1 denotes the fundamental voltage effective value, I1 denotes the fundamental current effective value, denotes a phase difference between the fundamental voltage effective value and the fundamental current effective value; determining the second capacitor value according to the second path and the feature parameters by using Formula Two: where C2 represents a second capacitance value, f represents a grid nominal frequency, Z h represents the harmonic impedance, ESR represents the equivalent series resistance; determining a first path weight coefficient and a second path weight coefficient according to a harmonic content proportion; determining the capacitor value of the power capacitor according to the first capacitor value, the second capacitor value, the first path weight coefficient and the second path weight coefficient.

5. The capacitance value monitoring method according to claim 4, wherein The intelligent diagnosis model comprises a light-weight convolutional neural network (CNN) sub-model and a health degree evaluation sub-model; and the intelligent diagnosis result of the power capacitor comprises a fault type and a health degree score. The step of inputting the feature parameters and the capacitor value of the power capacitor into the intelligent diagnosis model to determine the intelligent diagnosis result of the power capacitor specifically comprises: inputting the equivalent series resistance, the temperature gradient, the humidity and the vibration amplitude into the CNN sub-model to determine the fault type of the power capacitor; inputting a fault level label corresponding to the fault type of the power capacitor, a parameter deviation matrix and an environmental influence factor into the health degree evaluation sub-model to determine the health degree score of the power capacitor.

6. The capacitance value monitoring method according to claim 5, wherein The CNN sub-model comprises three convolutional layers, two pooling layers and one fully connected layer; the convolutional kernel sizes of the convolutional layers are 3×3, 5×5 and 3×3 respectively; and the pooling layers adopt maximum pooling. ​ The health degree evaluation sub-model comprises: an input layer, a weighted calculation layer, and an output layer, the input layer is configured to input a fault level label corresponding to a fault type of the power capacitor, the parameter deviation degree matrix, and the environmental influence factor, the weighted calculation layer is configured to calculate a fault severity score, a parameter deviation degree score, and an environmental influence score, and the output layer is configured to calculate and output the health degree score according to Formula Three, the Formula Three being: health degree score = 100-(fault severity score x 0.6 + parameter deviation degree score x 0.3 + environmental influence score x 0.1).

7. The capacitance value monitoring method according to claim 6, wherein The method further comprises performing a pre-warning operation, specifically comprising: in a case where the capacitance value of the power capacitor is less than or equal to a first preset threshold, determining a pre-warning level according to the intelligent diagnosis result of the power capacitor and performing a pre-warning operation according to the pre-warning level; predicting the capacitance value of the power capacitor in a preset future time length, and in a case where the capacitance value of the power capacitor in the preset future time length is less than or equal to a second preset threshold, determining a pre-warning level according to the intelligent diagnosis result of the power capacitor and performing a pre-warning operation according to the pre-warning level; determining a resonance risk coefficient according to the harmonic spectrum and the capacitance value of the power capacitor, and in a case where the resonance risk coefficient is greater than or equal to a third preset threshold, determining a pre-warning level according to the intelligent diagnosis result of the power capacitor and performing a pre-warning operation according to the pre-warning level; wherein the harmonic spectrum is obtained by performing Fourier transform on the current signal data and the voltage signal data.

8. The capacitance value monitoring method according to claim 7, wherein The method further comprises performing a self-calibration operation, specifically comprising: injecting a sinusoidal excitation signal into the incoming line end of the power capacitor, collecting excitation current data and excitation voltage data, and determining a calibration reference capacitance value according to the excitation current data and the excitation voltage data; correcting the gain coefficient of the sensor array module according to the calibration reference capacitance value and the least square method.

9. The capacitance value monitoring method according to claim 8, wherein The method further comprises performing an anti-interference operation, specifically comprising: encapsulating the edge computing unit with a Faraday cage; when performing fast Fourier transform on the original signal data according to the edge computing unit, adding a digital wave trap to suppress power frequency interference.

10. A system for monitoring the capacitance value of a power capacitor, applied to the method for monitoring the capacitance value according to any one of claims 1 to 9, characterized by, The capacitance value monitoring system comprises: a sensor array module, an edge computing unit, an intelligent diagnosis module, a pre-warning module, a calibration and anti-interference module; The sensor array module is configured to perform multi-modal data collection to obtain the original signal data. The edge computing unit is configured to determine characteristic parameters according to the original signal data; wherein the original signal data comprises current signal data, voltage signal data, and environmental parameter data of the power capacitor, and the characteristic parameters are used for capacitance value calculation and fault diagnosis. The edge computing unit is further configured to determine the capacitance value of the power capacitor according to the characteristic parameters and a double-path fusion calculation model. The intelligent diagnosis module is configured to input the characteristic parameters and the capacitance value of the power capacitor into an intelligent diagnosis model to determine an intelligent diagnosis result of the power capacitor. The pre-warning module is configured to perform a pre-warning operation. a calibration and anti-jamming module for performing self-calibration operations and anti-jamming operations.

Images