Insulation detection method and system for power equipment

Abstract

The invention provides an insulation detection method and system for power equipment. The method comprises the steps that a detection pulse signal is injected into a power supply loop, and a mixed response signal is collected; performing time-frequency domain conjoint analysis on the mixed response signal, and separating and extracting frequency band characteristic parameters of a dynamic noise signal in a current working period, including a noise center frequency drift distance, a harmonic energy peak value and a frequency spectrum coverage bandwidth; a filtering algorithm corresponding to the current frequency conversion working condition is obtained through similarity matching, and real-time dynamic filtering parameters are generated, so that the stop-band center and the stop-band width of the parameters are adaptively tracked and updated along with the noise center frequency drift distance and the frequency spectrum coverage bandwidth; filtering the mixed response signal based on the dynamic filtering parameter and a filtering algorithm to obtain a reconstruction signal for the insulation leakage current signal; and calculating a real-time insulation resistance value according to the amplitude and the phase of the reconstructed signal. According to the invention, adaptive tracking and filtering of dynamic noise are realized, and the detection precision of the insulation resistance value under a complex frequency conversion working condition is obviously improved.

Description

Technical Field

[0001] This invention relates to the field of power equipment testing, and more particularly to a method and system for testing the insulation of power equipment. Background Technology

[0002] With the rapid development of power systems and the improvement of industrial automation, the safe and stable operation of power supply circuits in plant facilities is crucial for continuous production. Insulation problems in electrical lines and equipment pose significant safety hazards; once insulation performance deteriorates, it severely threatens the stability and continuity of power supply. Currently, the power industry generally uses offline manual inspections and periodic preventative tests for the maintenance of power supply circuit equipment. However, traditional manual inspection methods suffer from long inspection cycles, heavy workloads, and a high risk of missed or incorrect inspections, failing to continuously and promptly reflect the dynamic insulation status of equipment during operation. To overcome the limitations of offline testing, online insulation monitoring technology is gradually being applied. This technology actively injects AC or low-frequency pulse signals into the system under test and measures the response current to achieve online monitoring of insulation resistance, driving the development of equipment insulation maintenance from preventative testing to predictive maintenance.

[0003] Existing online insulation detection technologies typically employ a fixed-frequency signal injection method, combined with traditional hardware low-pass / band-pass filters or software filtering algorithms with fixed parameters, to extract the leakage current signal of the insulation response. However, in modern industrial production sites, numerous frequency converters inject broadband and dynamically changing harmonic noise into the power grid during operation. Due to the real-time changes in the operating frequency and load of the frequency converter, the noise spectrum generated undergoes irregular dynamic drift. When existing technologies are applied to such conditions, the mismatch between the fixed, static filtering mechanism and the dynamically drifting noise leads to severe distortion of the final extracted response signal. This results in a significant deviation between the detected insulation resistance value and the actual value, easily triggering frequent false alarms in monitoring systems and seriously affecting the safety and reliability of power plant operations. Summary of the Invention

[0004] This invention application provides an insulation testing method and system for power equipment to solve the technical problem of how to reduce the deviation between the insulation resistance test result and the actual value.

[0005] To address the aforementioned technical problems, this invention provides a method for insulation testing of power equipment, comprising:

[0006] A detection pulse signal is injected into the power supply circuit of the target power equipment, and a mixed response signal of the power supply circuit is collected; wherein, the mixed response signal includes an insulation leakage current signal excited by the detection pulse signal and a dynamic noise signal generated by the operation of the frequency converter.

[0007] The hybrid response signal is subjected to joint time-domain and frequency-domain analysis, and the frequency band characteristic parameters of the dynamic noise signal within the current working cycle are extracted based on the joint analysis results; wherein, the frequency band characteristic parameters include noise center frequency drift, harmonic energy peak value, and spectral coverage bandwidth;

[0008] The frequency band feature parameters are matched with a pre-built frequency conversion interference feature library to obtain the filtering algorithm corresponding to the current frequency conversion operating condition; and based on the filtering algorithm and the noise center frequency drift, real-time dynamic filtering parameters are generated; wherein, the stopband center and stopband width of the dynamic filtering parameters are adaptively tracked and updated with the noise center frequency drift and the spectrum coverage bandwidth.

[0009] Based on the dynamic filtering parameters and filtering algorithm, the hybrid response signal is filtered to remove the dynamic noise signal that overlaps with the frequency band of the detection pulse signal, thereby obtaining a reconstructed signal for the insulation leakage current signal.

[0010] Based on the amplitude and phase of the reconstructed signal, and combined with the voltage parameters of the detection pulse signal, the real-time insulation resistance value of the power supply circuit of the target power equipment under operating conditions is calculated to achieve insulation detection.

[0011] As a preferred embodiment, the joint time-domain and frequency-domain analysis of the hybrid response signal, and the extraction of frequency band characteristic parameters of the dynamic noise signal within the current working cycle based on the joint analysis results, includes:

[0012] The hybrid response signal is subjected to synchronous compressed wavelet transform, and the signal energy is compressed and rearranged to instantaneous frequency points in the frequency domain to obtain the time-frequency energy distribution matrix.

[0013] Based on the time-frequency energy distribution matrix, the energy ridge trajectory of the dynamic noise signal on the time-frequency plane is extracted using an energy gradient-based ridge extraction algorithm.

[0014] The difference between the instantaneous maximum and minimum frequency values ​​of the energy ridge trajectory within the current working cycle is calculated as the noise center frequency drift of the dynamic noise signal; the maximum energy amplitude point on the energy ridge trajectory is extracted to obtain the harmonic energy peak value.

[0015] Using the energy ridge trajectory as the central axis, perform layer-by-layer energy integration along both sides of the frequency axis in the time-frequency energy distribution matrix to obtain the cumulative integrated energy;

[0016] When the accumulated integrated energy reaches a preset proportion threshold of the total noise energy at the current moment, the upper frequency boundary and the lower frequency boundary corresponding to the current moment are obtained, and the difference between the upper frequency boundary and the lower frequency boundary is used as the spectrum coverage bandwidth.

[0017] As a preferred embodiment, the frequency conversion interference feature library stores several standard operating condition modes, as well as adaptive filtering function templates associated with each of the standard operating condition modes.

[0018] The step of performing similarity matching between the frequency band feature parameters and a pre-built frequency conversion interference feature library to obtain the filtering algorithm corresponding to the current frequency conversion operating condition includes:

[0019] The frequency band characteristic parameters are characterized as a multi-dimensional operating condition feature vector representing the current noise distribution characteristics;

[0020] The multidimensional working condition feature vector is projected onto a preset high-dimensional feature space to obtain the projection result; and the weighted Mahalanobis distance between the projection result and the center of each standard working condition mode is calculated respectively.

[0021] The standard working condition mode with the smallest weighted Mahalanobis distance is selected as the target matching mode, and the adaptive filtering function template corresponding to the target matching mode is called.

[0022] Based on the real-time rate of change of the noise center frequency drift, the order and topology of the filter are determined in the adaptive filtering function template, thereby generating a filtering algorithm for the current frequency conversion condition based on the order and topology.

[0023] As a preferred embodiment, the step of filtering the hybrid response signal based on the dynamic filtering parameters and filtering algorithm to filter out the dynamic noise signal that overlaps with the frequency band of the probe pulse signal, and obtaining a reconstructed signal for the insulation leakage current signal, includes:

[0024] Based on the dynamic filtering parameters and filtering algorithm, a time-varying notch filter is constructed.

[0025] The hybrid response signal is input into the time-varying notch filter to obtain a preliminary filtered signal;

[0026] The preliminary filtered signals are rearranged in reverse time order to obtain the rearrangement result;

[0027] The rearrangement result is input into the time-varying notch filter for secondary filtering to obtain a zero-phase-shift denoised signal.

[0028] Using the detection pulse signal as a reference signal, a dynamic orthogonal reference coordinate system composed of in-phase axes and quadrature axes is constructed; the zero-phase-shift denoised signal is projected onto the dynamic orthogonal reference coordinate system to obtain in-phase components and quadrature components with the same frequency as the detection pulse signal;

[0029] For the overlapping regions of the frequency bands, vector synthesis and amplitude gain compensation calculations are performed using the in-phase and quadrature components to obtain the reconstructed signal.

[0030] As a preferred embodiment, the voltage parameters include the voltage change rate and amplitude;

[0031] The step of calculating the real-time insulation resistance of the power supply circuit of the target power equipment under operating conditions based on the amplitude and phase of the reconstructed signal and the voltage parameters of the detection pulse signal includes:

[0032] Based on the level flip trigger edge of the probe pulse signal, multiple nonlinear feature sampling points are selected within a single pulse period of the reconstructed signal; wherein, the nonlinear feature sampling points include transient sampling points during the capacitor charging transition process and static sampling points under the charging saturation trend;

[0033] An equivalent parallel circuit model including the insulation resistance to be measured and the system's distributed capacitance to ground is established. Combined with the voltage change rate of the probe pulse signal and the instantaneous current values ​​of the multiple nonlinear characteristic sampling points, a system of linear equations is constructed regarding the insulation resistance to be measured and the system's distributed capacitance to ground.

[0034] The linear equations are solved using the least squares iterative algorithm to calculate the capacitive current component in the reconstructed signal corresponding to the system's distributed capacitance to ground, and the capacitive current component is removed from the reconstructed signal to extract the purely resistive leakage current component.

[0035] The insulation resistance value is calculated based on the steady-state amplitude of the pure resistive leakage current component and the amplitude of the detection pulse signal.

[0036] As a preferred embodiment, the step of injecting a probe pulse signal into the power supply circuit of the target power equipment and acquiring the mixed response signal of the power supply circuit further includes:

[0037] A detection pulse signal is injected into the power supply circuit to obtain the system-to-ground distributed capacitance of the power supply circuit and the insulation resistance value obtained in the previous detection cycle; the frequency and amplitude of the detection pulse signal are adjusted according to the system-to-ground distributed capacitance of the power supply circuit and the insulation resistance value obtained in the previous detection cycle.

[0038] The system's distributed capacitance to ground is monitored, and when the system's distributed capacitance to ground increases, the frequency of the detection pulse signal is reduced, and the output amplitude of the detection pulse signal is increased accordingly, so as to ensure that the signal-to-noise ratio of the hybrid response signal is within a preset threshold range.

[0039] As a preferred embodiment, before injecting the probe pulse signal into the power supply circuit of the target electrical equipment, the method further includes:

[0040] Determine the grounding system type of the target power equipment;

[0041] If the grounding system type is a TN system, monitor the operating status of the target power equipment;

[0042] When the target power equipment is in operation, a standby enable command is sent to the insulation monitoring terminal via the MODBUS communication protocol to put the insulation monitoring terminal into standby mode.

[0043] When the target power equipment is detected to have exited the operating state, the insulation monitoring terminal is controlled to start operation to implement offline online monitoring;

[0044] If the grounding system type is an IT system, then the insulation monitoring terminal is controlled to directly enter the real-time online monitoring mode.

[0045] As a preferred embodiment, after calculating the real-time insulation resistance value of the power supply circuit of the target power equipment in its operating state, the method further includes:

[0046] The real-time insulation resistance value is compared with the preset first warning threshold and second alarm threshold respectively;

[0047] When the real-time insulation resistance value is lower than the first warning threshold or the second alarm threshold, the fault location is performed based on the polarity or phase characteristics of the reconstructed signal.

[0048] If the power supply circuit is a DC system, the insulation fault is determined and displayed as occurring at the positive or negative terminal based on the current flow direction of the reconstructed signal.

[0049] If the power supply circuit is a three-phase AC system, the phase correlation between the reconstructed signal and the voltage of each phase is analyzed to determine and display the phase line where the insulation fault occurred.

[0050] Accordingly, this invention application also provides an insulation detection system for power equipment, including a data acquisition module, an extraction module, a filter parameter generation module, a reconstruction module, and an insulation detection module; wherein,

[0051] The acquisition module is used to inject a detection pulse signal into the power supply circuit of the target power equipment and acquire the mixed response signal of the power supply circuit; wherein, the mixed response signal includes an insulation leakage current signal excited by the detection pulse signal and a dynamic noise signal generated by the operation of the frequency converter.

[0052] The extraction module is used to perform joint time-domain and frequency-domain analysis on the hybrid response signal, and extract the frequency band characteristic parameters of the dynamic noise signal within the current working cycle based on the joint analysis results; wherein, the frequency band characteristic parameters include noise center frequency drift, harmonic energy peak value, and spectral coverage bandwidth;

[0053] The filter parameter generation module is used to perform similarity matching between the frequency band feature parameters and a pre-built frequency conversion interference feature library to obtain the filter algorithm corresponding to the current frequency conversion operating condition; and to generate real-time dynamic filter parameters based on the filter algorithm and the noise center frequency drift; wherein, the stopband center and stopband width of the dynamic filter parameters are adaptively tracked and updated with the noise center frequency drift and the spectrum coverage bandwidth.

[0054] The reconstruction module is used to filter the hybrid response signal based on the dynamic filtering parameters and filtering algorithm to filter out the dynamic noise signal that overlaps with the frequency band of the detection pulse signal, and obtain a reconstructed signal for the insulation leakage current signal.

[0055] The insulation detection module is used to calculate the real-time insulation resistance value of the power supply circuit of the target power equipment in the operating state based on the amplitude and phase of the reconstructed signal and the voltage parameters of the detection pulse signal, so as to realize insulation detection.

[0056] As a preferred embodiment, the extraction module performs joint time-domain and frequency-domain analysis on the hybrid response signal, and extracts the frequency band characteristic parameters of the dynamic noise signal within the current working cycle based on the joint analysis results, including:

[0057] The extraction module applies synchronous compressed wavelet transform to the hybrid response signal, compressing and rearranging the signal energy to instantaneous frequency points in the frequency domain to obtain a time-frequency energy distribution matrix.

[0058] Based on the time-frequency energy distribution matrix, the energy ridge trajectory of the dynamic noise signal on the time-frequency plane is extracted using an energy gradient-based ridge extraction algorithm.

[0059] The difference between the instantaneous maximum and minimum frequency values ​​of the energy ridge trajectory within the current working cycle is calculated as the noise center frequency drift of the dynamic noise signal; the maximum energy amplitude point on the energy ridge trajectory is extracted to obtain the harmonic energy peak value.

[0060] Using the energy ridge trajectory as the central axis, perform layer-by-layer energy integration along both sides of the frequency axis in the time-frequency energy distribution matrix to obtain the cumulative integrated energy;

[0061] When the accumulated integrated energy reaches a preset proportion threshold of the total noise energy at the current moment, the upper frequency boundary and the lower frequency boundary corresponding to the current moment are obtained, and the difference between the upper frequency boundary and the lower frequency boundary is used as the spectrum coverage bandwidth.

[0062] As a preferred embodiment, the frequency conversion interference feature library stores several standard operating condition modes, as well as adaptive filtering function templates associated with each of the standard operating condition modes.

[0063] The filter parameter generation module performs similarity matching between the frequency band feature parameters and a pre-built frequency conversion interference feature library to obtain the filter algorithm corresponding to the current frequency conversion operating condition, including:

[0064] The filter parameter generation module characterizes the frequency band feature parameters as a multi-dimensional operating condition feature vector of the current noise distribution characteristics.

[0065] The multidimensional working condition feature vector is projected onto a preset high-dimensional feature space to obtain the projection result; and the weighted Mahalanobis distance between the projection result and the center of each standard working condition mode is calculated respectively.

[0066] The standard working condition mode with the smallest weighted Mahalanobis distance is selected as the target matching mode, and the adaptive filtering function template corresponding to the target matching mode is called.

[0067] Based on the real-time rate of change of the noise center frequency drift, the order and topology of the filter are determined in the adaptive filtering function template, thereby generating a filtering algorithm for the current frequency conversion condition based on the order and topology.

[0068] As a preferred embodiment, the reconstruction module filters the hybrid response signal based on the dynamic filtering parameters and filtering algorithm to remove dynamic noise signals that overlap with the frequency band of the probe pulse signal, thereby obtaining a reconstructed signal for the insulation leakage current signal, including:

[0069] The reconstruction module constructs a time-varying notch filter based on the dynamic filtering parameters and the filtering algorithm;

[0070] The hybrid response signal is input into the time-varying notch filter to obtain a preliminary filtered signal;

[0071] The preliminary filtered signals are rearranged in reverse time order to obtain the rearrangement result;

[0072] The rearrangement result is input into the time-varying notch filter for secondary filtering to obtain a zero-phase-shift denoised signal.

[0073] Using the detection pulse signal as a reference signal, a dynamic orthogonal reference coordinate system composed of in-phase axes and quadrature axes is constructed; the zero-phase-shift denoised signal is projected onto the dynamic orthogonal reference coordinate system to obtain in-phase components and quadrature components with the same frequency as the detection pulse signal;

[0074] For the overlapping regions of the frequency bands, vector synthesis and amplitude gain compensation calculations are performed using the in-phase and quadrature components to obtain the reconstructed signal.

[0075] As a preferred embodiment, the voltage parameters include the voltage change rate and amplitude;

[0076] The insulation detection module calculates the real-time insulation resistance of the power supply circuit of the target power equipment under operating conditions based on the amplitude and phase of the reconstructed signal and the voltage parameters of the detection pulse signal, including:

[0077] The insulation detection module selects multiple nonlinear feature sampling points within a single pulse period of the reconstructed signal based on the level flip trigger edge of the detection pulse signal; wherein, the nonlinear feature sampling points include transient sampling points during the capacitor charging transition process and static sampling points under the charging saturation trend;

[0078] An equivalent parallel circuit model including the insulation resistance to be measured and the system's distributed capacitance to ground is established. Combined with the voltage change rate of the probe pulse signal and the instantaneous current values ​​of the multiple nonlinear characteristic sampling points, a system of linear equations is constructed regarding the insulation resistance to be measured and the system's distributed capacitance to ground.

[0079] The linear equations are solved using the least squares iterative algorithm to calculate the capacitive current component in the reconstructed signal corresponding to the system's distributed capacitance to ground, and the capacitive current component is removed from the reconstructed signal to extract the purely resistive leakage current component.

[0080] The insulation resistance value is calculated based on the steady-state amplitude of the pure resistive leakage current component and the amplitude of the detection pulse signal.

[0081] As a preferred embodiment, the acquisition module injects a detection pulse signal into the power supply circuit of the target power equipment and acquires the mixed response signal of the power supply circuit, and further includes:

[0082] The acquisition module injects a detection pulse signal into the power supply circuit and acquires the system-to-ground distributed capacitance of the power supply circuit and the insulation resistance value obtained in the previous detection cycle; based on the system-to-ground distributed capacitance of the power supply circuit and the insulation resistance value obtained in the previous detection cycle, the frequency and amplitude of the detection pulse signal are adjusted.

[0083] The system's distributed capacitance to ground is monitored, and when the system's distributed capacitance to ground increases, the frequency of the detection pulse signal is reduced, and the output amplitude of the detection pulse signal is increased accordingly, so as to ensure that the signal-to-noise ratio of the hybrid response signal is within a preset threshold range.

[0084] As a preferred embodiment, the insulation detection system further includes a mode selection module, which is used before injecting a detection pulse signal into the power supply circuit of the target electrical equipment:

[0085] Determine the grounding system type of the target power equipment;

[0086] If the grounding system type is a TN system, monitor the operating status of the target power equipment;

[0087] When the target power equipment is in operation, a standby enable command is sent to the insulation monitoring terminal via the MODBUS communication protocol to put the insulation monitoring terminal into standby mode.

[0088] When the target power equipment is detected to have exited the operating state, the insulation monitoring terminal is controlled to start operation to implement offline online monitoring;

[0089] If the grounding system type is an IT system, then the insulation monitoring terminal is controlled to directly enter the real-time online monitoring mode.

[0090] As a preferred embodiment, the insulation detection system further includes a fault location module, which is used after calculating the real-time insulation resistance value of the power supply circuit of the target power equipment in its operating state:

[0091] The real-time insulation resistance value is compared with the preset first warning threshold and second alarm threshold respectively;

[0092] When the real-time insulation resistance value is lower than the first warning threshold or the second alarm threshold, the fault location is performed based on the polarity or phase characteristics of the reconstructed signal.

[0093] If the power supply circuit is a DC system, the insulation fault is determined and displayed as occurring at the positive or negative terminal based on the current flow direction of the reconstructed signal.

[0094] If the power supply circuit is a three-phase AC system, the phase correlation between the reconstructed signal and the voltage of each phase is analyzed to determine and display the phase line where the insulation fault occurred.

[0095] Compared with the prior art, this invention application has the following beneficial effects:

[0096] This invention application provides an insulation detection method and system for power equipment. The insulation detection method includes: injecting a probe pulse signal into the power supply circuit of the target power equipment and acquiring a mixed response signal of the power supply circuit; wherein the mixed response signal includes an insulation leakage current signal excited by the probe pulse signal and a dynamic noise signal generated by the operation of the frequency converter; performing joint time-domain and frequency-domain analysis on the mixed response signal, and extracting the frequency band feature parameters of the dynamic noise signal within the current working cycle based on the joint analysis results; wherein the frequency band feature parameters include noise center frequency drift, harmonic energy peak value, and spectral coverage bandwidth; and performing similarity matching between the frequency band feature parameters and a pre-constructed frequency converter interference feature library to obtain... The system employs a filtering algorithm corresponding to the current frequency conversion operating condition; and generates real-time dynamic filtering parameters based on the filtering algorithm and the noise center frequency drift. The stopband center and stopband width of the dynamic filtering parameters are adaptively updated based on the noise center frequency drift and the spectral coverage bandwidth. The system filters the hybrid response signal based on the dynamic filtering parameters and the filtering algorithm to remove dynamic noise signals overlapping with the frequency band of the probe pulse signal, obtaining a reconstructed signal for the insulation leakage current signal. Based on the amplitude and phase of the reconstructed signal, combined with the voltage parameters of the probe pulse signal, the system calculates the real-time insulation resistance value of the power supply circuit of the target power equipment under operating conditions, thereby achieving insulation detection. This invention application obtains a filtering algorithm corresponding to the current frequency conversion condition by separating and extracting the frequency band feature parameters of the dynamic noise signal within the current working cycle and matching them with a pre-constructed frequency conversion interference feature library. This solves the inherent contradiction in the existing fixed filtering mechanism that cannot simultaneously ensure the integrity of signal extraction and broadband interference suppression. It can adaptively track and update the stopband center and stopband width according to the noise center frequency drift and the spectrum coverage bandwidth. While accurately eliminating dynamic noise signals, it preserves the amplitude and phase of the relatively weak detection pulse signal, effectively eliminates the distortion of leakage current signals, significantly improves the detection accuracy of insulation resistance under complex frequency conversion conditions, and avoids false alarms in the system.

[0097] Furthermore, by performing joint time-frequency domain analysis on the hybrid response current signal, characteristic parameters such as noise drift, energy peak, and bandwidth can be extracted in real time. This allows the system to quantify noise characteristics in real time, even when faced with atypical noise caused by unknown frequency converter access, changes in power grid topology, or equipment aging, without relying on a pre-set fixed noise model. This endows the insulation testing equipment with strong field adaptability and operating condition generalization ability. Attached Figure Description

[0098] Figure 1 This is a schematic flowchart of an embodiment of the insulation testing method for power equipment provided in this application.

[0099] Figure 2 This is a flowchart illustrating a preferred embodiment of the insulation testing method for power equipment provided in this application.

[0100] Figure 3 This is a flowchart illustrating another preferred embodiment of the insulation testing method for power equipment provided in this application.

[0101] Figure 4 This is a flowchart illustrating another preferred embodiment of the insulation testing method for power equipment provided in this application.

[0102] Figure 5 This is a schematic diagram of an embodiment of the insulation detection system for power equipment provided in this application. Detailed Implementation

[0103] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0104] Example 1

[0105] According to relevant technical records, existing online insulation detection technologies typically employ a fixed-frequency signal injection method, combined with traditional hardware low-pass / band-pass filters or software filtering algorithms with fixed parameters, to extract the leakage current signal of the insulation response. However, in modern industrial production sites, numerous frequency converters inject broadband and dynamically changing harmonic noise into the power grid during operation. When existing insulation monitoring equipment using the fixed-period signal injection method is applied to such complex frequency converter systems, the weak measurement signal injected is easily overwhelmed and superimposed by the broadband interference generated by the frequency converter.

[0106] Furthermore, due to the real-time changes in the inverter's operating frequency and load, the noise spectrum it generates will dynamically drift irregularly. When existing technologies are applied to such operating conditions, inherent limitations are exposed: if the filter passband is wide, the dynamically drifting broadband noise is very likely to overlap with the fixed-frequency injection signal, resulting in a large amount of noise mixing into the effective signal; on the other hand, if the filter band is forcibly narrowed or the filter order is increased in order to completely filter out these complex broadband interferences, it will severely weaken the already extremely weak low-frequency effective leakage current signal, and may even cause severe distortion of the phase and amplitude of the useful signal.

[0107] Therefore, the mismatch between the fixed, static filtering mechanism and the dynamic drift noise will lead to serious distortion of the final extracted response signal, resulting in a large deviation between the detected insulation resistance value and the actual value. In practical applications, this can easily cause frequent false alarms in the monitoring system, seriously affecting the safety and reliability of power plant production and operation.

[0108] For one or more of the above technical issues, please refer to Figure 1 , Figure 1 This is a schematic flowchart of an embodiment of the insulation testing method for power equipment provided in this application.

[0109] Figure 1 The illustrated embodiment includes steps S101 to S105; each step is described in detail below:

[0110] Step S101: Inject a detection pulse signal into the power supply circuit of the target power equipment and collect the mixed response signal of the power supply circuit.

[0111] The hybrid response signal includes an insulation leakage current signal excited by the probe pulse signal and a dynamic noise signal generated by the operation of the frequency converter.

[0112] Specifically, the injection method of the probe pulse signal can be an adaptive pulse injection method. In this embodiment, the probe pulse signal is generated by an insulation monitor and applied to the power supply circuit of the target power equipment by superposition. That is, a probe pulse signal on the order of μA is superimposed and injected into the operating power supply system to achieve uninterrupted monitoring of the insulation resistance value without disconnecting the circuit under test. The phase lines L1, L2, and L3 of the target power equipment can be connected to the insulation monitor in sequence via a miniature circuit breaker and a contactor, thereby forming an injection path for the probe pulse signal without affecting the normal power supply of the system.

[0113] Before injecting the probe pulse signal, the injection timing needs to be determined based on the grounding system type of the target power equipment. The grounding system type can be a TN system or an IT system. If the grounding system type is an IT system, the insulation monitor can directly enter the real-time online monitoring mode, continuously injecting probe pulse signals into the power supply circuit and collecting mixed response signals. If the grounding system type is a TN system, it needs to be connected to the standby terminal of the insulation monitor through an offline / online module. Utilizing the logical relationship between the main and auxiliary contacts of multiple 380VAC / 220VAC intermediate relays (or multiple normally open contacts of contactors), a standby enable command is sent to the insulation monitor via the MODBUS communication protocol to achieve automatic control of the injection timing: when the power supply equipment is in normal operation, the insulation monitor remains in standby mode; when the power supply equipment is detected to be out of operation, the insulation monitor automatically starts operation, injecting probe pulse signals into the power supply circuit to implement offline-online insulation monitoring. The purpose of the aforementioned timing control mechanism is to avoid interference from injected signals to the normal operation of power supply equipment during the operation of the TN system, while eliminating monitoring gaps that may occur during the switching between operation and shutdown states, and ensuring continuous capture of insulation degradation risks.

[0114] The acquisition of the hybrid response signal can be accomplished by the dual-detection module built into the insulation monitor. The dual-detection module may include an AC detection module and a DC detection module, respectively suitable for acquiring the response current of AC power supply systems and DC power supply systems, to ensure effective capture of the hybrid response signal under different power supply systems. In this embodiment, the insulation resistance detection range of the insulation monitor can be 0 to 20 MΩ, and the detection accuracy can be ±5%. It should be noted that the range and accuracy are not limited to the above values, and those skilled in the art can adjust them according to actual operating conditions.

[0115] The hybrid response signal contains two components: one is the insulation leakage current signal excited by the probe pulse signal, flowing through the insulation path under test and returning, which carries the real-time insulation status information of the target power equipment; the other is the broadband dynamic harmonic noise injected into the grid by the frequency converter during operation, forming the dynamic noise signal. Because the operating frequency and load of the frequency converter are constantly changing, the spectral distribution of the dynamic noise signal undergoes irregular dynamic drift, and its noise frequency band may overlap with the frequency band of the probe pulse signal, thus interfering with the subsequent extraction of the leakage current signal. The hybrid response signal acquired in step S101 will be used as the input for subsequent steps S102 to S105. Through time-frequency joint analysis, dynamic filtering, and equivalent circuit modeling, the accurate calculation of the real-time insulation resistance value of the target power equipment is finally achieved.

[0116] In some preferred embodiments, injecting a probe pulse signal into the power supply circuit of the target power equipment in step S101 may include the following steps:

[0117] Determine the grounding system type of the target power equipment;

[0118] If the grounding system type is a TN system, monitor the operating status of the target power equipment;

[0119] When the target power equipment is in operation, a standby enable command is sent to the insulation monitoring terminal via the MODBUS communication protocol to put the insulation monitoring terminal into standby mode.

[0120] When the target power equipment is detected to have exited the operating state, the insulation monitoring terminal is controlled to start operation to implement offline online monitoring;

[0121] If the grounding system type is an IT system, then the insulation monitoring terminal is controlled to directly enter the real-time online monitoring mode.

[0122] In this embodiment, the TN system is commonly referred to in Chinese as a protective grounding system or a direct grounding system for the power supply neutral point. This system has one point directly grounded (usually the neutral point), and the exposed conductive parts of electrical equipment (such as metal casings) are connected to this grounding point via a protective conductor (PE or PEN wire). Here, T stands for Terra, indicating that the power supply terminal (transformer neutral point) is directly grounded, and N stands for Neutral, indicating that the exposed conductive parts of the equipment are directly electrically connected to the power supply terminal grounding point (neutral point).

[0123] In common Chinese, IT refers to a system with an ungrounded neutral point or an isolated power supply system. This means that the energized parts of the power supply system are not grounded or are grounded through a high impedance connection, while the exposed conductive parts of the electrical equipment are directly grounded. "I" stands for Isolation, indicating that the power supply is isolated from ground (ungrounded or grounded with high impedance), and "T" stands for Terra, indicating that the exposed conductive parts of the equipment are directly grounded.

[0124] In a preferred embodiment, injecting a probe pulse signal into the power supply circuit of the target power equipment and acquiring the mixed response signal of the power supply circuit includes:

[0125] A detection pulse signal is injected into the power supply circuit to obtain the system-to-ground distributed capacitance of the power supply circuit and the insulation resistance value obtained in the previous detection cycle; the frequency and amplitude of the detection pulse signal are adjusted according to the system-to-ground distributed capacitance of the power supply circuit and the insulation resistance value obtained in the previous detection cycle.

[0126] The system's distributed capacitance to ground is monitored, and when the system's distributed capacitance to ground increases, the frequency of the detection pulse signal is reduced, and the output amplitude of the detection pulse signal is increased accordingly, so as to ensure that the signal-to-noise ratio of the hybrid response signal is within a preset threshold range.

[0127] In this preferred embodiment, the system-to-ground distributed capacitance of the power supply circuit and the insulation resistance value obtained in the previous detection cycle are first acquired. The insulation resistance value obtained in the previous detection cycle can be read in real time by the insulation monitoring instrument from the insulation measurement results completed in the previous detection cycle. The system-to-ground distributed capacitance is obtained by collecting the instantaneous current values ​​of multiple nonlinear characteristic sampling points within a single pulse cycle, combining them with an equivalent parallel circuit model that includes the insulation resistance to be measured and the system-to-ground distributed capacitance, and solving it using a least squares iterative algorithm. This solution process and the insulation resistance value calculation process in step S105 can share the same equivalent circuit framework, thereby achieving synchronous identification of the system-to-ground distributed capacitance without the need for additional hardware.

[0128] Secondly, based on the system's distributed capacitance to ground and the real-time insulation resistance, the frequency and amplitude of the detection pulse signal are adjusted in a coordinated manner to ensure that the signal-to-noise ratio of the hybrid response signal is within a preset threshold range. The physical basis of this coordinated adjustment is as follows: In the equivalent circuit including the system's distributed capacitance to ground, the capacitive impedance is inversely proportional to the frequency of the detection pulse signal; when the system's distributed capacitance to ground increases, the impedance of the capacitive branch decreases accordingly, the current component flowing through the capacitive branch increases relatively, the proportion of the insulation leakage current signal in the hybrid response signal decreases, resulting in a decrease in the signal-to-noise ratio of the effective signal.

[0129] Therefore, when an increase in the system's distributed capacitance to ground is detected, the frequency of the probe pulse signal is reduced to increase the capacitive impedance and suppress the relative proportion of the capacitive current component, thereby restoring the discernibility of the insulation leakage current signal. Simultaneously, the output amplitude of the probe pulse signal is increased to compensate for the potential attenuation of the effective response amplitude due to the reduced frequency. Thus, through coordinated adjustment of frequency and amplitude, the signal-to-noise ratio of the mixed response signal is maintained within the preset threshold range. Conversely, when the system's distributed capacitance to ground decreases, or when a significant decrease in the real-time insulation resistance leads to an increase in the resistive leakage current component, the frequency of the probe pulse signal can be appropriately increased while the output amplitude is correspondingly reduced. This reduces the impact of the injected signal on the system under test while ensuring the signal-to-noise ratio meets requirements.

[0130] In this embodiment, the preset threshold range can be determined comprehensively based on the hardware detection accuracy (±5%) of the insulation monitoring instrument and the noise baseline level of the on-site working conditions; the specific adjustment range of the frequency and amplitude is not limited to fixed values, and those skilled in the art can make adaptive adjustments based on the rated voltage level of the system under test, the range of distributed capacitance, and the interference intensity of the frequency converter.

[0131] The purpose of this step is to dynamically maintain the impedance matching relationship between the probe pulse signal and the circuit under test by sensing the changes in the distributed parameters of the power supply circuit in real time, so as to ensure that the mixed response signal received in the subsequent signal processing steps always has sufficient effective signal components, thereby laying the signal quality foundation for the accurate extraction of dynamic noise characteristic parameters in step S102 and the accurate calculation of insulation resistance in step S105.

[0132] Step S102: Perform joint time-domain and frequency-domain analysis on the hybrid response signal, and extract the frequency band characteristic parameters of the dynamic noise signal within the current working cycle based on the joint analysis results.

[0133] The frequency band characteristic parameters include noise center frequency drift, harmonic energy peak value, and spectral coverage bandwidth.

[0134] Because the spectrum of the dynamic noise signal drifts irregularly with the real-time changes in the inverter's operating frequency and load, its instantaneous frequency exhibits a continuously changing, non-steady-state characteristic in the time domain. Traditional Fourier transforms, based on global time averaging, cannot capture this type of time-varying spectral structure. While conventional continuous wavelet transforms possess time-frequency localization capabilities, they are constrained by the Heisenberg uncertainty principle, resulting in inherent diffusion in their time-frequency energy distribution. This leads to blurred boundaries between different frequency components, making it difficult to accurately track the dynamic trajectory of the instantaneous frequency. Therefore, this embodiment employs synchronous compressed wavelet transform to perform joint time-frequency analysis on the hybrid response signal.

[0135] In some preferred embodiments, such as Figure 2 As shown, the joint analysis of the hybrid response signal in the time and frequency domains, and the extraction of the frequency band characteristic parameters of the dynamic noise signal within the current working cycle based on the joint analysis results, includes steps S201 to S205; each step is detailed below:

[0136] Step S201: Apply synchronous compressed wavelet transform to the hybrid response signal to compress and rearrange the signal energy to instantaneous frequency points in the frequency domain to obtain the time-frequency energy distribution matrix.

[0137] Step S202: Based on the time-frequency energy distribution matrix, the energy ridge trajectory of the dynamic noise signal on the time-frequency plane is extracted using an energy gradient-based ridge extraction algorithm.

[0138] Step S203: Calculate the difference between the maximum and minimum instantaneous frequency values ​​of the energy ridge trajectory within the current working cycle, and use this difference as the noise center frequency drift of the dynamic noise signal; extract the maximum energy amplitude point on the energy ridge trajectory to obtain the harmonic energy peak value;

[0139] Step S204: Using the energy ridge trajectory as the central axis, perform layer-by-layer energy integration along both sides of the frequency axis in the time-frequency energy distribution matrix to obtain the cumulative integrated energy;

[0140] Step S205: When the accumulated integrated energy reaches a preset proportion threshold of the total noise energy at the current moment, the upper frequency boundary and the lower frequency boundary corresponding to the current moment are obtained, and the difference between the upper frequency boundary and the lower frequency boundary is used as the spectrum coverage bandwidth.

[0141] Specifically, in step S201, the hybrid response signal is first subjected to synchronous compressed wavelet transform to compress and rearrange the signal energy to instantaneous frequency points in the frequency domain to obtain the time-frequency energy distribution matrix.

[0142] The synchronous compressed wavelet transform is performed in two stages. In the first stage, the hybrid response signal is used as input, a basic wavelet that meets the admissibility condition is selected, and the continuous wavelet transform coefficients are calculated to obtain the time-frequency representation of the signal under different scales and time shift parameters. Based on this, for each effective wavelet coefficient, the actual instantaneous frequency of the signal energy at that time-frequency point is estimated by calculating the ratio of its phase partial derivative with respect to the time shift parameter to the coefficient itself.

[0143] In the second stage, a synchronous compression operation is performed: based on the instantaneous frequencies of each time point estimated in the first stage, the corresponding wavelet coefficient energy is redistributed from the original scale domain coordinates to the discrete frequency grid points on the frequency axis that are closest to the instantaneous frequency, thus completing the energy compression and rearrangement.

[0144] After the above two-stage processing, the energy that was originally diffused and distributed in the scale domain is concentrated and compressed to the vicinity of each instantaneous frequency point. In the obtained time-frequency energy distribution matrix, the energy of each frequency component appears as clear and narrow high-energy stripes on the time-frequency plane. The time-frequency resolution is significantly better than that of conventional continuous wavelet transform, laying the foundation for subsequent ridge extraction.

[0145] In this embodiment, the basic wavelet can be a Morlet wavelet, and its center frequency and bandwidth parameters can be adaptively tuned according to the sampling rate of the mixed response signal and the operating frequency range of the frequency converter. It should be noted that the type of basic wavelet is not limited to the Morlet wavelet, and other analytic wavelets that meet the allowable conditions can be selected according to the actual signal characteristics.

[0146] Based on the time-frequency energy distribution matrix, the energy ridge trajectory of the dynamic noise signal on the time-frequency plane is extracted using an energy gradient-based ridge extraction algorithm.

[0147] The implementation process of the ridge extraction algorithm based on energy gradient is as follows: At each moment, the first-order gradient of the energy amplitude with respect to frequency in the time-frequency energy distribution matrix is ​​calculated along the frequency axis; the local maximum point of energy corresponds to the zero-crossing position of the gradient, that is, the first-order gradient turns from positive to negative, and the second derivative of the energy amplitude at that point is negative. The frequency position that satisfies the above conditions is identified as the ridge candidate point at that moment.

[0148] To eliminate spurious extrema introduced by low-energy background noise, an energy amplitude threshold is applied to the ridge candidate points, retaining only those with energy amplitudes exceeding a preset percentage of the total energy at the current moment. Subsequently, ridge candidate points at adjacent moments are continuously correlated along the time axis. A greedy matching strategy based on frequency proximity is used to sequentially connect adjacent candidate points with frequency deviations within a preset tolerance range, forming a continuous energy ridge trajectory. This energy ridge trajectory describes the complete path of the instantaneous frequency evolution of the dynamic noise signal over time in the time-frequency plane, and its shape directly reflects the dynamic changes in the inverter's operating frequency and harmonic components. The purpose of this step is to distinguish the main energy ridge of the dynamic noise signal from the fixed frequency components of the probe pulse signal from the time-frequency energy distribution matrix, providing an accurate time-frequency positioning reference for the subsequent quantitative extraction of frequency band characteristic parameters.

[0149] After obtaining the energy ridge trajectory, three frequency band characteristic parameters are calculated sequentially. Specifically, these include:

[0150] The instantaneous frequency values ​​corresponding to each moment of the energy ridge trajectory within the current working cycle are iterated, and the maximum and minimum values ​​are taken respectively. The difference between the two values ​​is the noise center frequency drift of the dynamic noise signal. The noise center frequency drift quantitatively describes the range of frequency components of the dynamic noise signal within a complete working cycle. The larger the value, the stronger the dynamics of the noise spectrum under the current frequency conversion condition, and the wider the stopband range that the subsequent dynamic filtering parameters need to cover. This parameter will be used as one of the core inputs for generating dynamic filtering parameters in step S103 to drive the real-time tracking and updating of the stopband center.

[0151] Along the energy ridge trajectory, the amplitude of the time-frequency energy distribution matrix at each ridge position is compared sequentially, and the point with the largest amplitude is taken as the harmonic energy peak value of the dynamic noise signal. The harmonic energy peak value reflects the strongest time-frequency energy concentration point of the harmonic interference injected by the frequency converter within the current working cycle. It can be used to normalize and compare the energy distribution of each standard operating condition mode in the frequency converter interference feature library during the similarity matching process in step S103, thereby improving the robustness of operating condition identification.

[0152] Furthermore, taking the energy ridge trajectory as the central axis, energy is integrated layer by layer along both sides of the frequency axis in the time-frequency energy distribution matrix to obtain the cumulative integrated energy; when the cumulative integrated energy reaches the preset proportion threshold of the total noise energy at the current moment, the corresponding upper frequency boundary and lower frequency boundary are obtained, and the difference between the two is taken as the spectrum coverage bandwidth.

[0153] Specifically, at each moment, with the current ridge frequency as the center, the integration window is expanded outward layer by layer with a fixed step size to the high-frequency side and low-frequency side of the frequency axis, and the square of the time-frequency energy amplitude at each frequency grid point within the window is accumulated; at the same time, the total noise energy in the full frequency range at the current moment is calculated.

[0154] After each layer of expansion is completed, the ratio of the current cumulative integrated energy to the total noise energy is compared with the preset ratio threshold. When the ratio reaches or exceeds the preset ratio threshold for the first time, the expansion is stopped, and the frequency positions corresponding to the two ends of the integration window at this time are recorded as the upper frequency boundary and the lower frequency boundary, respectively. The difference between the two is the spectrum coverage bandwidth at the current moment.

[0155] In this embodiment, the preset ratio threshold can be 0.95, that is, the frequency range width corresponding to covering 95% of the total noise energy at the current moment is used as the spectrum coverage bandwidth. It should be noted that the preset ratio threshold is not limited to 0.95. Those skilled in the art can adjust it according to the trade-off between noise suppression integrity and filter passband protection. For example, in the case of relatively concentrated noise spectrum, the threshold can be appropriately reduced to narrow the stopband, and in the case of relatively diffuse noise spectrum, the threshold can be appropriately increased to ensure the integrity of noise suppression.

[0156] The spectrum coverage bandwidth and the noise center frequency drift together constitute a complete quantitative description of the frequency band characteristics of the dynamic noise signal under the current frequency conversion condition: the former reflects the frequency expansion range of the noise at a certain moment, and the latter reflects the overall wandering amplitude of the noise center frequency throughout the entire working cycle.

[0157] This preferred embodiment, without relying on any pre-set fixed noise model, uses high-resolution time-frequency representation of synchronous compressed wavelet transform to quantify the frequency band characteristics of dynamic noise signals in real time. This endows the system with adaptive sensing capabilities for unknown frequency converter models, atypical noise conditions, and changes in power grid topology, providing reliable real-time input for the accurate generation of subsequent dynamic filtering parameters.

[0158] Step S103: Perform similarity matching between the frequency band feature parameters and the pre-built frequency conversion interference feature library to obtain the filtering algorithm corresponding to the current frequency conversion operating condition; and generate real-time dynamic filtering parameters based on the filtering algorithm and the noise center frequency drift.

[0159] The stopband center and stopband width of the dynamic filtering parameters are adaptively tracked and updated according to the noise center frequency drift and the spectral coverage bandwidth.

[0160] The frequency converter interference feature library is pre-built offline and stored in a local database. The frequency converter interference feature library stores several standard operating condition modes, as well as adaptive filtering function templates associated with each of the standard operating condition modes.

[0161] In some preferred embodiments, such as Figure 3 As shown, step S103, which involves performing similarity matching between the frequency band feature parameters and a pre-built frequency conversion interference feature library to obtain the filtering algorithm corresponding to the current frequency conversion operating condition, includes steps S301 to S304; each step is detailed below:

[0162] Step S301: The frequency band characteristic parameters are characterized as a multi-dimensional operating condition feature vector of the current noise distribution characteristics;

[0163] Step S302: Project the multidimensional working condition feature vector onto a preset high-dimensional feature space to obtain the projection result; and calculate the weighted Mahalanobis distance between the projection result and the center of each standard working condition mode.

[0164] Step S303: Select the standard working condition mode with the smallest weighted Mahalanobis distance as the target matching mode, and call the adaptive filtering function template corresponding to the target matching mode;

[0165] Step S304: Based on the real-time rate of change of the noise center frequency drift, determine the filter order and topology in the adaptive filter function template, and then generate a filter algorithm for the current frequency conversion condition based on the order and topology.

[0166] In this preferred embodiment, the standard operating condition mode is obtained through offline analysis and summarization of dynamic noise signals collected from a large number of typical frequency converters under different operating conditions (including different loads, different speeds, different frequency converter models, and different power grid topology conditions). During the offline construction phase, for each typical frequency converter operating condition, the corresponding noise center frequency drift, harmonic energy peak value, and spectral coverage bandwidth are extracted to form the original feature samples for that operating condition. A large number of feature samples under the same operating condition are statistically clustered, with the cluster centers used as representative features of the standard operating condition mode. Simultaneously, the feature covariance matrix of this type of sample is calculated for subsequent weighted Mahalanobis distance calculation. Each standard operating condition mode is stored in the feature library in the form of its cluster center coordinates and corresponding covariance matrix.

[0167] The adaptive filtering function template is designed independently for each standard operating mode, describing the basic form of the filter transfer function applicable to this type of frequency conversion operating condition. This includes the filter type (notch filter, band-stop filter, etc.), the parameterized expression framework of the transfer function, and alternative order and topology schemes corresponding to different noise dynamic rates. The adaptive filtering function template is not a complete filter with fixed parameters, but rather a function framework that can be driven by real-time noise characteristic parameters and instantiated with parameters. Its final parameters are determined during the online matching stage based on real-time frequency band characteristic parameters.

[0168] During the online operation phase, the frequency band feature parameters extracted in step S102 are first characterized as a multi-dimensional operating condition feature vector reflecting the current noise distribution characteristics. Specifically, step S301 uses the noise center frequency drift, harmonic energy peak value, and spectral coverage bandwidth as components to construct a three-dimensional original feature vector, which comprehensively describes the integrated characteristics of the dynamic noise signal in the current working cycle in terms of frequency travel range, energy intensity, and bandwidth.

[0169] Because feature components of different dimensions differ in their units, numerical ranges, and ability to distinguish operating conditions, directly measuring distance in the original three-dimensional space can easily lead to a single component dominating the matching result due to its large numerical range, thus reducing recognition accuracy. Therefore, the original three-dimensional feature vector is projected onto a preset high-dimensional feature space to obtain the projection result. This projection is achieved through kernel function mapping, which maps the linearly inseparable operating condition distribution in the original feature space to the high-dimensional feature space, resulting in clearer inter-class boundaries and improving the distinguishability between different standard operating condition modes. The kernel function can be a radial basis function (RBF) kernel, which uses the negative exponent of the Euclidean distance between the original feature vector and the cluster centers of each standard operating condition mode as the mapping weight. It should be noted that the type of kernel function is not limited to the RBF kernel; those skilled in the art can choose a multinomial kernel function or other suitable kernel function forms based on the distribution characteristics of each standard operating condition mode in the feature library.

[0170] After projection, the weighted Mahalanobis distance between the projection result and the modal centers of each standard operating condition in the frequency conversion interference feature library is calculated. The weighted Mahalanobis distance introduces an additional weight matrix on top of the standard Mahalanobis distance, assigning different matching weights to different feature dimensions to reflect the relative importance of each feature component in operating condition differentiation. A higher weight is assigned to the noise center frequency drift dimension because this component directly determines the tracking range of the stopband center in subsequent dynamic filtering parameters, and has the most significant impact on the filtering effect. Compared with the standard Euclidean distance, the weighted Mahalanobis distance can simultaneously eliminate the influence of inconsistent dimensions among feature dimensions, take into account the statistical dispersion of features in each dimension, and reflect the differences in the contribution of each dimension to operating condition identification, thus achieving accurate operating condition matching even in cases of uneven feature distribution.

[0171] The standard operating condition mode with the smallest weighted Mahalanobis distance is selected as the target matching mode, and the corresponding adaptive filtering function template is called. The target matching mode is the historical operating condition category that is most similar to the current frequency conversion operating condition in the noise feature space. Its corresponding adaptive filtering function template records a verified and effective filter structure framework, providing a basis for the instantiation of subsequent filtering algorithms.

[0172] Subsequently, based on the real-time rate of change of the noise center frequency drift, the filter order and topology are determined in the adaptive filtering function template, thereby generating a filtering algorithm for the current frequency conversion condition based on the order and topology. The real-time rate of change of the noise center frequency drift is obtained by comparing the ratio of the difference in noise center frequency drift within adjacent operating cycles to the time interval, reflecting the speed of dynamic changes in the noise spectrum under the current frequency conversion condition. When the real-time rate of change is low, it indicates that the current noise spectrum changes slowly, and the filter does not need to have fast tracking capability. In this case, a lower-order topology is selected in the adaptive filtering function template to reduce the computational complexity of the filter and reduce group delay. When the real-time rate of change is high, it indicates that the current noise spectrum changes drastically, and the filter needs to have stronger frequency selectivity and faster parameter response capability. In this case, a higher-order topology is selected to obtain a steeper stopband edge and more accurate notch filtering capability, ensuring that dynamic noise signals can still be effectively suppressed under the condition of rapid noise spectrum drift. The adaptive filtering function template pre-sets alternative schemes for order and topology corresponding to different rate of change intervals. The system automatically selects the corresponding scheme based on the interval in which the real-time rate of change falls, completes the final instantiation of the filtering algorithm, and generates a complete filtering algorithm for the current frequency conversion operating condition.

[0173] Based on the filtering algorithm generated above and the noise center frequency drift, real-time dynamic filtering parameters are generated. The core of these dynamic filtering parameters consists of two terms: stopband center and stopband width, which correspond to the noise frequency location and frequency range that the filter needs to suppress, respectively.

[0174] The stopband center is determined in real time based on the instantaneous frequency of the energy ridge trajectory at the current moment. The current frequency value of the energy ridge trajectory is used as the set value for the stopband center, enabling real-time tracking of the center frequency of dynamically drifting noise. Specifically, after completing the time-frequency analysis of each sampling point, the system extracts the instantaneous frequency corresponding to the energy ridge trajectory at that moment and updates the stopband center to that frequency position, ensuring that the notch center of the filter is always aligned with the frequency point where the noise energy is most concentrated. The stopband width is determined jointly based on the spectral coverage bandwidth and the noise center frequency drift: the current spectral coverage bandwidth is used as the base stopband width, and a margin proportional to the noise center frequency drift is added on top to cover the stopband boundary displacement caused by frequency drift. This ensures that the stopband can still completely enclose the noise frequency band when the noise center frequency changes rapidly, preventing the noise spectrum from exceeding the stopband boundary and remaining in the filter output.

[0175] The stopband center and stopband width of the dynamic filtering parameters are adaptively updated based on the noise center frequency drift and the spectral coverage bandwidth. The update frequency is synchronized with the sampling rate of the time-frequency analysis, enabling point-by-point real-time tracking of the dynamically drifting noise frequency band. Compared with fixed-parameter filters, the adaptive tracking and update mechanism allows the filter to maintain precise alignment with the noise frequency band under complex frequency conversion conditions where the noise spectrum undergoes irregular dynamic drift. This effectively suppresses dynamic noise signals while avoiding noise residue due to insufficient stopband coverage or weakening of the amplitude and phase integrity of the probe pulse signal due to excessively wide stopband.

[0176] Step S104: Based on the dynamic filtering parameters and filtering algorithm, the hybrid response signal is filtered to remove the dynamic noise signal that overlaps with the frequency band of the detection pulse signal, thereby obtaining a reconstructed signal for the insulation leakage current signal.

[0177] In some application scenarios, the frequency band of the dynamic noise signal partially overlaps with the frequency band of the detection pulse signal. In this overlapping area, the noise component and the effective insulation leakage current signal are deeply mixed. Simple amplitude domain filtering will inevitably weaken the effective response of the detection pulse signal simultaneously, resulting in distortion of the amplitude and phase of the leakage current signal.

[0178] Therefore, this embodiment divides the filtering process into multiple sub-steps and executes them sequentially: constructing a time-varying notch filter, implementing bidirectional zero-phase-shift filtering, establishing a dynamic orthogonal reference coordinate system and completing signal projection decomposition, and performing vector synthesis and amplitude gain compensation for overlapping frequency bands, ultimately obtaining a reconstructed signal with both amplitude and phase fully preserved. Specifically:

[0179] Based on the dynamic filtering parameters and filtering algorithm, a time-varying notch filter is constructed.

[0180] The hybrid response signal is input into the time-varying notch filter to obtain a preliminary filtered signal;

[0181] The preliminary filtered signals are rearranged in reverse time order to obtain the rearrangement result;

[0182] The rearrangement result is input into the time-varying notch filter for secondary filtering to obtain a zero-phase-shift denoised signal.

[0183] Using the detection pulse signal as a reference signal, a dynamic orthogonal reference coordinate system composed of in-phase axes and quadrature axes is constructed; the zero-phase-shift denoised signal is projected onto the dynamic orthogonal reference coordinate system to obtain in-phase components and quadrature components with the same frequency as the detection pulse signal;

[0184] For the overlapping regions of the frequency bands, vector synthesis and amplitude gain compensation calculations are performed using the in-phase and quadrature components to obtain the reconstructed signal.

[0185] In this embodiment, the time-varying notch filter can use the filter topology determined in step S103 as the basic framework, and the real-time updated stopband center and stopband width as the core parameters. At each sampling moment, the current dynamic filtering parameters are instantiated as filter coefficients, so that the notch center of the filter is always aligned with the instantaneous frequency of the dynamic noise signal at the current moment, and the stopband width always covers the actual extension range of the noise frequency band at the current moment.

[0186] Compared with fixed-parameter notch filters, the filter coefficients of the time-varying notch filter are updated point by point over time, rather than remaining constant throughout the filtering process. Therefore, it can track the dynamically drifting noise center frequency in real time and avoid noise residue caused by noise frequency wandering beyond the fixed stopband range.

[0187] In this embodiment, the steepness of the stopband edge of the time-varying notch filter can be set according to the filter order determined in step S304. The higher the order, the steeper the stopband edge, and the more accurate the protection of the detection pulse signal components near the noise frequency band boundary. It should be noted that there is a trade-off between the filter order and the computational complexity. Those skilled in the art can determine the order comprehensively based on the computing power constraints of real-time processing and the filtering accuracy requirements.

[0188] The initial filtered signal has completed the first suppression of the main energy of the dynamic noise signal. However, since any causal filter introduces a frequency-related phase delay when performing frequency-selective processing on the signal, the insulation leakage current signal component in the initial filtered signal has a nonlinear phase shift relative to the original mixed response signal. This phase shift will directly affect the accuracy of the insulation resistance value calculated based on the reconstructed signal phase in step S105, and therefore can be eliminated.

[0189] Therefore, the initial filtered signal is rearranged in reverse time order to obtain a rearranged result, that is, the time sequence of the initial filtered signal is flipped from forward to reverse. The rearranged result is then input into the time-varying notch filter for secondary filtering to obtain a zero-phase-shift denoised signal. The above-described bidirectional filtering mechanism of forward filtering—time flipping—reverse filtering—re-flipping and restoration ensures that the phase responses introduced by the two filtering processes are opposite in direction and equal in magnitude in the time domain. Thus, after the signal undergoes complete bidirectional filtering, the phase responses precisely cancel each other out. The final zero-phase-shift denoised signal is equivalent to a single filtering in frequency selectivity, while its phase response is zero relative to the original signal. That is, while removing noise components, the original phase information of the insulation leakage current signal is completely preserved.

[0190] It should be noted that the bidirectional filtering process is completed in a single sampling window using an offline batch processing method. That is, the preliminary filtered signal data of a complete working cycle is first cached, and then the time flip and reverse filtering operations are performed. Therefore, it does not affect the real-time detection capability of the system. The length of the sampling window can be consistent with the working cycle of the time-frequency analysis in step S102.

[0191] Preferably, the construction process of the dynamic orthogonal reference coordinate system is as follows: Using the real-time instantaneous phase of the probe pulse signal as a reference, an in-phase axis is defined along the phase direction of the probe pulse signal, and an orthogonal axis is defined along a direction orthogonal to it, forming a two-dimensional rotating reference coordinate system that rotates in real time with the phase of the probe pulse signal. The rotation frequency of the reference coordinate system is strictly synchronized with the injection frequency of the probe pulse signal, ensuring that the coordinate system is always aligned with the phase state of the probe pulse signal. Since the injection frequency of the probe pulse signal has been determined and is known in real time through adaptive adjustment in the above steps, the rotation state of the dynamic orthogonal reference coordinate system can be accurately calculated at each sampling moment, without the need for additional phase-locking hardware.

[0192] The zero-phase-shift denoised signal is projected onto the dynamic orthogonal reference coordinate system. Specifically, correlation operations are performed between the zero-phase-shift denoised signal and unit reference signals in the in-phase axis direction and the orthogonal axis direction, respectively. The components in phase with the in-phase axis direction and in phase with the orthogonal axis direction at the frequency of the probe pulse signal are extracted, resulting in in-phase and orthogonal components. The in-phase component reflects the energy component in the zero-phase-shift denoised signal that is in phase and at the same frequency as the probe pulse signal, while the orthogonal component reflects the energy component that is in phase with the probe pulse signal but at the same frequency but with a 90-degree phase difference. Together, they completely characterize the vector state of the zero-phase-shift denoised signal at the frequency of the probe pulse signal, i.e., all information regarding amplitude and phase.

[0193] The purpose of this step is to further separate the effective insulation leakage current response component, which is in the same frequency as the probe pulse signal, from the background by projecting the zero-phase-shift denoised signal mixed with residual noise components into a coordinate system that is strictly synchronized with the probe pulse signal. This provides an accurate amplitude and phase reference for subsequent vector compensation in the overlapping frequency band region.

[0194] For the overlapping regions of the frequency bands, vector synthesis and amplitude gain compensation calculations are performed using the in-phase and quadrature components to obtain the reconstructed signal.

[0195] The overlapping frequency band refers to the frequency range where the frequency band of the dynamic noise signal overlaps with the frequency band of the probe pulse signal. Within this region, while suppressing noise components, the time-varying notch filter inevitably attenuates the effective response component of the probe pulse signal to a certain extent, resulting in the amplitude of the zero-phase-shift denoising signal in this frequency band being lower than the actual leakage current response amplitude of the corresponding frequency band in the mixed response signal. If this amplitude attenuation is not compensated for, directly calculating the insulation resistance value using the amplitude of the zero-phase-shift denoising signal will introduce a systematic underestimation error.

[0196] Therefore, based on the in-phase and quadrature components, vector synthesis is performed: the in-phase and quadrature components are considered as projections of the same vector in two orthogonal directions. By taking the square root of the sum of their squares, the actual response vector amplitude of the zero-phase-shift denoised signal at the frequency of the probe pulse signal is obtained. This vector synthesis operation is essentially equivalent to recovering the complete complex envelope amplitude from the in-phase and quadrature components. The result is unaffected by the absolute phase of the signal, depending only on the sum of the energies of the two quadrature components, thus exhibiting inherent robustness to phase estimation errors.

[0197] Based on vector synthesis, amplitude gain compensation is further performed: According to the real-time amplitude-frequency response of the time-varying notch filter at the probe pulse signal frequency, the amplitude attenuation coefficient applied by the filter to the effective response component of the probe pulse signal at that moment is calculated; the reciprocal of the attenuation coefficient is used as a gain compensation factor to multiply the amplitude obtained from vector synthesis, quantitatively restoring the amplitude loss caused by frequency band overlap. The real-time amplitude-frequency response of the time-varying notch filter at the probe pulse signal frequency can be directly calculated based on the filter coefficients at the current moment, without the need for additional calibration.

[0198] After the above vector synthesis and amplitude gain compensation processing, the amplitude distortion of the obtained signal in the frequency band overlap region is quantitatively recovered, and the phase information is preserved by bidirectional zero phase shift filtering. Thus, while preserving the complete amplitude and phase information of the insulation leakage current signal, the dynamic noise signal is effectively eliminated, and the reconstructed signal is obtained.

[0199] The reconstructed signal is the insulation leakage current response signal after complete noise suppression and amplitude and phase recovery processing. Its amplitude and phase accurately reflect the real insulation leakage characteristics of the target power equipment under the current operating conditions. It will be used as the input of step S105 to calculate the real-time insulation resistance value of the target power equipment.

[0200] Step S105: Based on the amplitude and phase of the reconstructed signal and the voltage parameters of the detection pulse signal, calculate the real-time insulation resistance value of the power supply circuit of the target power equipment under operating conditions to achieve insulation detection.

[0201] In this step, the voltage parameters include the voltage change rate and amplitude.

[0202] Although the reconstructed signal has undergone dynamic noise suppression in step S104, it still contains two types of current components with different properties. The first type is the purely resistive leakage current on the insulation path of the target power equipment, whose amplitude is inversely proportional to the insulation resistance, and is an effective source of information for calculating the insulation resistance value. The second type is the capacitive charging and discharging current generated by the distributed capacitance of the system to ground in the power supply circuit under the excitation of the probe pulse signal. This current is independent of the insulation state. If it is not removed and the insulation resistance value is directly calculated based on the total current of the reconstructed signal, a systematic error caused by the charging and discharging effect of the distributed capacitance will be introduced, which is particularly significant in power supply circuits with large distributed capacitance. Therefore, this step uses a combination of equivalent circuit modeling, nonlinear feature sampling, and least squares iterative solution to quantitatively separate and remove the capacitive current component from the reconstructed signal, and finally extracts the purely resistive leakage current component for accurate calculation of the insulation resistance value. This step is divided into the following four sub-steps, which are executed sequentially.

[0203] Specifically, such as Figure 4As shown, step S105, which calculates the real-time insulation resistance value of the power supply circuit of the target power equipment under operating conditions based on the amplitude and phase of the reconstructed signal and the voltage parameters of the probe pulse signal, includes steps S401 to S404; each step is detailed below:

[0204] Step S401: Based on the level flip trigger edge of the probe pulse signal, select multiple nonlinear feature sampling points within a single pulse period of the reconstructed signal; wherein, the nonlinear feature sampling points include transient sampling points during the capacitor charging transition process and static sampling points under the charging saturation trend.

[0205] Step S402: Establish an equivalent parallel circuit model that includes the insulation resistance to be measured and the system's distributed capacitance to ground. Combine the voltage change rate of the probe pulse signal and the instantaneous current values ​​of the multiple nonlinear characteristic sampling points to construct a set of linear equations about the insulation resistance to be measured and the system's distributed capacitance to ground.

[0206] Step S403: Solve the linear equation system using the least squares iterative algorithm to calculate the capacitive current component in the reconstructed signal corresponding to the system's distributed capacitance to ground, and remove the capacitive current component from the reconstructed signal to extract the purely resistive leakage current component.

[0207] Step S404: Calculate the insulation resistance value based on the steady-state amplitude of the pure resistive leakage current component and the amplitude of the detection pulse signal.

[0208] In this preferred embodiment, step S401 selects multiple nonlinear feature sampling points within a single pulse period of the reconstructed signal based on the level-flipping trigger edge of the probe pulse signal, specifically as follows:

[0209] The probe pulse signal undergoes a level transition process within each pulse cycle, changing from a low level to a high level (or from a high level to a low level). At the instant the level transition occurs, the system's distributed capacitance to ground in the power supply circuit begins to charge (or discharge) in response to the voltage change, generating a capacitive current that decays exponentially with time. Simultaneously, the resistive leakage current in the insulation path adjusts synchronously with the change in the probe pulse signal voltage, tending towards a new steady-state value. The superposition of these two types of currents causes the reconstructed signal to exhibit a significant nonlinear transition characteristic after the level transition; that is, the current amplitude changes rapidly with time during the brief transition period, and then gradually approaches stability.

[0210] Using the level-flipping trigger edge as a time reference, sampling points are selected within the following two characteristic intervals during a single pulse period of the reconstructed signal:

[0211] One type is the transient sampling point, located during the capacitor charging transition after a level flip, specifically the nonlinear decay phase where the amplitude of the reconstructed signal current changes rapidly over time. The transient sampling points are positioned within a time window following the level flip trigger edge but before the capacitive current has fully decayed. During this phase, the capacitive current component dominates, and the reconstructed signal is most sensitive to changes in the distributed capacitance parameters. Therefore, the transient sampling points are primarily used to constrain the value of the system's distributed capacitance to ground in the equivalent circuit model.

[0212] The second type is the static sampling point, located under the charging saturation trend, that is, within the time window when the capacitive current has basically decayed and the reconstructed signal approaches steady state. At this stage, the capacitive current component can be ignored, and the reconstructed signal is mainly composed of pure resistive leakage current. The current amplitude at this stage is most sensitive to changes in insulation resistance. The static sampling point is mainly used to constrain the value of the insulation resistance to be measured in the equivalent circuit model.

[0213] By selecting multiple sampling points within the two aforementioned characteristic intervals, the resulting multiple nonlinear characteristic sampling points cover the complete dynamic process of the reconstructed signal from the transient state to the steady state. This ensures that the parameter identification of the subsequent equivalent circuit model is simultaneously constrained by data from both types of characteristic intervals, effectively avoiding the problems of using only steady-state data leading to the inability to identify distributed capacitance parameters, or using only transient data leading to deviations in the estimation of steady-state insulation resistance. In this embodiment, the total number of the multiple nonlinear characteristic sampling points can be determined based on the sampling rate within a single pulse period and the duration of the transient process. The ratio of transient sampling points to static sampling points can be adaptively adjusted according to the size of the distributed capacitance. It should be noted that the specific number and location distribution of the sampling points are not limited to a fixed scheme, and those skilled in the art can make adaptive designs based on the actual circuit parameter range and the accuracy requirements of the numerical solution.

[0214] Furthermore, an equivalent parallel circuit model including the insulation resistance to be measured and the system's distributed capacitance to ground is established. By combining the voltage change rate of the probe pulse signal and the instantaneous current values ​​of the multiple nonlinear characteristic sampling points, a system of linear equations concerning the insulation resistance to be measured and the system's distributed capacitance to ground is constructed.

[0215] The equivalent parallel circuit model treats the insulation path of the target power equipment as the insulation resistance to be measured, and the parasitic capacitance to ground of the power supply circuit as the distributed capacitance to ground of the system. These two components are connected in parallel and then subjected to voltage excitation from the probe pulse signal. In this equivalent parallel circuit model, the total current of the reconstructed signal is composed of the superposition of the resistive leakage current component flowing through the insulation resistance to be measured and the capacitive current component flowing through the distributed capacitance to ground of the system. The resistive leakage current component is proportional to the ratio of the current voltage to the insulation resistance to be measured at any given time; the capacitive current component is proportional to the product of the distributed capacitance to ground of the system and the rate of change of the voltage at the current time, i.e., the differential response of the distributed capacitance to the voltage.

[0216] Based on the above equivalent parallel circuit model, at each nonlinear characteristic sampling point, an equation can be established regarding the total current, resistive leakage current component, and capacitive current component: The instantaneous current value of the reconstructed signal at that moment (a known quantity, directly obtained from sampling) is used as the left-hand side of the equation. The insulation resistance to be measured and the system's distributed capacitance to ground are the unknown parameters. Combined with the instantaneous voltage value and voltage change rate of the probe pulse signal at that moment (both known quantities), the resistive leakage current component and the capacitive current component are expressed as linear functions of the two unknown parameters, thus forming a linear combination of the two unknown parameters on the right-hand side. This modeling process is repeated for all nonlinear characteristic sampling points. Since the number of sampling points exceeds the number of unknown parameters (the insulation resistance to be measured and the system's distributed capacitance to ground are two unknowns), an overdetermined linear equation system with more equations than unknowns is ultimately obtained. The redundant equations of this overdetermined linear equation system come from independent sampling constraints at different times, providing a statistical averaging basis for subsequent least-squares iterative solutions and effectively suppressing the influence of single-sampling-point measurement noise on the parameter identification results.

[0217] Then, the linear equations are solved using the least squares iterative algorithm to calculate the capacitive current component in the reconstructed signal corresponding to the system's distributed capacitance to ground, and the capacitive current component is removed from the reconstructed signal to extract the purely resistive leakage current component.

[0218] The least squares iterative algorithm aims to minimize the sum of squared residuals of the equation system at all nonlinear feature sampling points. It searches for the optimal parameter estimates that minimize the overall fitting error of the equation system within the parameter space of the insulation resistance to be measured and the system's distributed capacitance to ground. Specifically, starting from the initial parameter estimates, the algorithm calculates the residuals of the equation system at each sampling point based on the current parameter estimates in each iteration, and updates the parameter estimates along the direction of the fastest decrease in the sum of squared residuals. After several iterations, the parameter estimates converge to a stable solution. The initial parameter estimates can use the solution result of the previous pulse cycle as the initial value for a warm start, to accelerate the convergence speed and avoid getting trapped in local extrema. When there are no historical solution results during the initial power-on phase of the system, empirically preset values ​​based on the rated parameter range of the power supply circuit can be used as the initial value for a cold start. The convergence criterion of the least squares iterative algorithm can be set as follows: the relative change in parameter estimates between two adjacent iterations is lower than a preset accuracy threshold. After reaching the convergence criterion, the optimal parameter estimates are output, namely the estimated values ​​of the insulation resistance to be measured and the system's distributed capacitance to ground.

[0219] Based on the estimated system-to-ground capacitance obtained from the above solution, and combined with the voltage change rate of the probe pulse signal at each moment, the theoretical value of the capacitive current component of the reconstructed signal at each moment is recalculated. That is, the product of the estimated system-to-ground capacitance and the voltage change rate at the corresponding moment is used as the quantitative estimate of the capacitive current component at that moment. Subsequently, the capacitive current component is removed point by point from the total current of the reconstructed signal. That is, at each moment, the instantaneous value of the total current of the reconstructed signal is subtracted from the calculated value of the capacitive current component at the corresponding moment, resulting in the difference signal after capacitive current removal. This difference signal is the pure resistive leakage current component.

[0220] The purely resistive leakage current component contains only the effective leakage current response flowing through the insulation resistance under test. Its time-domain waveform, in the steady-state phase of the probe pulse signal (i.e., the interval where the static sampling point is located), approaches a stable DC or low-frequency component that is in phase with the probe pulse signal. Its amplitude is inversely proportional to the insulation resistance under test and is no longer affected by the charging and discharging effect of distributed capacitance, thus providing a reliable basis for the accurate calculation of the subsequent insulation resistance value.

[0221] Then, based on the steady-state amplitude of the purely resistive leakage current component and the amplitude of the detection pulse signal, the insulation resistance is calculated, specifically:

[0222] In the purely resistive leakage current component, the average current value of each sampling point within the steady-state interval of the static sampling point is taken as the steady-state amplitude of the purely resistive leakage current component. This averaging process further suppresses the influence of residual measurement noise on the steady-state amplitude estimation. According to Ohm's law, the real-time insulation resistance value of the target power equipment under the current operating conditions is obtained by dividing the steady-state voltage amplitude of the probe pulse signal by the steady-state amplitude of the purely resistive leakage current component.

[0223] The real-time insulation resistance value is calculated and updated once per pulse cycle, with the update frequency synchronized with the injection frequency of the detection pulse signal, thus achieving real-time continuous monitoring of the insulation status. In this embodiment, the detection range of the insulation resistance value can be 0 to 20 MΩ, and the detection accuracy can be ±5%. It should be noted that the above range and accuracy indicators are not limited to the above values, and those skilled in the art can adjust them according to the rated insulation level of the target power equipment and actual monitoring requirements.

[0224] The real-time insulation resistance value, as the final output of step S105, reflects the insulation health status of the target power equipment in real time, providing a quantitative basis for subsequent alarm threshold comparison and fault location. On the other hand, it serves as a feedback quantity sent back to the adaptive adjustment link of the detection pulse signal in step S101, and together with the synchronous identification result of the system's distributed capacitance to ground, it is used to drive the dynamic adjustment of the frequency and amplitude of the detection pulse signal, thereby realizing closed-loop optimization of the entire insulation detection process.

[0225] Preferably, after calculating the real-time insulation resistance value of the power supply circuit of the target power equipment in its operating state, the method further includes:

[0226] The real-time insulation resistance value is compared with the preset first warning threshold and second alarm threshold respectively;

[0227] When the real-time insulation resistance value is lower than the first warning threshold or the second alarm threshold, the fault location is performed based on the polarity or phase characteristics of the reconstructed signal.

[0228] If the power supply circuit is a DC system, the insulation fault is determined and displayed as occurring at the positive or negative terminal based on the current flow direction of the reconstructed signal.

[0229] If the power supply circuit is a three-phase AC system, the phase correlation between the reconstructed signal and the voltage of each phase is analyzed to determine and display the phase line where the insulation fault occurred.

[0230] It should be noted that the first warning threshold and the second alarm threshold can be named ALARM1 and AMARM2, respectively. These two thresholds represent different stages of insulation degradation and have different technical significance.

[0231] The first warning threshold can be set at a relatively high resistance value, which falls under the category of "preventive protection". When the insulation resistance value drops to this value, it indicates that the insulation performance has begun to decline significantly, but the system is still within the safe operating range. At this time, the system reminds the operation and maintenance personnel to pay attention so as to arrange planned maintenance and prevent the fault from expanding further.

[0232] The second alarm threshold can be set at a lower critical resistance value, which falls under the category of "safety assurance". When the insulation resistance value drops below this value, it indicates that the system has a serious risk of grounding fault or a single-phase grounding has occurred, which may endanger personal safety or burn out equipment. At this time, the system usually triggers audible and visual alarms, records the fault point, and may even trigger the circuit breaker.

[0233] Furthermore, in some application examples, the insulation detection method for power equipment can be applied to industrial control computers. In some application scenarios, an artificial intelligence-based insulation monitoring system can be configured to achieve integrated real-time monitoring, accurate diagnosis, intelligent early warning, and remote monitoring.

[0234] In some embodiments, the insulation monitoring system may include a cabinet, an insulation monitor, an industrial computer, an offline / online module, a system control module, and a communication module, and each module is fixed to the cabinet via metal rails.

[0235] The insulation detection method for power equipment described in this embodiment can be applied to an industrial control computer. The industrial control computer can control the insulation monitoring instrument and use the adaptive pulse injection method to realize online insulation detection of power equipment or electrical lines. The industrial control computer can model and analyze the data collected by the insulation monitoring instrument and perform fault early warning and real-time display through the built-in expert system.

[0236] The insulation monitor can control offline / online modules using multiple 380VAC / 220VAC intermediate relays (or multiple normally open contacts of contactors) via the standby command function.

[0237] Specifically, by utilizing the logical relationship between the main and auxiliary contacts of multiple 380VAC intermediate relays (multiple normally open contacts of contactors), and through the MODBUS enable command standby function of the insulation monitor, offline-to-online insulation monitoring is carried out on the power circuits and equipment of the TN system.

[0238] The power grid supply circuit and the phase lines L1, L2, and L3 of the electrical equipment are connected to the insulation monitoring instrument in sequence through miniature circuit breakers and contactors to continuously monitor the insulation resistance value of the entire system.

[0239] The system control module can be activated by the passive dry contacts (two sets of contacts for pre-alarm and main alarm) of the insulation monitor, and controlled by multiple intermediate relays to control the power supply, insulation early warning, insulation alarm, insulation abnormality, equipment failure, insulation monitoring in progress, insulation normal and other function indicator lights of the insulation monitoring system.

[0240] The communication module uses an RS485 serial server to collect and upload monitoring and control data from the insulation monitor via the ModBus RTU protocol, enabling human-machine interaction.

[0241] In some implementations, the insulation resistance of the system under test is continuously monitored. When the value is lower than the preset insulation resistance response value ALM1 or ALM2, an alarm is triggered, the alarm indicator light illuminates, the measured value is displayed on the LCD screen, and two sets of independent C / O programmable alarm relays are activated. Through their passive dry contact alarm output signals, they control the operation of intermediate relays and miniature circuit breakers, realizing functions such as automatic power control, insulation warning, insulation alarm, insulation abnormality, equipment failure, insulation monitoring in progress, and insulation normal. Each function is displayed through indicator lights.

[0242] The industrial control computer can realize human-machine interaction through the communication module, and use the ModBus RTU protocol to collect relevant data from the insulation monitor. Facing different application scenarios, it can automatically, in real time, and quickly respond to the status of the insulation monitor and realize alarm and fault reminder functions. It runs in guardian mode on the terminal, realizes real-time text push of alarms and faults, and has remote operation, event query, data acquisition and data display functions.

[0243] A database linking insulation resistance with its own insulation performance and external environmental interference is constructed, with a storage space of more than 500G. Real-time insulation monitoring data is stored in daily, monthly, weekly, and annual reports; it is displayed through a trend chart, and the file names are in date format and the file format is Excel.

[0244] For example, the cabinet can adopt a front and rear door structure, with dimensions of 2260×800×800mm (corresponding to height, width, and depth respectively). The front of the cabinet uses a protective door with plexiglass, and the door hinge is on the left side of the front of the screen. The back has a steel plate protective door; the top and side panels are 1.5mm thick, and the cabinet columns are 2.0mm thick. A tray (for placing the industrial control computer host) is included, along with a drawer for the computer keyboard and mouse; the gap between the door and the screen has a cross-section of not less than 4mm. 2 The system features reliable multi-strand soft copper wire connections. An 8-port PDU power strip and fan are installed inside the panel. Connection test brackets (independent test brackets for installing insulation testers) are provided to ensure adaptability to various operating conditions.

[0245] The selected insulation monitoring instrument has a range of 0-20MΩ and an accuracy of ±5%; the system control module uses contactors, relays, and miniature circuit breakers, and the alarm module includes a red LED; the industrial control computer uses the Linux system Ubuntu; the memory is greater than 8G; and the storage space is greater than 500G.

[0246] During assembly, the first step is to fix the guide rails inside the front and back frames of the cabinet. Then, install the insulation monitor, miniature circuit breaker, fuse, contactor, relay, and communication module on the guide rails in sequence. Connect the power supply (AC 220V), the monitored system (phase line L, neutral line N, ground line PE), and the external monitoring equipment (RS485 interface) to the bottom wiring terminals on the back of the cabinet, respectively. Then, install the function indicator lights and industrial control computer on the front of the cabinet. The protective doors on the front and back are normally closed.

[0247] After the equipment is powered on, for TN systems, the system control module first checks whether the power supply equipment is operating normally. If it is operating normally, the online insulation monitor is in standby mode; if the power supply equipment is deactivated, the online insulation monitor is activated. For IT systems, the system control module directly enters online monitoring mode.

[0248] For TN systems, the insulation tester monitors the insulation resistance of the power supply circuit and power supply equipment (threshold ≤ 300kΩ triggers an alarm).

[0249] For IT systems, insulation testers monitor the insulation resistance of power supply circuits and equipment (according to GB / T16895.22-2022 Low-voltage electrical installations Part 5-53: Selection and installation of electrical equipment for safety protection, isolation, switching, control and monitoring, the typical setting value corresponding to the rated system voltage is 100Ω / V (alarm value is 300Ω / V)).

[0250] Monitoring data is uploaded to the monitoring center via the communication module. Historical data from the insulation monitoring instrument is automatically stored (retained for one year) and can be queried via buttons. The platform management software has built a database linking insulation resistance with factors such as its own insulation performance and external environmental interference. Real-time insulation monitoring data is stored in daily, monthly, weekly, and annual reports; it is displayed through a trend chart. Files are named in date format and are in Excel format.

[0251] Accordingly, such as Figure 5 As shown, this invention application also provides an insulation detection system 500 for power equipment, including a data acquisition module 501, an extraction module 502, a filter parameter generation module 503, a reconstruction module 504, and an insulation detection module 505; wherein,

[0252] The acquisition module 501 is used to inject a detection pulse signal into the power supply circuit of the target power equipment and acquire the mixed response signal of the power supply circuit; wherein, the mixed response signal includes an insulation leakage current signal excited by the detection pulse signal and a dynamic noise signal generated by the operation of the frequency converter.

[0253] The extraction module 502 is used to perform joint time-domain and frequency-domain analysis on the hybrid response signal, and extract the frequency band characteristic parameters of the dynamic noise signal within the current working cycle based on the joint analysis results; wherein, the frequency band characteristic parameters include noise center frequency drift, harmonic energy peak value, and spectral coverage bandwidth;

[0254] The filter parameter generation module 503 is used to perform similarity matching between the frequency band feature parameters and a pre-built frequency conversion interference feature library to obtain the filter algorithm corresponding to the current frequency conversion operating condition; and to generate real-time dynamic filter parameters based on the filter algorithm and the noise center frequency drift; wherein the stopband center and stopband width of the dynamic filter parameters are adaptively tracked and updated with the noise center frequency drift and the spectrum coverage bandwidth.

[0255] The reconstruction module 504 is used to filter the hybrid response signal based on the dynamic filtering parameters and filtering algorithm to filter out the dynamic noise signal that overlaps with the frequency band of the detection pulse signal, and obtain a reconstructed signal for the insulation leakage current signal.

[0256] The insulation detection module 505 is used to calculate the real-time insulation resistance value of the power supply circuit of the target power equipment in the operating state based on the amplitude and phase of the reconstructed signal and the voltage parameters of the detection pulse signal, so as to realize insulation detection.

[0257] As a preferred embodiment, the extraction module 502 performs joint time-domain and frequency-domain analysis on the hybrid response signal, and extracts the frequency band characteristic parameters of the dynamic noise signal within the current working cycle based on the joint analysis results, including:

[0258] The extraction module 502 applies synchronous compressed wavelet transform to the hybrid response signal, compressing and rearranging the signal energy to instantaneous frequency points in the frequency domain to obtain a time-frequency energy distribution matrix.

[0259] Based on the time-frequency energy distribution matrix, the energy ridge trajectory of the dynamic noise signal on the time-frequency plane is extracted using an energy gradient-based ridge extraction algorithm.

[0260] The difference between the instantaneous maximum and minimum frequency values ​​of the energy ridge trajectory within the current working cycle is calculated as the noise center frequency drift of the dynamic noise signal; the maximum energy amplitude point on the energy ridge trajectory is extracted to obtain the harmonic energy peak value.

[0261] Using the energy ridge trajectory as the central axis, perform layer-by-layer energy integration along both sides of the frequency axis in the time-frequency energy distribution matrix to obtain the cumulative integrated energy;

[0262] When the accumulated integrated energy reaches a preset proportion threshold of the total noise energy at the current moment, the upper frequency boundary and the lower frequency boundary corresponding to the current moment are obtained, and the difference between the upper frequency boundary and the lower frequency boundary is used as the spectrum coverage bandwidth.

[0263] As a preferred embodiment, the frequency conversion interference feature library stores several standard operating condition modes, as well as adaptive filtering function templates associated with each of the standard operating condition modes.

[0264] The filter parameter generation module 503 performs similarity matching between the frequency band feature parameters and a pre-built frequency conversion interference feature library to obtain the filter algorithm corresponding to the current frequency conversion operating condition, including:

[0265] The filter parameter generation module 503 characterizes the frequency band feature parameters as a multi-dimensional operating condition feature vector of the current noise distribution characteristics.

[0266] The multidimensional working condition feature vector is projected onto a preset high-dimensional feature space to obtain the projection result; and the weighted Mahalanobis distance between the projection result and the center of each standard working condition mode is calculated respectively.

[0267] The standard working condition mode with the smallest weighted Mahalanobis distance is selected as the target matching mode, and the adaptive filtering function template corresponding to the target matching mode is called.

[0268] Based on the real-time rate of change of the noise center frequency drift, the order and topology of the filter are determined in the adaptive filtering function template, thereby generating a filtering algorithm for the current frequency conversion condition based on the order and topology.

[0269] As a preferred embodiment, the reconstruction module 504 filters the hybrid response signal based on the dynamic filtering parameters and filtering algorithm to remove dynamic noise signals that overlap with the frequency band of the detection pulse signal, thereby obtaining a reconstructed signal for the insulation leakage current signal, including:

[0270] The reconstruction module 504 constructs a time-varying notch filter based on the dynamic filtering parameters and the filtering algorithm;

[0271] The hybrid response signal is input into the time-varying notch filter to obtain a preliminary filtered signal;

[0272] The preliminary filtered signals are rearranged in reverse time order to obtain the rearrangement result;

[0273] The rearrangement result is input into the time-varying notch filter for secondary filtering to obtain a zero-phase-shift denoised signal.

[0274] Using the detection pulse signal as a reference signal, a dynamic orthogonal reference coordinate system composed of in-phase axes and quadrature axes is constructed; the zero-phase-shift denoised signal is projected onto the dynamic orthogonal reference coordinate system to obtain in-phase components and quadrature components with the same frequency as the detection pulse signal;

[0275] For the overlapping regions of the frequency bands, vector synthesis and amplitude gain compensation calculations are performed using the in-phase and quadrature components to obtain the reconstructed signal.

[0276] As a preferred embodiment, the voltage parameters include the voltage change rate and amplitude;

[0277] The insulation detection module 505 calculates the real-time insulation resistance value of the power supply circuit of the target power equipment under operating conditions based on the amplitude and phase of the reconstructed signal and the voltage parameters of the detection pulse signal, including:

[0278] The insulation detection module 505 selects multiple nonlinear feature sampling points within a single pulse cycle of the reconstructed signal based on the level flip trigger edge of the detection pulse signal; wherein, the nonlinear feature sampling points include transient sampling points during the capacitor charging transition process and static sampling points under the charging saturation trend;

[0279] An equivalent parallel circuit model including the insulation resistance to be measured and the system's distributed capacitance to ground is established. Combined with the voltage change rate of the probe pulse signal and the instantaneous current values ​​of the multiple nonlinear characteristic sampling points, a system of linear equations is constructed regarding the insulation resistance to be measured and the system's distributed capacitance to ground.

[0280] The linear equations are solved using the least squares iterative algorithm to calculate the capacitive current component in the reconstructed signal corresponding to the system's distributed capacitance to ground, and the capacitive current component is removed from the reconstructed signal to extract the purely resistive leakage current component.

[0281] The insulation resistance value is calculated based on the steady-state amplitude of the pure resistive leakage current component and the amplitude of the detection pulse signal.

[0282] As a preferred embodiment, the acquisition module 501 injects a detection pulse signal into the power supply circuit of the target power equipment and acquires the mixed response signal of the power supply circuit, and further includes:

[0283] The acquisition module 501 injects a detection pulse signal into the power supply circuit and acquires the system-to-ground distributed capacitance of the power supply circuit and the insulation resistance value obtained in the previous detection cycle; and adjusts the frequency and amplitude of the detection pulse signal according to the system-to-ground distributed capacitance of the power supply circuit and the insulation resistance value obtained in the previous detection cycle.

[0284] The system's distributed capacitance to ground is monitored, and when the system's distributed capacitance to ground increases, the frequency of the detection pulse signal is reduced, and the output amplitude of the detection pulse signal is increased accordingly, so as to ensure that the signal-to-noise ratio of the hybrid response signal is within a preset threshold range.

[0285] As a preferred embodiment, the insulation detection system 500 further includes a mode selection module, which is used before injecting a detection pulse signal into the power supply circuit of the target electrical equipment:

[0286] Determine the grounding system type of the target power equipment;

[0287] If the grounding system type is a TN system, monitor the operating status of the target power equipment;

[0288] When the target power equipment is in operation, a standby enable command is sent to the insulation monitoring terminal via the MODBUS communication protocol to put the insulation monitoring terminal into standby mode.

[0289] When the target power equipment is detected to have exited the operating state, the insulation monitoring terminal is controlled to start operation to implement offline online monitoring;

[0290] If the grounding system type is an IT system, then the insulation monitoring terminal is controlled to directly enter the real-time online monitoring mode.

[0291] As a preferred embodiment, the insulation detection system 500 further includes a fault location module, which is used after calculating the real-time insulation resistance value of the power supply circuit of the target power equipment in its operating state:

[0292] The real-time insulation resistance value is compared with the preset first warning threshold and second alarm threshold respectively;

[0293] When the real-time insulation resistance value is lower than the first warning threshold or the second alarm threshold, the fault location is performed based on the polarity or phase characteristics of the reconstructed signal.

[0294] If the power supply circuit is a DC system, the insulation fault is determined and displayed as occurring at the positive or negative terminal based on the current flow direction of the reconstructed signal.

[0295] If the power supply circuit is a three-phase AC system, the phase correlation between the reconstructed signal and the voltage of each phase is analyzed to determine and display the phase line where the insulation fault occurred.

[0296] Compared with the prior art, this invention application has the following beneficial effects:

[0297] This invention application provides an insulation detection method and system for power equipment. The insulation detection method includes: injecting a probe pulse signal into the power supply circuit of the target power equipment and acquiring a mixed response signal of the power supply circuit; wherein the mixed response signal includes an insulation leakage current signal excited by the probe pulse signal and a dynamic noise signal generated by the operation of the frequency converter; performing joint time-domain and frequency-domain analysis on the mixed response signal, and extracting the frequency band feature parameters of the dynamic noise signal within the current working cycle based on the joint analysis results; wherein the frequency band feature parameters include noise center frequency drift, harmonic energy peak value, and spectral coverage bandwidth; and performing similarity matching between the frequency band feature parameters and a pre-constructed frequency converter interference feature library to obtain... The system employs a filtering algorithm corresponding to the current frequency conversion operating condition; and generates real-time dynamic filtering parameters based on the filtering algorithm and the noise center frequency drift. The stopband center and stopband width of the dynamic filtering parameters are adaptively updated based on the noise center frequency drift and the spectral coverage bandwidth. The system filters the hybrid response signal based on the dynamic filtering parameters and the filtering algorithm to remove dynamic noise signals overlapping with the frequency band of the probe pulse signal, obtaining a reconstructed signal for the insulation leakage current signal. Based on the amplitude and phase of the reconstructed signal, combined with the voltage parameters of the probe pulse signal, the system calculates the real-time insulation resistance value of the power supply circuit of the target power equipment under operating conditions, thereby achieving insulation detection. This invention application obtains a filtering algorithm corresponding to the current frequency conversion condition by separating and extracting the frequency band feature parameters of the dynamic noise signal within the current working cycle and matching them with a pre-constructed frequency conversion interference feature library. This solves the inherent contradiction in the existing fixed filtering mechanism that cannot simultaneously ensure the integrity of signal extraction and broadband interference suppression. It can adaptively track and update the stopband center and stopband width according to the noise center frequency drift and the spectrum coverage bandwidth. While accurately eliminating dynamic noise signals, it preserves the amplitude and phase of the relatively weak detection pulse signal, effectively eliminates the distortion of leakage current signals, significantly improves the detection accuracy of insulation resistance under complex frequency conversion conditions, and avoids false alarms in the system.

[0298] Furthermore, by performing joint time-frequency domain analysis on the hybrid response current signal, characteristic parameters such as noise drift, energy peak, and bandwidth can be extracted in real time. This allows the system to quantify noise characteristics in real time, even when faced with atypical noise caused by unknown frequency converter access, changes in power grid topology, or equipment aging, without relying on a pre-set fixed noise model. This endows the insulation testing equipment with strong field adaptability and operating condition generalization ability.

[0299] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for insulation testing of electrical equipment, characterized in that, include: A detection pulse signal is injected into the power supply circuit of the target power equipment, and a mixed response signal of the power supply circuit is collected; wherein, the mixed response signal includes an insulation leakage current signal excited by the detection pulse signal and a dynamic noise signal generated by the operation of the frequency converter. The hybrid response signal is subjected to joint time-domain and frequency-domain analysis, and the frequency band characteristic parameters of the dynamic noise signal within the current working cycle are extracted based on the joint analysis results; wherein, the frequency band characteristic parameters include noise center frequency drift, harmonic energy peak value, and spectral coverage bandwidth; The frequency band feature parameters are matched with a pre-built frequency conversion interference feature library to obtain the filtering algorithm corresponding to the current frequency conversion operating condition; and based on the filtering algorithm and the noise center frequency drift, real-time dynamic filtering parameters are generated; wherein, the stopband center and stopband width of the dynamic filtering parameters are adaptively tracked and updated with the noise center frequency drift and the spectrum coverage bandwidth. Based on the dynamic filtering parameters and filtering algorithm, the hybrid response signal is filtered to remove the dynamic noise signal that overlaps with the frequency band of the detection pulse signal, thereby obtaining a reconstructed signal for the insulation leakage current signal. Based on the amplitude and phase of the reconstructed signal, and combined with the voltage parameters of the detection pulse signal, the real-time insulation resistance value of the power supply circuit of the target power equipment under operating conditions is calculated to achieve insulation detection.

2. The insulation testing method for power equipment as described in claim 1, characterized in that, The joint time-domain and frequency-domain analysis of the hybrid response signal, and the extraction of frequency band characteristic parameters of the dynamic noise signal within the current working cycle based on the joint analysis results, includes: The hybrid response signal is subjected to synchronous compressed wavelet transform, and the signal energy is compressed and rearranged to instantaneous frequency points in the frequency domain to obtain the time-frequency energy distribution matrix. Based on the time-frequency energy distribution matrix, the energy ridge trajectory of the dynamic noise signal on the time-frequency plane is extracted using an energy gradient-based ridge extraction algorithm. The difference between the instantaneous maximum and minimum frequency values ​​of the energy ridge trajectory within the current working cycle is calculated as the noise center frequency drift of the dynamic noise signal; the maximum energy amplitude point on the energy ridge trajectory is extracted to obtain the harmonic energy peak value. Using the energy ridge trajectory as the central axis, perform layer-by-layer energy integration along both sides of the frequency axis in the time-frequency energy distribution matrix to obtain the cumulative integrated energy; When the accumulated integrated energy reaches a preset proportion threshold of the total noise energy at the current moment, the upper frequency boundary and the lower frequency boundary corresponding to the current moment are obtained, and the difference between the upper frequency boundary and the lower frequency boundary is used as the spectrum coverage bandwidth.

3. The insulation testing method for power equipment as described in claim 1, characterized in that, The frequency conversion interference feature library stores several standard operating condition modes, as well as adaptive filtering function templates associated with each of the standard operating condition modes. The step of performing similarity matching between the frequency band feature parameters and a pre-built frequency conversion interference feature library to obtain the filtering algorithm corresponding to the current frequency conversion operating condition includes: The frequency band characteristic parameters are characterized as a multi-dimensional operating condition feature vector representing the current noise distribution characteristics; The multidimensional working condition feature vector is projected onto a preset high-dimensional feature space to obtain the projection result; and the weighted Mahalanobis distance between the projection result and the center of each standard working condition mode is calculated respectively. The standard working condition mode with the smallest weighted Mahalanobis distance is selected as the target matching mode, and the adaptive filtering function template corresponding to the target matching mode is called. Based on the real-time rate of change of the noise center frequency drift, the order and topology of the filter are determined in the adaptive filtering function template, thereby generating a filtering algorithm for the current frequency conversion condition based on the order and topology.

4. The insulation testing method for power equipment as described in claim 1, characterized in that, The process of filtering the hybrid response signal based on the dynamic filtering parameters and filtering algorithm to remove dynamic noise signals that overlap with the frequency band of the detection pulse signal, and obtaining a reconstructed signal for the insulation leakage current signal, includes: Based on the dynamic filtering parameters and filtering algorithm, a time-varying notch filter is constructed. The hybrid response signal is input into the time-varying notch filter to obtain a preliminary filtered signal; The preliminary filtered signals are rearranged in reverse time order to obtain the rearrangement result; The rearrangement result is input into the time-varying notch filter for secondary filtering to obtain a zero-phase-shift denoised signal. Using the detection pulse signal as a reference signal, a dynamic orthogonal reference coordinate system composed of in-phase axes and quadrature axes is constructed; the zero-phase-shift denoised signal is projected onto the dynamic orthogonal reference coordinate system to obtain in-phase components and quadrature components with the same frequency as the detection pulse signal; For the overlapping regions of the frequency bands, vector synthesis and amplitude gain compensation calculations are performed using the in-phase and quadrature components to obtain the reconstructed signal.

5. The insulation testing method for power equipment as described in claim 1, characterized in that, The process of injecting a detection pulse signal into the power supply circuit of the target power equipment and acquiring the mixed response signal of the power supply circuit further includes: A detection pulse signal is injected into the power supply circuit to obtain the system-to-ground distributed capacitance of the power supply circuit and the insulation resistance value obtained in the previous detection cycle; the frequency and amplitude of the detection pulse signal are adjusted according to the system-to-ground distributed capacitance of the power supply circuit and the insulation resistance value obtained in the previous detection cycle. The system's distributed capacitance to ground is monitored, and when the system's distributed capacitance to ground increases, the frequency of the detection pulse signal is reduced, and the output amplitude of the detection pulse signal is increased accordingly, so as to ensure that the signal-to-noise ratio of the hybrid response signal is within a preset threshold range.

6. The insulation testing method for power equipment as described in claim 1, characterized in that, Before injecting a probe pulse signal into the power supply circuit of the target electrical equipment, the method further includes: Determine the grounding system type of the target power equipment; If the grounding system type is a TN system, monitor the operating status of the target power equipment; When the target power equipment is in operation, a standby enable command is sent to the insulation monitoring terminal via the MODBUS communication protocol to put the insulation monitoring terminal into standby mode. When the target power equipment is detected to have exited the operating state, the insulation monitoring terminal is controlled to start operation to implement offline online monitoring; If the grounding system type is an IT system, then the insulation monitoring terminal is controlled to directly enter the real-time online monitoring mode.

7. The insulation testing method for power equipment as described in claim 1, characterized in that, After calculating the real-time insulation resistance value of the power supply circuit of the target power equipment in its operating state, the method further includes: The real-time insulation resistance value is compared with the preset first warning threshold and second alarm threshold respectively; When the real-time insulation resistance value is lower than the first warning threshold or the second alarm threshold, the fault location is performed based on the polarity or phase characteristics of the reconstructed signal. If the power supply circuit is a DC system, the insulation fault is determined and displayed as occurring at the positive or negative terminal based on the current flow direction of the reconstructed signal. If the power supply circuit is a three-phase AC system, the phase correlation between the reconstructed signal and the voltage of each phase is analyzed to determine and display the phase line where the insulation fault occurred.

8. An insulation testing system for power equipment, characterized in that, It includes a data acquisition module, an extraction module, a filter parameter generation module, a reconstruction module, and an insulation detection module; among which, The acquisition module is used to inject a detection pulse signal into the power supply circuit of the target power equipment and acquire the mixed response signal of the power supply circuit; wherein, the mixed response signal includes an insulation leakage current signal excited by the detection pulse signal and a dynamic noise signal generated by the operation of the frequency converter. The extraction module is used to perform joint time-domain and frequency-domain analysis on the hybrid response signal, and extract the frequency band characteristic parameters of the dynamic noise signal within the current working cycle based on the joint analysis results; wherein, the frequency band characteristic parameters include noise center frequency drift, harmonic energy peak value, and spectral coverage bandwidth; The filter parameter generation module is used to perform similarity matching between the frequency band feature parameters and a pre-built frequency conversion interference feature library to obtain the filter algorithm corresponding to the current frequency conversion operating condition; and to generate real-time dynamic filter parameters based on the filter algorithm and the noise center frequency drift; wherein, the stopband center and stopband width of the dynamic filter parameters are adaptively tracked and updated with the noise center frequency drift and the spectrum coverage bandwidth. The reconstruction module is used to filter the hybrid response signal based on the dynamic filtering parameters and filtering algorithm to filter out the dynamic noise signal that overlaps with the frequency band of the detection pulse signal, and obtain a reconstructed signal for the insulation leakage current signal. The insulation detection module is used to calculate the real-time insulation resistance value of the power supply circuit of the target power equipment in the operating state based on the amplitude and phase of the reconstructed signal and the voltage parameters of the detection pulse signal, so as to realize insulation detection.

9. The insulation detection system for power equipment as described in claim 8, characterized in that, The extraction module performs joint time-domain and frequency-domain analysis on the hybrid response signal, and extracts the frequency band characteristic parameters of the dynamic noise signal within the current working cycle based on the joint analysis results, including: The extraction module applies synchronous compressed wavelet transform to the hybrid response signal, compressing and rearranging the signal energy to instantaneous frequency points in the frequency domain to obtain a time-frequency energy distribution matrix. Based on the time-frequency energy distribution matrix, the energy ridge trajectory of the dynamic noise signal on the time-frequency plane is extracted using an energy gradient-based ridge extraction algorithm. The difference between the instantaneous maximum and minimum frequency values ​​of the energy ridge trajectory within the current working cycle is calculated as the noise center frequency drift of the dynamic noise signal; the maximum energy amplitude point on the energy ridge trajectory is extracted to obtain the harmonic energy peak value. Using the energy ridge trajectory as the central axis, perform layer-by-layer energy integration along both sides of the frequency axis in the time-frequency energy distribution matrix to obtain the cumulative integrated energy; When the accumulated integrated energy reaches a preset proportion threshold of the total noise energy at the current moment, the upper frequency boundary and the lower frequency boundary corresponding to the current moment are obtained, and the difference between the upper frequency boundary and the lower frequency boundary is used as the spectrum coverage bandwidth.

10. The insulation detection system for power equipment as described in claim 8, characterized in that, The frequency conversion interference feature library stores several standard operating condition modes, as well as adaptive filtering function templates associated with each of the standard operating condition modes. The filter parameter generation module performs similarity matching between the frequency band feature parameters and a pre-built frequency conversion interference feature library to obtain the filter algorithm corresponding to the current frequency conversion operating condition, including: The filter parameter generation module characterizes the frequency band feature parameters as a multi-dimensional operating condition feature vector of the current noise distribution characteristics. The multidimensional working condition feature vector is projected onto a preset high-dimensional feature space to obtain the projection result; and the weighted Mahalanobis distance between the projection result and the center of each standard working condition mode is calculated respectively. The standard working condition mode with the smallest weighted Mahalanobis distance is selected as the target matching mode, and the adaptive filtering function template corresponding to the target matching mode is called. Based on the real-time rate of change of the noise center frequency drift, the order and topology of the filter are determined in the adaptive filtering function template, thereby generating a filtering algorithm for the current frequency conversion condition based on the order and topology.

Images