A method and system for on-line monitoring of insulation of a medium voltage bus in a nuclear power plant

Abstract

The application discloses a kind of nuclear power plant medium voltage bus insulation on-line monitoring method and system, and the insulation response section is identified to bus operation data by executing low-frequency detection signal injection, impedance sensitivity is preferentially collected and frames are formed to form insulation response frame;Adaptive filtering matching is carried out to insulation response frame, and frequency domain feature weighted extraction is corrected by temperature drift, to build impedance calculation unit;Deterioration feature is mapped to form insulation reference spectrum by impedance calculation unit, to generate impedance deterioration sequence by evaluating deterioration rate and acceleration, to generate early warning signal frame by hierarchical marking;Impedance recovery accumulation feature is identified by early warning signal frame deterioration continuity detection, to activate hierarchical early warning response strategy to build insulation correction scheme;Form impedance trend curve based on insulation correction scheme insulation period inspection, determine data sending node by adaptive transmission time delay evaluation, execute centralized control response check to generate insulation monitoring report, improve the accuracy and early warning timeliness of nuclear power plant medium voltage bus insulation on-line monitoring.

Description

Technical Field

[0001] This invention relates to the field of power system insulation testing technology, and in particular to an online monitoring method and system for the insulation of medium-voltage busbars in nuclear power plants. Background Technology

[0002] The medium-voltage ungrounded power distribution system of a nuclear power plant is subjected to both electrical and thermal stresses over a long period of operation. The insulation performance of the busbars gradually degrades with the accumulation of years of operation. While a single grounding fault in an ungrounded system will not immediately interrupt power supply, if it is not detected and addressed in a timely manner, continued insulation degradation will lead to phase-to-phase short circuits or multi-point grounding faults, threatening the safe and stable operation of the unit. Existing insulation monitoring schemes rely on periodic power outages for detection or simple ground insulation impedance threshold alarms. The former cannot achieve continuous online monitoring, and the latter lacks sensitivity in the early stages of insulation degradation, making it difficult to provide reliable early warning information before significant insulation deterioration occurs.

[0003] The measured value of busbar insulation impedance can fluctuate reversibly due to changes in ambient temperature and the self-heating effect of load current. These reversible fluctuations are superimposed on the actual degradation signal in the monitoring data, making it easy for existing methods to confuse the two. In addition, the accelerated phase of insulation degradation, from slow to rapid, is difficult to identify at the data level. When impedance shows a phased rebound, there is also a lack of means to distinguish between the real improvement and the false recovery due to polarization effect, resulting in insufficient reliability of early warnings and a lack of data support for the allocation of operation and maintenance resources. Summary of the Invention

[0004] This invention discloses an online monitoring method and system for the insulation of medium-voltage busbars in nuclear power plants, aiming to solve problems such as misjudgment of insulation status caused by temperature drift interference, delayed early warning during the accelerated deterioration stage, and difficulty in distinguishing between false recovery and real improvement. By establishing a complete monitoring closed loop from signal injection and acquisition, temperature drift correction, deterioration trend assessment to early warning response and data transmission, continuous quantitative monitoring and graded early warning of the insulation status of medium-voltage busbars in nuclear power plants can be achieved.

[0005] The first aspect of this invention proposes an online monitoring method for the insulation of medium-voltage busbars in nuclear power plants, comprising the following steps:

[0006] Obtain busbar operation data of medium-voltage ungrounded system, perform low-frequency detection signal injection to identify insulation response sections based on the busbar operation data, and form insulation response frames based on impedance sensitivity priority acquisition frames based on the insulation response sections.

[0007] Adaptive filtering matching is performed based on the insulation response frame to form a filtering configuration. Temperature drift correction and frequency domain feature weighting extraction are performed on the filtering configuration to form weighted feature components. An impedance calculation unit is constructed through the weighted feature components.

[0008] The impedance calculation unit is subjected to degradation feature mapping to form an insulation reference spectrum. Based on the insulation reference spectrum, degradation rate and degradation acceleration are evaluated to generate an impedance degradation sequence. The impedance degradation sequence is graded and marked to generate a warning signal frame.

[0009] The degradation continuity detection of the warning signal frame is performed to identify the impedance rise accumulation characteristics. Based on the impedance rise accumulation characteristics, a graded warning response strategy is activated to form warning trigger parameters. The warning trigger parameters are used to extract the impedance deviation increment of the impedance degradation sequence to construct an insulation correction scheme.

[0010] Based on the insulation correction scheme, insulation periodic inspections are performed to form an impedance trend curve. An adaptive transmission delay assessment based on the early warning level is implemented on the impedance trend curve to determine the data uploading node. Based on the data uploading node, centralized control response verification is performed to generate an insulation monitoring report.

[0011] A second aspect of this invention provides an online monitoring system for the insulation of medium-voltage busbars in nuclear power plants, comprising:

[0012] The signal injection module is used to acquire the bus operation data of the medium-voltage ungrounded system, perform low-frequency detection signal injection to identify the insulation response section based on the bus operation data, and perform impedance sensitivity priority acquisition and framing based on the insulation response section to form an insulation response frame.

[0013] The feature extraction module is used to perform adaptive filtering matching based on the insulation response frame to form a filtering configuration, perform temperature drift correction and frequency domain feature weighting extraction on the filtering configuration to form weighted feature components, and construct an impedance calculation unit through the weighted feature components;

[0014] The degradation assessment module is used to map the degradation features of the impedance calculation unit to form an insulation reference spectrum, perform degradation rate and degradation acceleration assessment based on the insulation reference spectrum to generate an impedance degradation sequence, and implement graded marking for the impedance degradation sequence to generate a warning signal frame.

[0015] The early warning response module is used to detect and identify the impedance rise accumulation characteristics of the early warning signal frame, activate the hierarchical early warning response strategy based on the impedance rise accumulation characteristics to form early warning trigger parameters, and use the early warning trigger parameters to extract the impedance deviation increment of the impedance degradation sequence to construct an insulation correction scheme.

[0016] The uploading verification module is used to perform insulation periodic inspections based on the insulation correction scheme to form an impedance trend curve, perform adaptive transmission delay assessment based on the early warning level on the impedance trend curve to determine the data uploading node, and perform centralized control response verification to generate an insulation monitoring report based on the data uploading node.

[0017] The beneficial effects of this invention are reflected in the following points: 1. In response to the problem of significant differences in the response sensitivity of low-frequency detection signals in different bus sections, a weighted distribution analysis of response amplitude deviation is introduced to rank the insulation response sections by sensitivity, so that the acquisition resources are concentrated in the high-sensitivity sections rather than uniformly distributed; in the filtering configuration stage, the ambient temperature drift and the load self-heating temperature drift are calculated separately and correction coefficients are established separately, so as to eliminate the interference of reversible impedance fluctuations on the insulation status judgment from two independent mechanisms; when extracting frequency domain features, weights are assigned differently according to the acquisition priority, so that the construction of the impedance calculation unit can maintain a stronger perception of the insulation status changes in the high-sensitivity sections. 2. Deterioration rate assessment selects data from light-load periods to eliminate the contamination of rate estimation by self-heating effects. Based on this, statistical analysis of adjacent rate differences is introduced to identify abrupt changes in the rate plateau period. The fitting residual convergence screening mechanism is combined with acceleration fitting analysis, and only abrupt change segments with linear model fitting quality that meet the requirements are retained for the combined calculation of deterioration acceleration, so that the confidence of the deterioration acceleration value can be quantified. The warning level division boundary dynamically floats with the overall offset of the insulation reference spectrum, ensuring that the sensitivity of the acceleration signal judgment under the background of overall insulation deterioration is linked with the current safety margin in real time. 3. By combining the normalized rise amplitude and rise duration, a distinction is made between pseudo-recovery of polarization effect and actual insulation improvement. Differentiated early warning response strategies are activated for different pseudo-recovery levels to prevent pseudo-recovery from unreasonably lowering the early warning level. During the data transmission stage, a reverse-driven correction mechanism is introduced, which combines the impedance rise slope and the overall insulation offset to determine the relaxation range of the time delay threshold. This ensures that the rise under the background of overall degradation will not excessively reduce the transmission frequency. At the same time, the time delay threshold correction during pseudo-recovery is shielded, so that the monitoring data transmission strategy is consistent with the actual state of insulation risk. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating an online monitoring method for the insulation of medium-voltage busbars in nuclear power plants, according to the present invention.

[0019] Figure 2 This is a structural block diagram of an online monitoring system for the insulation of medium-voltage busbars in nuclear power plants, according to the present invention. Detailed Implementation

[0020] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0021] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0022] The technical solutions of the embodiments of this application will be described below.

[0023] like Figure 1 As shown, this embodiment of the invention provides a method for online monitoring of the insulation of a medium-voltage busbar in a nuclear power plant, including the following steps S110-S150:

[0024] Step S110: Obtain bus operation data of medium-voltage ungrounded system, perform low-frequency detection signal injection to identify insulation response sections based on bus operation data, and form insulation response frames based on impedance sensitivity priority acquisition of insulation response sections.

[0025] Specifically, busbar operation data of the medium-voltage ungrounded system is acquired. This data is continuously collected from the online monitoring devices of the nuclear power plant's medium-voltage distribution system. The data collection covers three types of quantities: three-phase-to-ground voltage, line current, and neutral point displacement voltage. The synchronous acquisition of these three types of quantities covers a complete representation of the busbar's electrical state. The sampling frequency of the three-phase-to-ground voltage is set based on the highest injection frequency of the low-frequency detection signal, satisfying the Nyquist sampling theorem while allowing at least a two-fold frequency margin to ensure that subsequent frequency domain feature extraction does not produce aliasing distortion. Line current sampling and three-phase voltage sampling are synchronously triggered under the same clock source, with the synchronization deviation controlled within 1% of the sampling period to avoid distortion of the voltage-current phase relationship during impedance calculation due to sampling timing misalignment. The neutral point displacement voltage is obtained from the open delta winding of the zero-sequence voltage transformer in the ungrounded system. This channel is the core quantity for detecting changes in the system's insulation symmetry. Under normal insulation symmetry, the neutral point displacement voltage is close to zero; a decrease in insulation impedance of any phase will cause a shift in the neutral point displacement voltage. If a phase voltage loss or transformer saturation anomaly occurs during busbar operation data acquisition, the corresponding acquisition period is marked as a data anomaly, and subsequent processing results within the data anomaly period are further annotated with an anomaly label. The amplitude range and harmonic content of busbar operation data vary significantly under different operating conditions of nuclear power plant units. During full-power operation, the three-phase voltage waveform is stable, but the line current is large. During power reduction and start-up / shutdown periods, the voltage waveform contains more transient components. The range and resolution settings of the acquisition device must cover the dynamic range of data under all operating conditions.

[0026] Low-frequency detection signal injection is used to identify insulation response sections based on bus operation data. The low-frequency detection signal is coupled to the bus zero-sequence circuit via an isolation transformer from a signal injection device. The injection frequency is selected from a specific low-frequency point other than the odd harmonics of the power frequency to avoid aliasing with system operation harmonics. The injection amplitude is adaptively adjusted based on the line current level in the current bus operation data. During periods of heavy load, the injection amplitude is appropriately increased to ensure that the signal-to-noise ratio of the insulation response signal meets the identification requirements. The coupling method of the isolation transformer ensures that the injected signal does not affect the normal power supply operation of the bus. At the beginning of each injection cycle, the injection device outputs a standard synchronization pulse. After the synchronization pulse is aligned with the clock of the acquisition device, the start and end times of the insulation response section can be accurately locked within the effective action window of the injected signal, eliminating the interference of injection time uncertainty on the boundary determination of the response section. In the bus operation data, the periodic amplitude variation segment of the zero-sequence voltage channel after low-frequency probe signal injection is identified as an insulation response segment. Only when the response amplitude exceeds three times the average background noise is the segment confirmed as a valid insulation response segment. This validity determination avoids misidentifying system transient disturbances as insulation responses. The brief zero-sequence voltage fluctuations caused by the switching operation of the nuclear power plant's auxiliary power supply differ significantly from the true insulation response in terms of duration and frequency characteristics; responses lasting less than one injection signal cycle are not included in the insulation response segment. The identification of insulation response segments is performed immediately after each injection cycle. The temporal sequence of the response amplitude of each segment is input to the deviation weighted distribution analysis stage as input for deviation calculation. The start and end sampling point numbers of each segment are mapped to the intra-frame segment location index during the framing stage.

[0027] In some embodiments, the step of prioritizing impedance sensitivity acquisition and framing to form an insulation response frame based on the insulation response segment includes: performing a weighted distribution analysis of the response amplitude deviation of the insulation response segment to obtain a sensitivity ranking result; performing acquisition priority calibration based on the sensitivity ranking result to generate a priority acquisition sequence; performing dynamic frame structure mapping based on the warning level on the priority acquisition sequence to form frame grouping parameters; and performing priority framing on the insulation response segment according to the frame grouping parameters to form an insulation response frame.

[0028] A weighted distribution analysis of response amplitude deviation was performed on the insulation response segments to obtain sensitivity ranking results. Response amplitude deviation measures the degree to which the measured response amplitude in each insulation response segment deviates from the reference amplitude of the normal insulation state. A higher deviation indicates that the insulation impedance change at the corresponding location in that segment is more sensitive to the injected signal. After the time sequence of the response amplitude of each insulation response segment is extracted, the difference is calculated with the normal insulation reference amplitude. The difference is divided by the reference amplitude and normalized to obtain the single deviation. The deviation of the same segment within multiple injection cycles is weighted by time to obtain the representative value of the segment deviation. The deviation with more recent time weights has a higher weight to highlight the latest changes in the insulation state. Weighted distribution analysis arranges the representative values ​​of the deviation of each insulation response segment according to their amplitude to form a deviation distribution. Segment clusters with high deviation values ​​indicate insulation degradation accumulation in the corresponding busbar segment or feeder circuit. In a nuclear power plant, a section of the auxiliary power busbar, due to prolonged exposure to a humid environment, exhibits persistently low insulation impedance, resulting in a clear high-value clustering of the representative deviation values ​​for the corresponding insulation response segment. The sensitivity ranking is constructed by arranging all insulation response segments in descending order of their representative deviation values, with the segment with the highest deviation ranked first. The sensitivity ranking results are output as a sequence of segment numbers. The representative deviation values ​​of segments with data anomalies are multiplied by a reduction factor before ranking to minimize the impact of outlier data on the sensitivity ranking results.

[0029] Priority acquisition sequences are generated based on the sensitivity ranking results. Insulation response segments with higher rankings are more sensitive to changes in insulation impedance, and their acquisition priority is directly determined by their ranking in the sensitivity ranking results. After removing segments marked as having abnormal data from the sensitivity ranking results, the percentile boundaries are recalculated based on the total number of non-abnormal segments. The top 30% of non-abnormal segments are designated as Level 1 acquisition priority, those between 30% and 60% are designated as Level 2, and the bottom 40% are designated as Level 3 acquisition priority. The Level 3 boundaries are dynamically updated based on the sensitivity ranking results. When the acquisition priority of the same insulation response segment jumps between two adjacent injection cycles, a jump label is added. For example, if a feeder circuit experiences a rapid increase in response amplitude due to early signs of insulation degradation, the corresponding segment's ranking quickly moves forward, triggering a jump. During frame grouping parameter mapping, additional frame capacity is added for this segment based on the jump label. Data anomaly-marked segments are not included in priority grading and are uniformly placed in the pending data collection queue. Once the data returns to normal in the next injection cycle, they will re-enter the ranking and receive the corresponding priority label. If data cannot be recovered for two consecutive injection cycles, a sensor status verification alarm is triggered. For example, a loose terminal of a feeder transformer caused continuous data anomalies. After the alarm was triggered, maintenance personnel checked and tightened the terminal on-site. Once the data returns to normal in the next cycle, this segment will re-enter the ranking. In the priority acquisition sequence, first-priority segments receive the maximum frame capacity allocation, and third-priority segments receive the minimum frame capacity allocation. The priority acquisition sequence's level distribution is updated synchronously with each round of sensitivity ranking results.

[0030] Frame grouping parameters are generated by dynamically mapping the priority acquisition sequence based on the warning level to a frame structure. The insulation warning level requirements for the medium-voltage ungrounded system in a nuclear power plant differ significantly between normal operation and warning trigger states. Under normal conditions, the frame structure focuses on coverage, while under warning conditions, it emphasizes high-density sampling of high-risk sections. Therefore, the generation of frame grouping parameters requires dynamically adjusting two core parameters—frame capacity allocation ratio and frame interval time—after sensing the current warning level. The current system warning level is read from the warning status register of the monitoring device. When the warning level is normal, the frame capacity allocation ratio in the frame grouping parameters is set as follows: 50% for the first-priority section, and 25% each for the second and third-priority sections. This even distribution across the three levels ensures that the insulation status of the entire busbar section can be periodically acquired under normal conditions. When the warning level is raised to Level 1, the frame capacity allocation ratio in the frame grouping parameters is adjusted so that Level 1 priority segments account for 70% and Level 3 priority segments are compressed to 10% to free up frame capacity resources. Concentrating frame capacity on high-risk segments improves the acquisition resolution of highly sensitive segments under warning conditions. When a rapid decrease in the insulation impedance of a certain busbar triggers a Level 1 warning, the corresponding insulation response segment achieves finer-grained continuous sampling by expanding the frame capacity ratio in the frame grouping parameters, helping the monitoring device capture detailed characteristics of the accelerated insulation degradation stage. The frame interval time in the frame grouping parameters shortens as the warning level increases, with the shortest frame interval limited to no less than the duration of a single complete cycle of the low-frequency detection signal.

[0031] Based on the frame grouping parameters, insulation response segments are prioritized and framed to form insulation response frames. During the framing process, according to the frame capacity allocation ratio for each priority level in the frame grouping parameters, the corresponding number of insulation response segment sampling points are sequentially extracted from the priority acquisition sequence to fill the frame capacity. First-priority segment sampling points are filled into the first part of the frame first, followed by second and third-priority segments in sequence. Sampling points from each priority segment within the frame are arranged in chronological order to maintain temporal continuity. The frame header of the insulation response frame carries the current warning level identifier, the injected signal frequency, and the start and end position indices of each priority segment within the frame. When processing frame data, the impedance calculation unit can directly locate the data boundaries of each priority segment based on the header indices, directly accessing the first-priority segment data without traversing the entire frame, thus improving the processing and response speed of high-risk signals. In insulation response sections, data to be supplemented in areas marked with data anomalies is packaged separately as supplementary frames after data recovery. The frame structure of the supplementary frames is the same as that of normal insulation response frames, but a supplementary sampling mark is added to the header. The supplementary sampling mark distinguishes between normal acquisition frames and supplementary acquisition frames. Supplementary acquisition frames after a brief saturation recovery of a certain section of bus transformer carry a supplementary sampling mark. During degradation trend analysis, it can accurately identify that the batch of data belongs to historical supplementation rather than the latest acquisition, avoiding the impact of timing misalignment on rate estimation. The insulation response frame generation rate is dynamically adjusted according to the warning level. Under normal conditions, one frame is generated per low-frequency detection signal injection cycle. Under warning conditions, multiple frames are generated within a single injection cycle according to the frame interval time parameter in the frame grouping parameters.

[0032] Step S120: Adaptive filtering matching is performed based on the insulation response frame to form a filtering configuration. Temperature drift correction and frequency domain feature weighting are applied to the filtering configuration to form weighted feature components. Impedance calculation unit is constructed through the weighted feature components.

[0033] Specifically, an adaptive filtering configuration is formed based on the insulation response frame. The adaptive filtering matching is based on the injected signal frequency carried in the insulation response frame header and the current warning level identifier. Different injection frequencies correspond to different filter passband widths and stopband attenuation requirements. When the injection frequency is low, the suppression of background power frequency harmonic interference is stronger, requiring a deeper stopband attenuation. When the injection frequency is high, the phase linearity within the passband must be considered to ensure accurate extraction of response phase information. The data quality of the first-priority segment in the insulation response frame directly determines the accuracy requirements of the filter parameter matching. The current signal-to-noise ratio (SNR) is estimated from the sampling points of the first-priority segment. When the SNR is low, the filter passband is adaptively narrowed to enhance the extraction accuracy of the injected signal; when the SNR is sufficient, the passband is appropriately widened to retain more complete response spectrum information. The filtering configuration, determined by the filter type, passband center frequency, and passband width, establishes the signal separation scheme under the current injection conditions. The filter type dictates the overall architecture of signal separation, the passband center frequency locks the frequency of the injected signal, and the passband width controls the frequency accuracy of signal extraction. The filter type is selected between bandpass and notch filter combinations based on the current system harmonic environment. In nuclear power plant auxiliary power systems, when harmonic content is high during periods of frequent load switching, notch filter combinations are preferred to actively suppress specific harmonic interference. The sampled data corresponding to the jump-marked sections in the insulation response frame triggers special processing during filter matching. Drastic signal amplitude changes in the jump-marked sections may cause overshoot in fixed-parameter filters. The special processing method involves re-estimating the filter parameters for this section of data and generating a local filter configuration. This local configuration only applies to the jump-marked sections and does not affect the filtering processing of other sections. The generated filter configuration is bound to the current insulation response frame number. During the temperature drift correction and frequency domain feature extraction stages, the corresponding parameters are directly read according to the frame number, without cross-frame calls.

[0034] In some embodiments, the step of performing temperature drift correction and frequency domain feature weighted extraction to form weighted feature components on the filtering configuration includes: extracting the current ambient temperature sample value and load current sample value based on the filtering configuration; performing ambient temperature drift and self-heating temperature drift separation calculation based on the ambient temperature sample value and the load current sample value to obtain a comprehensive temperature drift correction coefficient; performing deviation compensation based on the comprehensive temperature drift correction coefficient and the frequency domain features of the filtering configuration to generate compensated frequency domain features; and performing response intensity weighted extraction to form weighted feature components based on the compensated frequency domain features.

[0035] The ambient temperature and load current samples are extracted based on the filtering configuration. The ambient temperature is read in real-time from temperature sensors installed inside the medium-voltage switchgear, with sensor locations covering representative positions of each section of the busbar. The temperature readings from each point are weighted by distance to form the overall representative ambient temperature of the current busbar. Points closer to the low-frequency detection signal injection point have higher weights, as temperature changes near the injection point have the most direct impact on insulation impedance. The load current is extracted from the busbar current transformer readings within the sampling period corresponding to the insulation response frame of the filtering configuration. The average of the effective current values ​​within the sampling period is used as the representative value, and averaging eliminates the interference of short-term load fluctuations on temperature drift estimation. Both the ambient temperature and load current samples are aligned with the insulation response frame number to ensure that the temperature and current data used for temperature drift separation calculation are strictly synchronized with the current frame's acquisition time. During the unit power increase process in the medium-voltage system of a nuclear power plant, the bus current rises rapidly. If the temperature and current data are not synchronized, it will lead to a significant deviation in the estimation of self-heating temperature drift, thus affecting the accuracy of the correction coefficient. When an abnormal sensor reading occurs in the ambient temperature sampling value, the corresponding point is replaced by the interpolation of the adjacent normal point. The interpolation replacement is marked in the temperature data record. When the interpolation replacement ratio exceeds 50% of the total number of points, a low confidence label is added to the temperature drift correction reliability of the current frame. This label is passed into the temperature drift separation calculation stage with the frame. After the comprehensive temperature drift correction coefficient is generated, it is converted into a low confidence inherited label and added to the corresponding frame.

[0036] The comprehensive temperature drift correction coefficient is obtained by separately calculating environmental temperature drift and self-heating temperature drift based on the sampled values ​​of ambient temperature and load current. The temperature dependence of insulation impedance stems from two independent mechanisms: ambient temperature changes cause overall temperature rise and fall of the insulation material, resulting in environmental temperature drift; load current generates Joule heat through the conductor and conducts it to the insulation layer, forming self-heating temperature drift. The contribution ratio of these two factors to the total temperature drift dynamically changes with the load condition. During light load periods, the contribution of self-heating temperature drift is small, while during heavy load periods, the contribution of self-heating temperature drift may exceed that of ambient temperature drift. Therefore, they need to be estimated separately and then superimposed. The environmental temperature drift coefficient α_env, in units of 1 / ℃, is obtained by interpolation from the ambient temperature sampling value and the thermal resistance reference model of the insulation material. The self-heating temperature drift coefficient α_self, in units of 1 / A², is estimated by multiplying the square of the load current sampling value by the normalized thermal resistance coefficient R_thermal. The two coefficients are weighted and superimposed according to the current temperature and current level to obtain the comprehensive temperature drift correction coefficient K, i.e., K=α_env×(T_env-T_ref)+α_self×I²×R_thermal, where T_env is the current ambient temperature sampling value in ℃; T_ref is the standard reference temperature in ℃; I is the load current sampling value in A; R_thermal is the normalized thermal resistance coefficient, dimensionless; and K is the dimensionless comprehensive temperature drift correction coefficient. The calculation of the comprehensive temperature drift correction factor is performed independently on a frame-by-frame basis. Temperature drift factors from different acquisition periods are not averaged across frames. This is because the current change rate is relatively fast during power changes in nuclear power plant units, and cross-frame averaging would mask the sudden changes in self-heating temperature drift caused by short-term load changes, resulting in the correction factor lagging behind the actual temperature drift state. After the comprehensive temperature drift correction factor for the frame corresponding to the low-confidence inherited temperature drift data is calculated, a low-confidence inherited label is added. The deviation compensation range for the low-confidence inherited labeled frame is conservatively limited to no more than 15% of the measured amplitude.

[0037] Deviation compensation is performed based on the frequency domain characteristics of the comprehensive temperature drift correction coefficient and the filtering configuration to generate compensated frequency domain features. The compensated frequency domain features are extracted from the insulation response signal after filtering. The frequency domain processing result of the filtering configuration includes two terms: response amplitude and response phase at the injection frequency. The response amplitude reflects the degree of attenuation of the injected signal by the insulation impedance, and the response phase reflects the ratio of capacitive to resistive components of the impedance. These two features jointly describe the electrical characteristics of the current insulation state. Deviation compensation applies the comprehensive temperature drift correction coefficient K to the response amplitude. The compensation formula is A_corrected = A_measured / (1 + K), where A_measured is the measured response amplitude of the current frame (in amperes); A_corrected is the compensated amplitude (in amperes). The temperature drift correction term in the denominator removes the impedance shift caused by temperature from the measured amplitude, making the compensated amplitude more accurately reflect the intrinsic degradation state of the insulation material rather than the reversible changes caused by temperature fluctuations. The phase drift compensation for the response phase uses a phase correction coefficient β, which is determined by the temperature coefficient of the dielectric constant of the insulating material. β is a phase temperature drift sensitivity coefficient reflecting the temperature sensitivity of the material's dielectric response, with dimensions in rad. The corrected phase φ_corrected = φ_measured - β × K, where φ_measured is the measured response phase in rad; φ_corrected is the compensated phase in rad; and β × K has the same dimension as φ_measured. Phase correction ensures that the estimation of the capacitive component is not affected by temperature changes. During the startup and heating phase of a nuclear power plant unit, the rapid temperature rise of the bus insulation material leads to an overestimation of the capacitive component. Phase temperature drift compensation can distinguish this reversible change from the phase shift caused by actual insulation degradation. The compensation amplitude for the low-confidence inherited annotation frame is limited to no more than 15% of the measured amplitude.

[0038] Weighted feature components are formed by weighted extraction of response intensity based on the compensated frequency domain characteristics. The response amplitude A_corrected and response phase φ_corrected of each insulation response segment in the compensated frequency domain characteristics are assigned weight coefficients w_p according to the acquisition priority level of each segment. First-priority segments receive the highest weight, and third-priority segments receive the lowest weight. This weight differentiation ensures that the weighted feature components are more sensitive to changes in the insulation state of high-sensitivity segments. After weighting, the amplitude and phase components of each segment are superimposed according to vector synthesis rules. The superposition result is the weighted feature component of the current frame. Vector synthesis preserves the phase information of the response of each segment. When the response phases of different segments are similar, the superposition is enhanced; when the phases differ significantly, the superposition cancels out. Phase cancellation indicates a difference in the direction of insulation degradation in different busbar segments. When a busbar segment is predominantly resistively degraded while an adjacent segment is predominantly capacitively degraded, the opposite directions of the response vectors of the two segments suppress the superposition amplitude. In this case, the compensated frequency domain characteristics of each segment need to be judged independently. When extracting weighted segments, the jump-marked segments are additionally multiplied by a jump intensity coefficient. This coefficient is obtained by normalizing the change in sensitivity ranking of the segment between the previous and current injection cycles. The larger the jump amplitude, the higher the coefficient, thus amplifying the contribution of the jump-marked segments in the weighted feature components. The real part of the complex number aggregates the weighted response in the equivalent resistance direction of each segment, while the imaginary part aggregates the weighted response in the equivalent capacitance direction. These two components correspond to the perception dimensions of insulation resistance degradation and capacitive degradation, respectively.

[0039] Impedance calculation unit is constructed using weighted characteristic components. The impedance calculation unit takes the complex expression of the weighted characteristic components as input and calculates the equivalent insulation impedance magnitude by the ratio of the known voltage amplitude of the injected signal to the measured response amplitude of the weighted characteristic components. The phase difference is used to separate the specific values ​​of the resistive and capacitive components. The impedance magnitude Z is determined by the formula Z=V_inject / A_corrected_weighted, where V_inject is the voltage amplitude of the low-frequency probe signal injected in V; A_corrected_weighted is the equivalent response current amplitude of the weighted characteristic components in A; and the dimension of Z is Ω, i.e., [V] / [A]=[Ω]. The resistive component R_ins and capacitive component C_ins, corresponding to the real and imaginary parts of the weighted characteristic component, are obtained through trigonometric decomposition of Z and the response phase: R_ins = Z × cos(φ_corrected_weighted), C_ins = 1 / (ω × Z × sin(φ_corrected_weighted)), where ω is the angular frequency of the injected signal in rad / s; φ_corrected_weighted is the equivalent response phase of the weighted characteristic component, which is synthesized by vector synthesis of φ_corrected in each segment according to priority weights in rad. The physical meanings of the two components correspond to the leakage resistance and equivalent capacitance of the insulating material, respectively. A continuous decrease in R_ins is a typical characteristic of insulation moisture or carbonization degradation, and the unit of R_ins is Ω; a continuous increase in C_ins indicates a reduction in the thickness of the insulation layer or an increase in the dielectric constant of the material, and the unit of C_ins is F. The impedance calculation unit uses the R_ins and C_ins of the current frame as core parameters, along with corresponding warning level indicators, frame numbers, and low-confidence inheritance labeling status. For frames with low-confidence inheritance labels, the R_ins and C_ins values ​​of the corresponding impedance calculation unit participate in the degradation feature mapping but are weighted down during the reference spectrum construction stage to ensure that unreliable temperature drift correction data does not have a dominant impact on the establishment of the insulation reference spectrum. The impedance calculation units are arranged sequentially by frame number; a rapid change labeling is triggered when the change in R_ins between adjacent frames exceeds 10% of the current value.

[0040] Step S130: The impedance calculation unit is subjected to degradation feature mapping to form an insulation reference spectrum. Based on the insulation reference spectrum, degradation rate and degradation acceleration are evaluated to generate an impedance degradation sequence. The impedance degradation sequence is graded and marked to generate a warning signal frame.

[0041] Specifically, the impedance calculation unit is subjected to degradation feature mapping to form an insulation reference spectrum. The degradation feature mapping arranges the R_ins and C_ins values ​​of each frame in the impedance calculation unit in chronological order, and extracts the statistical benchmark under normal insulation conditions. The normal benchmark is confirmed by referring to the insulation test data during the cold shutdown of the nuclear power plant unit. During the cold shutdown period, the load current is zero and the temperature is stable, at which time the R_ins and C_ins of the impedance calculation unit best reflect the intrinsic state of the insulation material. The insulation reference spectrum uses the mean R_ins and mean C_ins as the reference points in a dual-axis coordinate system. The offset of each frame of the impedance calculation unit in the reference coordinate system reflects the degree to which the current insulation state deviates from the normal benchmark. The direction of the offset distinguishes between two degradation modes: resistive degradation and capacitive degradation. Resistive degradation corresponds to a continuous decrease in R_ins while C_ins does not change significantly, while capacitive degradation corresponds to a continuous increase in C_ins while R_ins changes relatively steadily. The distribution of offsets in each frame of the insulation reference spectrum evolves over time, forming a degradation trajectory. A monotonically extending trajectory towards lower resistance indicates moisture absorption or continuous accumulation of surface contaminants in the insulation. In a nuclear power plant building, during the rainy season, persistently high humidity leads to increased leakage current on the insulation surface of a certain busbar. The corresponding impedance calculation unit's R_ins slowly declines over several weeks, resulting in a gradual degradation trajectory extending towards lower resistance in the insulation reference spectrum. Offsets from low-confidence inherited annotation frames participate in the construction of the insulation reference spectrum, but their weights are reduced. This reduction decreases the impact of the offsets on the trajectory, preventing unreliable frame data from temperature drift correction from distorting the overall shape of the reference spectrum.

[0042] In some embodiments, the step of performing degradation rate and degradation acceleration assessment based on the insulation reference spectrum to generate an impedance degradation sequence includes: extracting continuous-time impedance change records from the insulation reference spectrum; performing a rate change analysis of the continuous-time impedance change records during light-load periods to obtain a degradation rate sequence; identifying abrupt changes in the rate plateau phase of the degradation rate sequence to obtain a degradation acceleration value; and performing threshold comparison and sorting based on the degradation acceleration value to form an impedance degradation sequence.

[0043] Extract continuous-time impedance change records from the insulation reference spectrum. After arranging the frame offsets in the insulation reference spectrum in ascending order according to the acquisition timestamp, frames that are temporally continuous and without data interruption are extracted as continuous time periods. A data interruption is defined when the time interval between frames exceeds two injection cycles. The interruption points divide the temporal sequence of the insulation reference spectrum into several independent continuous time periods. The R_ins and C_ins change sequences of each time period in the continuous-time impedance change record are independent and do not merge across interruption points. The change is defined as the difference between the corresponding parameters of adjacent frames. A negative difference indicates a decrease in R_ins, i.e., the progression of resistive degradation, while a positive difference indicates an increase in C_ins, i.e., the progression of capacitive degradation. The stability of the absolute value of the difference reflects the uniformity or acceleration of the degradation progress. In the early stage of moisture absorption of a certain bus insulation, the absolute value of the difference remains at a low level with small fluctuations for a long time. However, after the initiation of partial discharge, the absolute value of the difference suddenly jumps, and the change sequence shows obvious non-uniform characteristics. The minimum effective length of the continuous impedance change record is set to cover at least one complete load change cycle. Nuclear power plant units typically have a load dispatch cycle of 8 to 12 hours. When the length of a continuous period is less than one load cycle, a short period is added and merged with adjacent continuous periods to ensure that the time span of the rate estimation is sufficiently stable. In the continuous impedance change record, the period containing the rapidly changing frame is additionally marked with the boundary of the rapidly changing segment. The boundary of the rapidly changing segment separates the change sequence into a rapidly changing interval and a stable changing interval. If a bus section experiences a rapid drop in R_ins due to a brief overload, the rapidly changing segment boundaries on both sides of the corresponding frame segment isolate this extreme change. The changes in the stable changing interval can still objectively reflect the deterioration progression rhythm under normal operating conditions.

[0044] A degradation rate sequence was obtained by analyzing the rate of change of impedance changes over continuous periods during light load periods. The change sequences within each continuous period of the impedance change records were filtered by light load segments and then subjected to linear regression. The self-heating effect was largely eliminated during light load periods, and the rate of change more closely approximated the intrinsic degradation progress of the insulation material. Heavy load periods were not included in the main regression due to the superposition of self-heating effects. Light load periods were identified from the load current sampling values ​​corresponding to the impedance calculation units. Segments with load currents lower than 30% of the rated current were determined to be light load segments. The distribution of light load segments within continuous periods typically corresponds to nighttime off-peak load periods. During the nighttime period when the nuclear power plant's auxiliary power is operating at full power in a steady state, the auxiliary power load is relatively stable and at a light load level. The change in the insulation reference spectrum offset during this period best reflects the true degradation progress of the insulation. Data at the boundaries of rapidly changing segments in the continuous impedance change records were regressed separately and not merged with stable segments. The slope of the rapidly changing intervals was recorded separately and marked for isolation in the degradation rate sequence, and was not included in the main analysis of the difference statistics. The degradation rate sequence is composed of the linear regression slopes of the changes in light-load frames within each consecutive time period, arranged chronologically. The regression slope v_deg = ΔR_ins / Δt, where ΔR_ins is the change in R_ins between adjacent time periods in ohms; Δt is the corresponding time interval in hours; and v_deg is in ohms per hour. The larger the absolute value of the slope, the faster the insulation degradation rate in the current time period. A slope turning from negative to positive indicates a phased recovery in insulation impedance, which needs to be determined in conjunction with the insulation reference spectrum trajectory morphology to determine whether it is a false recovery. When the number of light-load frames in a consecutive time period is less than 3, rate regression is not performed for that consecutive time period. The corresponding position in the degradation rate sequence is replaced by the average rate of adjacent effective time periods with an estimated label. During the short-term power reduction operation of nuclear power plant units, the acquisition window is compressed, and the number of light-load frames may not meet the regression conditions. In this case, the average rate of adjacent effective time periods can provide a reasonable reference replacement value to avoid gaps in the sequence.

[0045] For example, the step of identifying rate stationary abrupt changes in the degradation rate sequence to obtain degradation acceleration values ​​includes: statistically analyzing the differences between adjacent rates in the degradation rate sequence to generate difference distribution features; identifying accelerated abrupt change segments that exceed the rate baseline boundary based on the difference distribution features to generate abrupt change segment identifiers; performing acceleration fitting analysis on the abrupt change segment identifiers to form acceleration fitting parameters; and performing fitting residual convergence screening based on the acceleration fitting parameters to generate degradation acceleration values.

[0046] The degradation rate sequence is statistically analyzed using the differences between adjacent rates to generate a difference distribution characteristic. The difference between adjacent rates reflects the change in degradation rate from one period to the next; a positive difference indicates an increasing rate (accelerated degradation) and a negative difference indicates a decreasing rate (slowing degradation). The absolute value of the difference directly reflects the severity of the rate change. The differences between all adjacent rate pairs in the degradation rate sequence are arranged chronologically to form a difference sequence. The mean, standard deviation, and range are extracted from this sequence to constitute the difference distribution characteristic. The mean reflects whether the overall rate is accelerating, the standard deviation reflects the stability of the rate change, and the range captures the most dramatic single rate jump in the sequence. A large standard deviation in the difference distribution characteristics indicates that the rate changes within the degradation rate sequence are uneven, with stable rates in some periods and drastic rate fluctuations in others. This non-uniform distribution is a typical background for identifying abrupt changes during the stationary period. After long-term stable operation, the bus insulation of a nuclear power plant may experience a sudden rate jump due to the initiation of partial discharge, resulting in a large difference value in the difference sequence that is much higher than the mean. The range value is significantly high. The rate differences at estimated marked positions in the degradation rate sequence are weighted down during statistical analysis. Differences at estimated positions may originate from data splicing errors rather than actual rate changes. After weighting down, the contribution of estimated differences to the mean and standard deviation is reduced, ensuring that the three statistical measures of the difference distribution characteristics mainly reflect the true rate change patterns of the measured data and are not affected by artificial differences introduced by data gaps.

[0047] Based on the difference distribution characteristics, accelerated mutation segments exceeding the rate baseline boundary are identified, generating mutation segment identifiers. The rate baseline boundary is determined by adding a specific multiple of the standard deviation to the mean of the difference distribution characteristics. Differences exceeding the baseline boundary correspond to rate jumps exceeding the normal fluctuation range of the stationary period and are identified as mutation candidate points. The multiple coefficient is set to 1.5 to suppress false detections caused by random fluctuations within the stationary period while maintaining the sensitivity of true mutation detection. When the standard deviation of the difference distribution characteristics is large, the multiple coefficient of the baseline boundary is adaptively increased to 2.0. A large standard deviation means that the rate itself fluctuates more frequently, and a low multiple coefficient will cause a large number of normal fluctuations to be misjudged as mutations. After increasing the multiple coefficient, only differences with amplitudes far exceeding the average fluctuation level trigger mutation candidate determination. When two or more consecutive adjacent positions are mutation candidate points, they are merged into one accelerated mutation segment. The start and end positions jointly define the time range of this segment. In the acceleration fitting stage, corresponding segments are extracted from the degradation rate sequence based on the start and end positions for regression analysis. The abrupt change segment identifier consists of descriptions of all accelerated abrupt change segments arranged in ascending order of occurrence time. Abrupt change segments closer to the current time are prioritized in the fitting analysis phase because recent abrupt changes are more indicative of the current insulation condition. If a busbar section frequently experiences accelerated abrupt change segments at increasingly shorter intervals towards the end of its operation, the continuously increasing segment density in the abrupt change segment identifier itself is an early warning signal of rapid insulation degradation. A degradation rate sequence with a negative mean in the difference distribution characteristics does not trigger accelerated abrupt change identification. A negative mean indicates that the overall rate is decreasing, meaning the degradation is slowing down, and the abrupt change segment identifier is an empty set. This situation usually corresponds to a stage of improved insulation environmental conditions. After the rainy season ends, the humidity in the busbar environment decreases, insulation surface leakage decreases, R_ins shows a phased rebound, the mean of the rate sequence difference turns negative, and the insulation condition enters a phased stable window.

[0048] Acceleration fitting analysis is performed on the abrupt change segment markers to generate acceleration fitting parameters. Acceleration fitting uses the degradation rate sequence sub-segments covered by each accelerated abrupt change segment in the abrupt change segment marker as the fitting object. Linear regression is performed on the rate time series within each abrupt change segment in the abrupt change segment marker, and the regression slope is the estimated degradation acceleration value of that segment. The fitting formula is v(t) = v_0 + a × t, where v(t) is the rate at time t, in ohms per hour; v_0 is the initial rate of the segment, in ohms per hour; a is the acceleration fitting value, in ohms per hour squared; t is the cumulative time since the start of the segment, in hours; the least squares method is used to solve for a to minimize the sum of squared regression residuals. The acceleration fitting parameters record the acceleration fitting value a, the coefficient of determination R², and the number of effective sample points for each abrupt change segment. a reflects the rate growth intensity, R² measures the quality of the linear model's fit to the rate change of that segment, and the number of effective sample points assesses the sufficiency of the regression data; a higher R² indicates a better fit. Abrupt change segments with an R² below 0.5 are labeled as having low fitting quality. Low fitting quality indicates that the rate change in this segment exhibits nonlinear characteristics. For example, in the initial stage of partial discharge, the rate change of a certain bus insulation segment follows a nonlinear pattern of rapid initial change followed by a slower rate change, resulting in a low linear fitting R². This segment is labeled as having low fitting quality in the acceleration fitting parameters, with a reduction factor of 0.7, and is included in the acceleration candidate set after weight reduction. Abrupt change segments with fewer than 3 valid sample points are not subjected to linear regression. The corresponding positions of the acceleration fitting parameters are filled with the mean of the acceleration estimates of adjacent valid segments, and a insufficient sample label is added. This situation usually occurs at the end of a continuous time period due to data truncation leading to fewer sample points. Filling with the mean of adjacent valid segments can maintain the temporal continuity of the acceleration fitting parameter sequence.

[0049] Degraded acceleration values ​​are generated by filtering out the fitting residuals based on the acceleration fitting parameters. The purpose of this filtering is to eliminate abrupt changes in the acceleration fitting parameters that result in unreliable acceleration estimates due to insufficient fitting quality, retaining high-quality fitting results as the source of the final degraded acceleration values. The fitting residual convergence is determined using the autocorrelation coefficient of the residual sequence as the convergence index. A residual convergence is considered when the absolute value of the autocorrelation coefficient is below 0.3; a value exceeding 0.5 indicates residual divergence, suggesting that there are systematic components outside the linear range of the rate change that have not been captured by the model. For example, during the rainy season, continuous dampness in the insulation leads to an exponential rather than nonlinear increase in the rate, resulting in a significant positive autocorrelation in the corresponding residual sequence, thus indicating residual divergence and requiring weight reduction. Low-fit quality and insufficient sample sections are not included in the residual convergence determination but are directly included in the acceleration fitting parameter candidate set after weight reduction according to the corresponding reduction coefficients: 0.7 for low-fit quality and 0.6 for insufficient sample sections. The fitted values ​​of each segment in the acceleration fitting parameters determined by residual convergence are merged according to the R² weighted average. The merging formula is a_final=Σ(R²_i×a_i) / Σ(R²_i), where a_final is the deterioration acceleration value in ohms per hour squared. The deterioration acceleration value after merging multiple abrupt change segments is accompanied by a confidence assessment. The confidence level is jointly determined by the number of residual convergence segments participating in the merging and the proportion of reduced-weight segments. When reduced-weight segments account for the majority, the confidence level is lowered. In the late stage of deterioration of a certain busbar in a nuclear power plant, multiple abrupt change segments are marked as low-fit quality due to the nonlinear characteristics of rate changes. The confidence level of the merged deterioration acceleration value is lowered accordingly, and the corresponding threshold comparison results are conservative, reflecting the prudent handling of the estimated values ​​from the source of low-quality fitting.

[0050] An impedance degradation sequence is formed by comparing and ranking thresholds based on degradation acceleration values. The degradation acceleration values ​​are compared with preset acceleration grading thresholds to determine the severity level of each abrupt change segment. The acceleration grading thresholds are statistically determined based on the degradation acceleration characteristics of historical insulation accidents in similar nuclear power plants before such accidents, reflecting the empirical correlation between different degradation rate growth levels and insulation failure risk. The thresholds are adjusted downwards in tandem with the overall offset of the insulation reference spectrum. A larger offset indicates that the insulation has entered the middle to late stages of degradation, and the actual risk corresponding to the same acceleration value is higher; therefore, lowering the threshold triggers a higher severity judgment. Abrupt changes with degradation acceleration values ​​exceeding the high threshold are classified as high-level degradation events; those between the high and low thresholds are classified as medium-level degradation events; and those below the low threshold are classified as low-level degradation events. This three-level classification covers the entire degradation range from slow degradation to rapid deterioration, with the warning response intensity increasing sequentially for each level. The impedance degradation sequence arranges all abrupt change segments in descending order of degradation acceleration value, with higher-level degradation events at the front. Within the same level, events are sorted by the order of occurrence time of the abrupt change segment, prioritizing events closer to the current moment to highlight the latest degradation dynamics. Historical events from a long time ago are retained in the impedance degradation sequence, but their weight decays over time during the grading and marking stage. When there are no abrupt change segments in the degradation rate sequence, the impedance degradation sequence only contains stationary period entries. The degradation acceleration value of stationary period entries is assigned to zero. A zero value corresponds to a stable insulation state with no signs of accelerated degradation. After the nuclear power plant busbar has undergone thorough drying treatment and been put back into operation following the annual overhaul, the mean difference of the rate sequence is close to zero and no abrupt change candidate points are generated within multiple consecutive injection cycles. The impedance degradation sequence only presents stationary period entries, and the insulation state is in a short-term healthy operation phase.

[0051] In some embodiments, the step of generating a warning signal frame by implementing graded marking for the impedance degradation sequence includes: dividing the impedance degradation sequence into degradation acceleration level intervals to obtain a grade division result; performing warning threshold matching based on the grade division result to generate a graded warning identifier; performing adjacent level cross-level jump constraint verification on the graded warning identifier to generate a constraint graded warning identifier; and associating the constraint graded warning identifier with the impedance degradation sequence to form a warning signal frame.

[0052] The impedance degradation sequence is divided into degradation acceleration level intervals to obtain the classification results. The boundary of the level interval dynamically depends on the overall offset of the insulation reference spectrum. When the overall offset of the insulation reference spectrum has deviated significantly from the normal reference, it indicates that the insulation is in the middle stage of degradation. At this time, the actual risk corresponding to the same degradation acceleration value is higher than that when the overall offset of the reference spectrum is smaller. Therefore, the boundary of the level interval is adjusted downward in sync with the overall offset of the insulation reference spectrum to maintain consistency between the classification judgment and the overall state of the insulation. High-level degradation events in the impedance degradation sequence are classified into the high-risk interval, medium-level degradation events into the warning interval, and low-level degradation events and stable period events into the normal interval. The classification results mark the interval to which each item belongs. The interval mark directly drives the determination of the corresponding warning level during the warning threshold matching stage. When the overall offset of the insulation reference spectrum exceeds the initial reference value by 20%, the lower boundary of the high-risk zone for the corresponding time period entry in the impedance degradation sequence is lowered. This means that a lower degradation acceleration value is sufficient to trigger the high-risk zone determination. The lower boundary adjustment reflects a higher alertness to acceleration signals in the context of overall insulation degradation. After many years of cumulative operation, the reference spectrum offset of a certain busbar continues to accumulate. Even if the current acceleration value is not high, it may still be judged as entering the high-risk zone due to the overall offset exceeding the threshold. The corresponding physical meaning at this time is that although the insulation has not deteriorated rapidly, it is already in a vulnerable state with a low insulation margin. The grade classification result of the estimated marked entries in the impedance degradation sequence is downgraded by one level. The acceleration uncertainty from the estimation source is transformed into a more conservative zone determination through downgrading. Entries that are still judged as high-risk zones after downgrading retain the high-risk mark during the verification stage and are not downgraded again. This indicates that even under conservative estimation, the severity of degradation of the entry still meets the high-risk determination criteria.

[0053] Based on the classification results, warning threshold matching is performed to generate tiered warning labels. Warning threshold matching maps each interval label in the classification results to the warning level definition of the warning system: high-risk intervals correspond to Level 1 warnings, alert intervals to Level 2 warnings, and normal intervals to a no-warning state. The warning level definition is consistent with the classification standards for insulation status in nuclear power plant safety regulations, ensuring that the tiered warning labels directly drive the response procedures of maintenance personnel without requiring additional level conversion. When multiple different interval entries exist within the same impedance degradation sequence in the classification results, the tiered warning label takes the warning level corresponding to the highest interval entry. The principle of higher levels covering lower levels ensures that the most severe degradation signal is always reflected in the tiered warning label and is not diluted by the majority of lower-level entries. The graded early warning indicator includes the core basis for triggering the current early warning level. The core basis covers the degradation acceleration value, the time of occurrence of the abrupt change section, and the current overall offset of the insulation reference spectrum. The three present the current degradation progress intensity, the time and location of degradation acceleration, and the degree of global insulation deviation, respectively. When the maintenance personnel consult the graded early warning indicator, they can directly determine the driving source of the early warning without having to trace back the original data. The core basis of a certain level of early warning indicator shows that the degradation acceleration value has increased rapidly in the past 48 hours and the overall offset of the reference spectrum has approached the warning limit. Based on this, the maintenance personnel can quickly determine that the insulation status has the dual risks of accelerated degradation and overall offset, making the decision-making basis more sufficient.

[0054] Constrained graded warning labels are generated by verifying the cross-level jump constraints between adjacent warning labels. Insulation degradation is a gradual physical process; jumping directly from no warning to a level one warning requires certain degradation accumulation conditions. If these conditions are not met, the jump is judged as a false jump caused by data anomalies or transient interference. When the warning level difference between two adjacent labels in the graded warning label time series exceeds one level, cross-level jump detection is triggered. The detection method is to trace back the impedance degradation sequence entry that triggered the jump and check the confidence level of the corresponding degradation acceleration value. The confidence level comes from whether the number of lightly loaded frame segments in the continuous time period of the entry is sufficient and whether there are estimated labels. If the confidence level is insufficient, the cross-level jump is judged as an abnormal jump. The graded early warning indicator corresponding to an abnormal jump is downgraded to the previous level plus one level. The downgraded indicator is marked with a "to be verified" label, which is removed after data confirmation in the next complete injection cycle. If the confidence level is still insufficient, the downgrade remains. For example, a current transformer on a feeder in a nuclear power plant experienced persistently low confidence in its data acquisition due to contact aging; the corresponding early warning level remained at the conservative downgrade level until the sensor was maintained and its acquisition accuracy was restored. The constraint-graded early warning indicator merges the verified indicator with the downgraded indicator processed by the abnormal jump, forming a temporally continuous and physically reasonable sequence of early warning levels. The "to be verified" label in the constraint-graded early warning indicator is retained for manual review by maintenance personnel. The review conclusion is fed back to the monitoring device, updating the final early warning level record at the corresponding time.

[0055] The constraint grading early warning identifier is associated with the impedance degradation sequence to form an early warning signal frame. The associated frame assembly aligns the constraint grading early warning identifier with the corresponding impedance degradation sequence entry by timestamp and encapsulates them into a frame unit. Each frame unit simultaneously encapsulates the current early warning level, degradation acceleration value, current values ​​of R_ins and C_ins, and offset, covering the urgency of the risk, the intensity of degradation progression, real-time levels of insulation electrical parameters, and the cumulative deviation from the baseline. The centralized control system can fully assess the current insulation risk level without additional queries after receiving the frame. The generation frequency of the early warning signal frame is dynamically adjusted according to the constraint grading early warning identifier level. In the absence of an early warning, one frame is generated per low-frequency detection signal injection cycle. The generation frequency doubles during a Level 1 early warning. High-frequency generation ensures that maintenance personnel can obtain more frequent insulation status updates during a Level 1 early warning to track whether the degradation rate is still accelerating. In the impedance degradation sequence, the warning signal frame corresponding to the item to be verified has a verification flag appended to its frame header. This verification flag causes the centralized control system to automatically trigger a manual confirmation request upon receiving the frame, rather than directly driving the automatic response program. This avoids unnecessary maintenance operations triggered by warning flags with insufficient confidence, while preserving access to the frame data for maintenance personnel to review the original degradation characteristics before manual confirmation. The warning signal frame header synchronously carries the current overall offset of the insulation reference spectrum. When the overall offset approaches the historical insulation failure precursor threshold, the centralized control system generates additional emergency handling suggestions based on the warning signal frame. The suggestions are dynamically generated based on the core criteria in the graded warning flags. If the overall offset of a certain busbar section continues to rise and approaches the historical failure precursor threshold within three consecutive months, the centralized control system makes a comprehensive judgment based on the current degradation acceleration value and offset trend, and marks that busbar section as a priority for power outage inspection in the emergency handling suggestions appended to the warning signal frame.

[0056] Step S140: Degradation continuity detection is performed on the warning signal frame to identify impedance rise accumulation characteristics. Based on the impedance rise accumulation characteristics, a graded warning response strategy is activated to form warning trigger parameters. The warning trigger parameters are used to extract the impedance deviation increment of the impedance degradation sequence to construct an insulation correction scheme.

[0057] Specifically, the continuous degradation detection of the warning signal frame identifies the cumulative characteristics of impedance recovery. The continuous degradation detection focuses on the temporal changes in the R_ins value within the warning signal frame. Under warning conditions, the direction of R_ins change is not monotonically decreasing; there are periods of phased recovery. These recoverys may originate from three reasons: the fading of polarization effects, improvement in temperature and humidity, or a temporary improvement in the actual insulation condition. Recoverys caused by the fading of polarization effects typically terminate within a few hours, after which degradation restarts. Recoverys caused by improved temperature and humidity resume as the environment deteriorates. Improved insulation conditions cause R_ins to maintain the improved level for a longer period. The turning point where R_ins changes from decreasing to increasing in the warning signal frame timeline is defined as the recovery start point, and the moment when R_ins returns to a decrease or remains stable is defined as the recovery end point. The increment of R_ins between the start and end points is the recovery amplitude, and the time span is the recovery duration. These two values ​​together describe the scale and speed of a single recovery event. The impedance rise cumulative feature extracts the rise amplitude and duration of all rise events within the most recent observation window. The observation window is 24 hours under Level 1 warning and 48 hours under Level 2 warning. The average amplitude and median duration of multiple rise events within the observation window together constitute the two representative values ​​of the impedance rise cumulative feature. Rise events corresponding to the frame segments to be verified in the warning signal frame are not included in the amplitude accumulation calculation of the cumulative feature; including them would artificially inflate the representative value of the cumulative amplitude, leading to deviations in the judgment of false recovery. A larger representative value of the cumulative amplitude and a shorter average duration of the impedance rise cumulative feature indicate that although the rise amplitude is considerable, each rise is short-lived. This pattern is highly consistent with the false recovery caused by the polarization effect. In the initial stage of restarting after a nuclear power plant bus insulation shutdown and maintenance, a brief R_ins rise occurs due to the dissipation of the polarization effect. This rise terminates within a few hours as the polarization effect completely dissipates.

[0058] In some embodiments, the step of activating a graded early warning response strategy based on the impedance rise accumulation characteristics to form early warning trigger parameters includes: extracting the rise amplitude and rise duration based on the impedance rise accumulation characteristics; performing a cumulative rise false recovery determination on the rise amplitude and rise duration to generate a false recovery identifier; matching the false recovery identifier with the corresponding early warning response level to generate a graded response configuration; and determining the early warning trigger parameters according to the graded response configuration.

[0059] The rise amplitude and duration are extracted based on the cumulative impedance rise characteristics. The amplitude sequences of each rise event in the cumulative impedance rise characteristics are arranged in ascending order of occurrence time. The overall mean of the sequence is used as the representative value of the rise amplitude. Mean processing eliminates the excessive bias of a single abnormally large rise on the representative value, making the representative value of the rise amplitude more stably reflect the typical level of rise capability within the current observation window. The rise duration is determined by the median of the duration of each rise event. Compared with the mean, the median is more resistant to extreme short-term rises. Short-term rises usually correspond to measurement noise or acquisition jitter rather than actual improvement in insulation condition. Median estimation excludes the interference of such short-term disturbances on the representative value of the rise duration. When the number of rise events within the observation window of the cumulative impedance rise characteristics is less than 3, low-sample annotations are added to the representative values ​​of rise amplitude and rise duration. Under low-sample conditions, the particularity of a single rise event may cause the representative value to deviate from the true level. Low-sample annotations trigger a conservative adjustment of the judgment threshold during the false recovery judgment stage, avoiding overly certain false recovery conclusions based on insufficient samples. The recovery magnitude representative value is normalized within the context of the overall offset of the current insulation reference spectrum. The normalization method is to divide the recovery magnitude by the difference between the current R_ins and the normal reference R_ins. The normalized recovery magnitude representative value reflects the proportion of the recovery amount to the total current degradation. The higher the proportion, the more significant the recovery is relative to the cumulative degradation. A recovery occurring when a section of busbar R_ins has deteriorated to 60% of the normal value may have a considerable relative magnitude after normalization, even if the absolute magnitude is not large. The recovery magnitude and recovery duration representative values ​​constitute the output of this step.

[0060] For example, the step of cumulatively determining a false recovery based on the magnitude of the rebound and the duration of the rebound to generate a false recovery identifier includes: comparing the magnitude of the rebound with a polarization effect threshold to obtain an amplitude determination result; performing a polarization time window comparison on the duration of the rebound to obtain a duration determination result; performing a normalized joint determination based on the amplitude determination result and the duration determination result to form a comprehensive determination conclusion; and performing a polarization effect level assessment based on the comprehensive determination conclusion to generate a false recovery identifier.

[0061] The amplitude of the recovery was compared with the polarization effect threshold to obtain the amplitude determination result. The polarization effect threshold is derived from the polarization characteristic test data of similar medium-voltage bus insulation materials. It reflects the maximum possible R_ins recovery amplitude caused purely by the fading of the polarization effect. The recovery amplitude exceeding this threshold cannot be explained solely by the polarization effect, and the contribution of the actual insulation condition improvement needs to be considered. The normalized recovery amplitude representative value was compared with the polarization effect threshold. If it was below 50% of the threshold, it was determined that the amplitude fully conformed to the polarization effect characteristics. If it was between 50% and 100% of the threshold, it was determined that the amplitude partially conformed to the polarization effect characteristics. If it exceeded the threshold, it was determined that the amplitude was outside the range that the polarization effect could explain. When the amplitude determination result is that it fully conforms to the polarization effect characteristics, it indicates that the current recovery is highly consistent with the pseudo-recovery in terms of amplitude dimension. Under low load conditions, the polarization degree of the insulation material of nuclear power plant units is relatively deep. After returning to full load operation, the R_ins recovery amplitude caused by the fading of the polarization effect usually falls within 50% of the threshold. This characteristic is a typical pseudo-recovery amplitude pattern. The recovery magnitude of low-sample labeled values ​​shifts the comparison boundary by 10% in a conservative direction during threshold comparison. This shift makes the amplitude judgment result tend to reduce the confidence of false recovery, and to treat the amplitude judgment under insufficient sample conditions with a conservative attitude.

[0062] The duration of the recovery is compared with the polarization time window to obtain the duration determination result. The polarization time window is determined by the dielectric relaxation time constant of the insulating material. The larger the relaxation time constant, the slower the polarization effect fades, and the wider the corresponding polarization time window. The typical range of polarization time windows for commonly used insulating materials in medium-voltage busbars is obtained from the system parameter database based on the material model. The median recovery duration is compared with the polarization time window. If the duration falls within the polarization time window range, it is determined that the duration conforms to the characteristics of the polarization effect. If it is below the lower limit of the time window, it is determined that the recovery duration is too short, which may indicate measurement noise. If it exceeds the upper limit of the time window, it is determined that the duration is outside the interpretation range of the polarization effect. The latter suggests that the recovery may come from the improvement of the actual insulation condition rather than the fading of the polarization effect. The duration determination result is jointly evaluated with the amplitude determination result when the duration exceeds the polarization time window. If the amplitude also exceeds the polarization effect threshold, both features indicate true recovery. If the amplitude is still within the polarization effect range but the duration exceeds it, it is determined to be a partial false recovery. In a certain bus section, after the improvement of dry weather, the reduction of insulation surface leakage caused R_ins to rise continuously for more than 72 hours, which is far longer than the typical polarization time window. The duration determination result is determined to be beyond the range that the polarization effect can explain, indicating that this rise includes a component of real insulation condition improvement. The median duration of the rise with low sample annotation has low stability. The corresponding duration determination result is marked with low confidence, and the contribution weight of the low confidence duration determination result is reduced in the joint determination.

[0063] A comprehensive judgment conclusion is formed by normalizing and jointly judging the amplitude and duration results. The joint judgment maps the three judgment states of each result to a quantitative score within the range of 0 to 1. A full match of the amplitude effect characteristics corresponds to an amplitude score of 0.9, partial match to 0.5, and exceeding the interpretable range to 0.1. The duration judgment results are similarly mapped: falling within the polarization time window corresponds to a duration score of 0.9, below the lower limit to 0.6, and exceeding the upper limit to 0.2. The comprehensive judgment score S_joint is determined by the formula S_joint = w_A × S_amplitude + w_T × S_duration, where S_amplitude is the amplitude score, S_duration is the duration score, and w_A and w_T are weighting coefficients of 0.6 and 0.4 respectively. A higher amplitude weight reflects the dominant contribution of the recovery amplitude to the judgment of false recovery, and a higher S_joint indicates that the current recovery characteristics are more consistent with the characteristic combination of false recovery of the polarization effect. When a low-confidence annotation is added to the duration determination result, the corresponding S_duration is multiplied by a confidence reduction factor before weighting. After reduction, the contribution of S_duration decreases, and the overall S_joint is biased towards S_amplitude. The comprehensive determination conclusion is generated by the quantitative fusion of the amplitude determination result and the duration determination result. The S_joint value and the corresponding textual description of the possibility of false recovery are combined. An S_joint higher than 0.7 is considered a high probability of false recovery, between 0.4 and 0.7 is considered a partial false recovery, and below 0.4 is considered a high probability of true recovery. When the comprehensive determination conclusion is a high probability of true recovery, it indicates that the current warning level can be considered for downgrading, but it is not automatically downgraded.

[0064] Based on the comprehensive judgment conclusion, a polarization effect level assessment is implemented to generate a false recovery label. The polarization effect level assessment maps the S_joint value in the comprehensive judgment conclusion to a grading system of the degree of polarization effect influence. A high probability of false recovery corresponds to a strong polarization effect level, a partial false recovery corresponds to a medium polarization effect level, and a relatively high probability of true recovery corresponds to a weak polarization effect level. The strong polarization effect level comes from the judgment branch in the comprehensive judgment conclusion where S_joint is higher than 0.7. Under the strong polarization effect level, the current rise is considered to be a reversible change without substantial significance. After the nuclear power plant unit resumes normal operating load, the accelerated release of polarization charge causes a temporary rise in R_ins. This type of rise naturally terminates after the polarization effect completely dissipates and resumes its downward trend in the direction of degradation. Therefore, the warning level remains unchanged, and a polarization interference label is added to the false recovery label. The polarization interference label informs the centralized control system that the currently observed rise does not represent a reduction in insulation risk, and maintenance personnel should not postpone maintenance plans based on this. Under a weak polarization effect level, the current recovery is considered to include a genuine insulation improvement component. A genuine improvement label is added to the false recovery identifier. This genuine improvement label triggers dynamic correction of the normal reference point for the insulation reference spectrum. The correction method involves incorporating the current R_ins mean into the rolling update of the reference point, preventing historical degradation states from permanently lowering the reference value. The false recovery identifier simultaneously carries the S_joint value and the polarization effect level label. The S_joint value is used for precise matching of the warning response intensity during the level response configuration phase. The polarization effect level label allows maintenance personnel to quickly understand the physical nature of the current recovery. If an R_ins recovery under a Level 1 warning is determined to be a strong polarization effect level, maintenance personnel can check the false recovery identifier to understand that the current recovery does not represent the elimination of insulation risk, and the established maintenance plan remains unaffected.

[0065] The level response configuration is generated based on the matching of false recovery identifiers with corresponding early warning response levels. For false recovery identifiers, the strong polarization effect level corresponds to no downgrade of the early warning response; the current early warning level is maintained and the response action remains active. The medium polarization effect level corresponds to a half-level downgrade of the early warning response; that is, under medium polarization conditions, the response intensity of a Level 1 early warning is adjusted to an intermediate level between Level 1 and Level 2, with the half-level downgrade reflecting the limited mitigation of the severity of the early warning due to partial false recovery. The weak polarization effect level allows for a one-level downgrade of the early warning response for evaluation, but downgrading requires the additional condition that the overall offset of the insulation reference spectrum does not further increase. The level response configuration determines the strategy for extending or shortening the current observation window while matching the early warning response level. Under the strong polarization effect level, the observation window is shortened to increase the acquisition density and track the degradation progress, ensuring that the acceleration signal after the polarization effect fades can be captured in a timely manner. Under the weak polarization effect level, the observation window is appropriately extended to accumulate more recovery event samples. When the R_ins of a certain bus section continues to rise during the dry season, downgrading can only be considered after confirming improved stability through multiple recovery event statistics. In cases where the false recovery indicator has a low sample size, the response level configuration does not lower the warning level. Under low sample conditions, the confidence level for false recovery is insufficient. A conservative strategy requires maintaining the current warning level until sufficient samples are available for reassessment. This conservative strategy reflects a zero-tolerance attitude towards insulation failure risks in nuclear power plant insulation safety management. The response level configuration consists of the current warning level, the response intensity level, and the direction of the observation window adjustment. The warning level locks in the basic response level, the response intensity level refines the specific levels of collected and transmitted parameters, and the direction of the observation window adjustment determines the duration and adjustment strategy of subsequent acquisition cycles. The combination of these three elements forms a complete response configuration scheme.

[0066] The warning trigger parameters are determined based on the graded response configuration. The warning response level and response intensity level in the graded response configuration are mapped to three specific parameters: acquisition frequency, injected signal amplitude, and data transmission trigger interval. Under the Level 1 warning maintenance state, the acquisition frequency is set to the highest level, and the injected signal amplitude is adaptively adjusted to the lowest injection level that meets the signal-to-noise ratio requirements based on the current load level to reduce interference with system operation. The warning trigger parameters corresponding to a half-level reduction in response are determined by linear interpolation between the Level 1 and Level 2 warning parameters. The interpolation result ensures a smooth parameter transition, avoiding abrupt changes in acquisition parameters that could affect data continuity when the response level is adjusted. When the observation window in the graded response configuration is shortened, the frame interval time parameter in the warning trigger parameters is simultaneously compressed. The shortened frame interval increases the frequency of warning signal frame generation, allowing the centralized control system to obtain more frame data in a short time to track the latest dynamics of degradation progress. In a nuclear power plant, after the strong polarization effect interference subsides, the degradation acceleration value of a certain bus section remains high. At this time, the acquisition frequency in the warning trigger parameters remains at the highest level, and the frame interval is compressed to 50% of the normal value, ensuring that the fine characteristics of the degradation acceleration stage are fully recorded. After the warning triggering parameter takes effect, it triggers the extraction of impedance deviation increment of the impedance degradation sequence. The deviation increment extraction is based on the deviation calculation period specified by the current warning triggering parameter, and extracts the change of R_ins in the corresponding time period from the impedance degradation sequence. The change is multiplied by the increment amplification factor in the warning triggering parameter to obtain the impedance deviation increment.

[0067] An insulation correction scheme is constructed by extracting the impedance deviation increment from the impedance degradation sequence using early warning trigger parameters. The deviation increment extraction uses the deviation calculation period specified by the early warning trigger parameters as the time window. Within this period, the measured R_ins values ​​of each time period in the impedance degradation sequence are extracted from the deviation sequence. The deviation increment per unit time is obtained through successive difference calculations of the deviation sequence. A positive increment indicates that the deviation is expanding, i.e., the degradation is continuing, while a negative increment indicates that the deviation is narrowing, i.e., the degradation in the current time period is slowing down. The deviation increment corresponding to the time period of the high-level degradation event in the impedance degradation sequence is multiplied by the increment amplification factor in the early warning trigger parameters. The amplification factor reflects that the rate of consumption of insulation safety margin by the deviation increment per unit time is faster in the high-level degradation stage. The insulation correction scheme uses the temporal distribution of the impedance deviation increment as input. Time periods where the increment is consistently positive and the absolute value exceeds the correction trigger threshold are identified as the time periods requiring correction intervention. The correction trigger threshold is determined by the level response level in the early warning trigger parameters; higher response levels correspond to lower correction trigger thresholds. The recommended maintenance actions for the intervention period in the insulation correction plan are determined in three levels based on the magnitude of the deviation increment. For small increments, it is recommended to shorten the next inspection interval; for medium increments, it is recommended to conduct online partial discharge detection; and for large increments, it is recommended to apply for a planned power outage for maintenance. These three levels of recommended actions correspond to the handling levels in the nuclear power plant insulation maintenance procedures. For false recovery indicators, the negative increment corresponding to the actual improvement period is marked with a natural recovery indicator. The natural recovery period does not trigger recommended maintenance actions, ensuring that the intervention recommendations of the insulation correction plan only apply to the period where the degradation is confirmed to be progressing.

[0068] Step S150: Based on the insulation correction scheme, periodic insulation inspection is performed to form an impedance trend curve. An adaptive transmission delay assessment based on the early warning level is performed on the impedance trend curve to determine the data uploading node. Based on the data uploading node, centralized control response verification is performed to generate an insulation monitoring report.

[0069] Specifically, an impedance trend curve is generated based on the insulation correction scheme through periodic insulation inspections. The inspection frequency is determined according to the recommended maintenance actions during the correction intervention period in the insulation correction scheme. Higher inspection frequencies correspond to periods where shorter review intervals are recommended, and additional specialized testing tasks are added to the standard inspections during periods where partial discharge detection is recommended. Inspection data from these two types of periods are prioritized for inclusion in the impedance trend curve construction to ensure the accuracy of trend description during high-risk periods. The inspection frequency for periods marked as natural recovery in the insulation correction scheme is executed according to the normal cycle, without reducing the inspection density due to improved insulation condition during the recovery period, ensuring that the trend curve can promptly capture the turning point when degradation resumes after the recovery ends. The impedance trend curve uses the inspection time as the horizontal axis and the normalized values ​​of R_ins and C_ins as the vertical axis. The R_ins and C_ins values ​​from each inspection's impedance calculation unit are connected sequentially to form a continuous curve reflecting the long-term evolution of the insulation condition. The slope of the curve directly reflects the rate increase / decrease trend in the degradation rate sequence. A slope change from negative to positive indicates a recovery segment, and a slope change from positive to negative indicates a degradation restart segment. In the insulation correction scheme, correction intervention labels are added to the impedance trend curve segments corresponding to the correction intervention period. The correction intervention labels record the types of maintenance actions implemented during that period. The comparison of the curve slopes before and after the maintenance actions directly reflects the effect of the intervention measures on the degradation progress. After a section of busbar was partially discharged and targeted cleaning was carried out, the slope of R_ins on the corresponding impedance trend curve changed from negative to flat after the correction intervention label. Before the cleaning treatment, the slope decreased by about 0.8% per hour. After the treatment, the slope remained within ±0.1% for three consecutive inspection cycles. The intervention effect is directly quantified in the curve shape.

[0070] In some embodiments, the step of performing adaptive transmission delay assessment based on warning level to determine the data uploading node on the impedance trend curve includes: extracting a warning level identifier from the impedance trend curve; performing a delay threshold classification mapping based on the warning level identifier to obtain a level delay threshold; performing a reverse-driven correction on the level delay threshold and the impedance trend curve to generate a corrected delay threshold; and determining the data uploading node by comparing the corrected delay threshold with the impedance trend curve.

[0071] The warning level identifier is extracted from the impedance trend curve. This identifier is embedded in the attribute fields of each data point on the impedance trend curve and is recorded synchronously with the corresponding R_ins and C_ins values. The current warning level identifier is read directly from the latest data point without needing to backtrack to historical data points, ensuring extraction efficiency unaffected by the amount of data on the curve. For adjacent data points on the impedance trend curve, the warning level switching time is recorded simultaneously. This switching time serves as the trigger for dynamic threshold adjustment during the time-delay threshold grading mapping stage. A higher level triggers threshold tightening, while a lower level triggers threshold widening. This dynamic threshold adjustment ensures that the data transmission strategy responds synchronously to changes in the warning status. The warning level identifier for the corrected intervention-annotated data points is appended with an intervention period annotation during extraction. During maintenance intervention, the insulation status may fluctuate briefly due to operational disturbances. The intervention period annotation triggers threshold widening during the time-delay threshold mapping stage, preventing unnecessary high-frequency data transmissions driven by operational disturbances. The start and end times of the intervention period annotation are strictly aligned with the timestamps of the maintenance operation records to ensure the widening window accurately covers the intervention period. After the warning level indicator of the latest data point of the impedance trend curve is extracted, the overall offset of the current insulation reference spectrum is read synchronously. The larger the offset, the closer the insulation is to the precursor of failure. When a section of the busbar in a nuclear power plant has been in continuous operation for eight years, the overall offset of the insulation reference spectrum has accumulated to 35% of the initial reference value. At this time, the offset value read synchronously is used as the basis for calculating the correction coefficient in the subsequent reverse drive correction. This makes the data transmission frequency of the busbar section with a large offset more strictly constrained when an upward trend appears, so as not to excessively reduce the transmission density due to local rebound.

[0072] The time delay threshold is obtained by hierarchical mapping based on the warning level identifier. The mapping is performed according to the level information in the warning level identifier. The real-time requirements for insulation monitoring differ significantly under different warning levels. In the absence of a warning, the batch transmission of inspection data after a complete inspection cycle can meet the monitoring requirements. In the first-level warning state, continuous transmission at shorter intervals is required to support the dynamic decision-making of the centralized control system. The hierarchical mapping transforms this difference in requirements into a quantified time delay parameter. The level delay threshold is set to the maximum level in the absence of a warning state, compressed to 50% of the level in the second-level warning state, and compressed to 20% of the level in the first-level warning state. After a section of the busbar in a nuclear power plant enters the first-level warning state, the centralized control system can receive the latest impedance trend curve data every few minutes. Based on this, maintenance personnel continuously assess whether to initiate an emergency power outage maintenance procedure. When an intervention period is added to the warning level identifier, the level delay threshold is relaxed by one level based on the corresponding warning level. After the intervention period ends, it automatically reverts to the original warning level threshold. The switching between relaxation and restoration is based on the time of removal of the intervention period label, ensuring that the adjustment of the time delay threshold is strictly synchronized with the progress of maintenance operations. When the time interval between the level switching moment and the adjacent data point in the impedance trend curve is too short, the level delay threshold is smoothly switched using a linear transition method at the switching position. When a certain bus section experiences a level change of two levels to one level and then down two levels in two hours, the linear transition makes the level delay threshold adjust gradually between the three switching. The transmission interval parameter received by the centralized control system changes smoothly, and there is no oscillation phenomenon of repeated switching between high and low frequency modes on the acquisition side.

[0073] A corrected delay threshold is generated by reversing the delay threshold and impedance trend curve. The reversal point where the slope of R_ins in the impedance trend curve changes from negative to positive indicates that the insulation impedance is showing an upward trend. If the warning level is still high when the reversal occurs, there is a contradiction between the two information. A correction mechanism is needed to balance the high-frequency transmission requirements under high warning levels with the reasonable need to appropriately reduce the transmission frequency under the upward trend. The correction mechanism enables the data transmission frequency to simultaneously sense both the warning level and the insulation recovery trend information. At the reversal point, the normalized value of the overall offset of the insulation reference spectrum is used as the constraint coefficient M_reverse. The modified time delay threshold T_modified = T_level × (1 + γ × (1 − M_reverse)), where T_level is the original level time delay threshold in seconds; T_modified is the modified time delay threshold in seconds; γ is the reversal intensity coefficient, which is determined by the normalization of the slope change amplitude. The larger the overall offset, the larger M_reverse, and the smaller (1 − M_reverse) is, the more constrained the relaxation range is, ensuring that the recovery under the background of overall insulation degradation will not excessively reduce the data transmission frequency. In the false recovery flag, the reversal point corresponding to the strong polarization effect level does not undergo reversal-driven correction. The correction delay threshold maintains its original value during the strong polarization effect period, ensuring that the data transmission frequency is not unreasonably suppressed due to polarization interference during the false recovery period. When the polarization effect subsides and R_ins briefly rises during the initial restart of the nuclear power plant unit, the corresponding reversal point is identified as a false recovery and the delay threshold relaxation is not triggered. The centralized control system continues to receive the latest data at the high frequency corresponding to the first-level warning. The correction delay threshold is evaluated on a rolling basis after new data points are added to the impedance trend curve. It is recalculated when the slope reversal disappears or the warning level changes to maintain dynamic correspondence with the latest insulation state.

[0074] The data upload node is determined by comparing the correction delay threshold with the impedance trend curve. Tracing back from the latest data point on the impedance trend curve, the most recent historical data point exceeding the correction delay threshold is determined as the starting node for this data upload. All data between the starting node and the latest data point constitutes the data packet for this upload. The warning level identifier and R_ins value of each data point within the data packet are fully preserved to support the historical state backtracking of the centralized control system. Data points on the impedance trend curve where the warning level changes are forcibly set as data upload nodes. Data at the moment of level change must be uploaded immediately without waiting for the correction delay threshold to be triggered. When a bus section is upgraded from a level 2 warning to a level 1 warning, the centralized control system can receive data containing the latest degradation characteristics within seconds of the upgrade, ensuring that the centralized control system is updated immediately upon the change in warning level. The data upload node for the time period of the correction intervention-marked data point is further delayed until after the maintenance operation is completed. This avoids confusion between intermediate state data during the intervention period and steady-state data after the intervention, ensuring that each uploaded data received by the centralized control system corresponds to a complete insulation status acquisition cycle. The data packet header embeds a warning level identifier, a correction delay threshold, and a data time range. Based on this, the centralized control system can directly determine the urgency of the current insulation risk, the timeliness requirements of this transmission, and the start and end interval of the data without parsing the data content. When the nuclear power plant's centralized control system receives a Level 1 warning data packet, it immediately initiates the insulation status assessment process based on the warning level identifier in the header. On-duty personnel can obtain a complete summary of the current deterioration acceleration value and offset within 30 seconds of the data packet's arrival, without waiting for the complete data parsing to be completed.

[0075] The centralized control system performs response verification based on the data transmission node to generate an insulation monitoring report. The centralized control response verification uses the response status of the data packet corresponding to the data transmission node after it is transmitted to the centralized control system as the verification basis. After receiving the complete data packet, the centralized control system returns a response message containing three items: a reception confirmation code, the data integrity verification result, and the reception time. If the data packet corresponding to the data transmission node does not receive a response within the specified time limit, a retransmission is triggered. The retransmission interval is dynamically adjusted according to the warning level indicator, with the shortest retransmission interval under the first-level warning state. If there is still no response after more than three retransmissions, the system switches to the backup communication link and records the communication failure event. The communication failure event is separately marked in the insulation monitoring report. The system automatically switches to the backup link when thunderstorms cause the main link signal to attenuate. Maintenance personnel check the report to see if there are any changes in the warning level or accelerated degradation events during the fault period, eliminating potential risks masked by data interruption. If the data integrity verification fails in the response message, the data packet is retransmitted, and a verification failure mark is added to the retransmitted data packet. The verification failure mark is recorded in the local log for subsequent link quality analysis. Two consecutive verification failures trigger a self-test of the local acquisition device to eliminate systemic errors in the data packaging process. The insulation monitoring report is generated based on the data packets that have passed the centralized control response verification. The report comprehensively presents the key evolution characteristics of the impedance trend curve, the current level of the constraint classification early warning, the insulation correction plan, the maintenance actions recommended, and the communication quality assessment. It covers the insulation status trend, risk level, handling suggestions, and transmission reliability, providing complete decision support information for the centralized control system and maintenance personnel.

[0076] To implement the above-described method embodiments, a method for online monitoring of the insulation of medium-voltage busbars in nuclear power plants is provided to achieve the corresponding functions and technical effects. See also... Figure 2 , Figure 2 This diagram illustrates a structural block diagram of an online monitoring system 200 for the insulation of a medium-voltage busbar in a nuclear power plant, according to an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The online monitoring system 200 for the insulation of a medium-voltage busbar in a nuclear power plant, according to an embodiment of this application, includes:

[0077] Signal injection module 201 is used to acquire bus operation data of medium voltage ungrounded system, perform low frequency detection signal injection to identify insulation response sections based on the bus operation data, and perform impedance sensitivity priority acquisition and framing based on the insulation response sections to form insulation response frames;

[0078] Feature extraction module 202 is used to perform adaptive filtering matching to form a filtering configuration based on the insulation response frame, perform temperature drift correction and frequency domain feature weighting extraction on the filtering configuration to form weighted feature components, and construct an impedance calculation unit through the weighted feature components;

[0079] The degradation assessment module 203 is used to map the degradation features of the impedance calculation unit to form an insulation reference spectrum, perform degradation rate and degradation acceleration assessment based on the insulation reference spectrum to generate an impedance degradation sequence, and implement graded marking for the impedance degradation sequence to generate a warning signal frame.

[0080] The early warning response module 204 is used to detect and identify impedance rise accumulation characteristics of the early warning signal frame, activate a graded early warning response strategy based on the impedance rise accumulation characteristics to form early warning trigger parameters, and use the early warning trigger parameters to extract impedance deviation increments from the impedance degradation sequence to construct an insulation correction scheme.

[0081] The uploading verification module 205 is used to perform insulation periodic inspections based on the insulation correction scheme to form an impedance trend curve, perform adaptive transmission delay assessment based on the early warning level on the impedance trend curve to determine the data uploading node, and perform centralized control response verification to generate an insulation monitoring report based on the data uploading node.

[0082] The above-described online monitoring system 200 for the insulation of medium-voltage busbars in nuclear power plants can implement the online monitoring method for the insulation of medium-voltage busbars in nuclear power plants described in the above-described method embodiments. The options described in the above method embodiments are also applicable to this embodiment and will not be detailed here. The remaining contents of this application's embodiments can be referred to the contents of the above method embodiments, and will not be repeated in this embodiment.

[0083] The purpose of the above embodiments is to reproduce and derive the technical solution of the present invention by way of example, and to fully describe the technical solution, purpose and effect of the present invention. The purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosure of the present invention, and not to limit the scope of protection of the present invention.

Claims

1. A method for online monitoring of the insulation of medium-voltage busbars in nuclear power plants, characterized in that, include: Obtain busbar operation data of medium-voltage ungrounded system, perform low-frequency detection signal injection to identify insulation response sections based on the busbar operation data, and form insulation response frames based on impedance sensitivity priority acquisition frames based on the insulation response sections. Adaptive filtering matching is performed based on the insulation response frame to form a filtering configuration. Temperature drift correction and frequency domain feature weighting extraction are performed on the filtering configuration to form weighted feature components. An impedance calculation unit is constructed through the weighted feature components. The impedance calculation unit is subjected to degradation feature mapping to form an insulation reference spectrum. Based on the insulation reference spectrum, degradation rate and degradation acceleration are evaluated to generate an impedance degradation sequence. The impedance degradation sequence is graded and marked to generate a warning signal frame. The degradation continuity detection of the warning signal frame is performed to identify the impedance rise accumulation characteristics. Based on the impedance rise accumulation characteristics, a graded warning response strategy is activated to form warning trigger parameters. The warning trigger parameters are used to extract the impedance deviation increment of the impedance degradation sequence to construct an insulation correction scheme. Based on the insulation correction scheme, insulation periodic inspections are performed to form an impedance trend curve. An adaptive transmission delay assessment based on the early warning level is implemented on the impedance trend curve to determine the data uploading node. Based on the data uploading node, centralized control response verification is performed to generate an insulation monitoring report.

2. The method according to claim 1, characterized in that, The process of prioritizing impedance sensitivity acquisition and framing based on the insulation response segment to form an insulation response frame includes: The sensitivity ranking results were obtained by performing a weighted distribution analysis of the response amplitude deviation of the insulation response segment. Based on the sensitivity ranking results, priority acquisition sequence is generated by prioritizing the acquisition process. The priority acquisition sequence is dynamically mapped based on the warning level to form frame grouping parameters; The insulation response segment is prioritized and framed according to the frame grouping parameters to form an insulation response frame.

3. The method according to claim 1, characterized in that, The step of performing temperature drift correction and frequency domain feature weighting extraction on the filter configuration to form weighted feature components includes: Based on the aforementioned filtering configuration, the current ambient temperature sample value and the load current sample value are extracted; Based on the ambient temperature sampling value and the load current sampling value, the ambient temperature drift and self-heating temperature drift are calculated separately to obtain the comprehensive temperature drift correction coefficient; Based on the comprehensive temperature drift correction coefficient and the frequency domain characteristics of the filter configuration, deviation compensation is performed to generate compensated frequency domain characteristics; Based on the compensated frequency domain features, a response intensity weighted extraction is performed to form weighted feature components.

4. The method according to claim 1, characterized in that, The process of generating an impedance degradation sequence by performing degradation rate and degradation acceleration assessment based on the insulation reference spectrum includes: Extract the impedance change records for continuous time periods from the insulation reference spectrum; The degradation rate sequence is obtained by performing a rate of change analysis on the impedance change records during light load periods in the continuous time period. The degradation rate sequence is subjected to rate stationary period abrupt change identification to obtain degradation acceleration value; Based on the degradation acceleration value, a threshold comparison and sorting process is performed to form an impedance degradation sequence.

5. The method according to claim 1, characterized in that, The step of generating a warning signal frame by performing graded marking on the impedance degradation sequence includes: The impedance degradation sequence is divided into degradation acceleration level intervals to obtain the level division results; Based on the classification results, early warning thresholds are matched to generate graded early warning identifiers; The hierarchical early warning identifier is subjected to adjacent level cross-level jump constraint verification to generate a constraint hierarchical early warning identifier; The constraint classification early warning identifier is associated with the impedance degradation sequence and framed to form an early warning signal frame.

6. The method according to claim 1, characterized in that, The method for generating early warning trigger parameters based on the impedance rise accumulation characteristic to activate the graded early warning response strategy includes: Based on the accumulated characteristics of impedance recovery, the recovery amplitude and recovery duration are extracted; The cumulative recovery magnitude and recovery duration are used to determine whether a false recovery has occurred and generate a false recovery identifier. Based on the false recovery identifier, a level response configuration is generated by matching the corresponding early warning response level; The warning trigger parameters are determined based on the level response configuration.

7. The method according to claim 1, characterized in that, The step of performing adaptive transmission delay assessment based on early warning level on the impedance trend curve to determine the data uploading node includes: Extract the warning level indicator from the impedance trend curve; Based on the warning level identifier, a time delay threshold is mapped to obtain the level time delay threshold; The level delay threshold and the impedance trend curve are reverse-driven to generate a corrected delay threshold; The data uploading node is determined by comparing the corrected delay threshold with the impedance trend curve.

8. The method according to claim 4, characterized in that, The step of identifying rate stationary abrupt changes in the degradation rate sequence to obtain degradation acceleration values ​​includes: The degradation rate sequence is statistically analyzed to generate a difference distribution feature. Based on the difference distribution characteristics, accelerate abrupt transition sections exceeding the rate reference boundary are identified, and abrupt transition section identifiers are generated. Acceleration fitting analysis is performed on the identified abrupt change segments to generate acceleration fitting parameters; Based on the acceleration fitting parameters, the fitting residuals are converged and filtered to generate degraded acceleration values.

9. The method according to claim 6, characterized in that, The step of cumulatively determining whether a false recovery has occurred based on the magnitude of the rebound and the duration of the rebound, and generating a false recovery identifier, includes: The amplitude of the rebound is compared with the polarization effect threshold to obtain the amplitude determination result; The duration of the rebound is compared using a polarization time window to obtain the duration determination result; A comprehensive judgment conclusion is formed by normalizing and jointly judging the amplitude determination result and the duration determination result. Based on the comprehensive judgment conclusion, a polarization effect level assessment is performed to generate a false recovery identifier.

10. An online monitoring system for the insulation of medium-voltage busbars in nuclear power plants, characterized in that, include: The signal injection module is used to acquire the bus operation data of the medium-voltage ungrounded system, perform low-frequency detection signal injection to identify the insulation response section based on the bus operation data, and perform impedance sensitivity priority acquisition and framing based on the insulation response section to form an insulation response frame. The feature extraction module is used to perform adaptive filtering matching based on the insulation response frame to form a filtering configuration, perform temperature drift correction and frequency domain feature weighting extraction on the filtering configuration to form weighted feature components, and construct an impedance calculation unit through the weighted feature components; The degradation assessment module is used to map the degradation features of the impedance calculation unit to form an insulation reference spectrum, perform degradation rate and degradation acceleration assessment based on the insulation reference spectrum to generate an impedance degradation sequence, and implement graded marking for the impedance degradation sequence to generate a warning signal frame. The early warning response module is used to detect and identify the impedance rise accumulation characteristics of the early warning signal frame, activate the hierarchical early warning response strategy based on the impedance rise accumulation characteristics to form early warning trigger parameters, and use the early warning trigger parameters to extract the impedance deviation increment of the impedance degradation sequence to construct an insulation correction scheme. The uploading verification module is used to perform insulation periodic inspections based on the insulation correction scheme to form an impedance trend curve, perform adaptive transmission delay assessment based on the early warning level on the impedance trend curve to determine the data uploading node, and perform centralized control response verification to generate an insulation monitoring report based on the data uploading node.

Images