A security isolation linkage control method and system for a strong and weak electricity integrated system

Abstract

The application relates to the technical field of strong and weak electric integration control, and discloses a safety isolation linkage control method and system for a strong and weak electric integration system. The method comprises the following steps: collecting strong and weak electric system physical layer error code events and strong electric side electrical transient data; establishing a timing error code correlation array through time stamp alignment and event correlation coding; adaptively dividing the strong electric side transient data into frequency windows and identifying the phase coupling based on the array directivity query result to establish a frequency spectrum coupling phase chain; matching the two cross-domain isomorphism and constructing a strong and weak electric coupling causal grid through causal sequential reasoning; predicting a safety margin decay trajectory and feeding forward adjusting a linkage type isolation parameter based on the causal grid; the problems that a weak electric communication error code microscopic mode caused by strong electric disturbance is covered by a protocol error correction mechanism, the existing monitoring cannot capture, and the control lacks causal mapping are solved; and closed-loop control from sub-health state perception to prior active regulation is realized.

Description

Technical Field

[0001] This invention relates to the field of integrated strong and weak current control technology, and in particular to a safety isolation and linkage control method and system for integrated strong and weak current systems. Background Technology

[0002] Integrated high-voltage and low-voltage systems are widely used in buildings, factories, and other scenarios. They achieve energy optimization and intelligent operation and maintenance by integrating high-voltage power distribution and low-voltage control functions. Existing technologies mainly rely on forward error correction and automatic retransmission mechanisms of communication protocols to ensure data transmission, coupled with insulation detection devices, electromagnetic shielding structures, and end-to-end latency monitoring. They focus on link connectivity and macroscopic bit error rate statistics. The core is to trigger alarms or isolation actions after a fault becomes apparent. The technical characteristics are biased towards passive response and macroscopic state judgment.

[0003] In actual data center operation and maintenance, engineers often encounter the following situation: after long-term operation of integrated strong and weak current cabinets, under conditions such as load switching or inverter modulation, the weak current side occasionally experiences control command execution lag or feedback data deviation, but the communication link does not trigger an interruption alarm, and routine bit error rate statistics do not show any abnormalities. Multiple troubleshooting attempts fail to reproduce and locate the problem. The essence of this problem is that the common-mode current generated by power fluctuations on the strong current side flows through the grounding impedance, forming a voltage disturbance. This disturbance is coupled to the weak current communication line through parasitic capacitance and mutual inductance, causing the physical layer bit error rate to exhibit periodic and sudden micro-mode changes. The error correction mechanism of the communication protocol masks this change. Existing monitoring only focuses on macroscopic indicators and does not analyze fine-grained characteristics such as the timing distribution of bit errors and codeword correlation. It also lacks a cross-domain mapping model between strong current disturbances and weak current communication quality degradation. This problem can lead to the failure to identify early hidden dangers such as aging of insulation materials and deterioration of contact resistance on the high-voltage side. Under critical operating conditions such as grid connection and off-grid switching, it may cause the protection action timing disorder. At the same time, occasional faults are difficult to trace, forcing operation and maintenance to expand the scope of shutdown investigation and increase the unplanned downtime of equipment and maintenance costs. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, this invention provides a safety isolation and linkage control method for integrated strong and weak current systems, comprising:

[0005] S1: Collect the original differential signal sequence, original ideal codeword sequence and original transient electrical data of the strong and weak current integrated system, and establish a timing error correlation array through timestamp alignment and event correlation coding;

[0006] S2, based on the directional query results of the timing error correlation array, adaptive window segmentation and phase coupling identification are performed on the transient data of the high-voltage side to establish a spectrum-coupled phase chain;

[0007] S3, cross-domain isomorphic matching and causal sequential inference are performed between the timing error correlation array and the spectrum-coupled phase chain to construct a strong-weak electrical coupled causal lattice;

[0008] S4, based on strong and weak electrical coupling causal lattice, performs safety margin decay trajectory prediction and linkage isolation parameter feedforward adjustment.

[0009] Furthermore, the step of establishing the timing error correlation array includes:

[0010] Based on the original sequence of differential signals in weak current communication and the original ideal codeword sequence, the initial set of bit error events is obtained by decoding and comparing with the bit error monitoring probe.

[0011] Based on the initial set of bit error events and the original set of electrical transient data from the high-voltage side, a unified time reference sequence is generated using a unified time base generation method, resulting in a time-aligned set of bit error events and a time-aligned set of electrical transient data from the high-voltage side.

[0012] Based on the time-aligned bit error event set and the high-voltage side electrical transient alignment data set, a bit error event base matrix is ​​generated through an event correlation coding method;

[0013] Based on the error event matrix, a set of higher-order correlation terms is generated using a higher-order correlation term construction method.

[0014] Based on the fundamental matrix of bit error events and the set of higher-order correlation terms, a timing bit error correlation array is constructed through an array fusion generation method, and the timing bit error correlation array and the set of electrical transient alignment data from the power supply side are output.

[0015] Furthermore, the error monitoring probe decoding comparison method includes:

[0016] Using the original sequence of differential signals for weak current communication as input, the physical layer signal flip point sequence is extracted by an edge detection algorithm with adaptive threshold, and a physical layer sampling clock sequence is constructed based on the physical layer signal flip point sequence.

[0017] Using the physical layer sampling clock sequence and the original sequence of the weak current communication differential signal as joint inputs, a decision bit sequence is generated through symbol decision;

[0018] Using the decision bit sequence and the original ideal codeword sequence as joint inputs, an error flag sequence is generated by bit-by-bit comparison and forward error correction flag extraction rules.

[0019] Using the error flag sequence and the physical layer sampling clock sequence as joint inputs, an initial set of error events is generated through error event encapsulation rules.

[0020] Furthermore, the method for constructing higher-order association terms includes:

[0021] Using the timestamp column in the bit error event matrix as input, a candidate period set for the bit error time series is obtained through a periodic candidate frequency search algorithm;

[0022] Using the candidate period set and the timestamp column in the error event matrix as joint inputs, the phase distribution sequence under each candidate period is obtained through phase folding operation;

[0023] Using the phase distribution sequence as input, a set of periodic clustering results is generated through phase clustering and burstiness assessment algorithms;

[0024] Using the set of periodic clustering results as input, the codeword distance distribution entropy value of each phase cluster is obtained through codeword distance statistics and information entropy calculation methods;

[0025] Using the periodic clustering result set and codeword distance distribution entropy as joint inputs, a mode weight value is generated for each phase cluster according to the mode weight calculation rule;

[0026] Using all higher-order related terms as input, a set of higher-order related terms is generated through set construction operations and combined with a first preset threshold for filtering.

[0027] Furthermore, the step of establishing the spectral coupling phase chain includes:

[0028] Using a timing error correlation array as input, a set of high-order correlation term pointing query results is generated through a high-order correlation term pointing query method;

[0029] Using the set of high-order correlation term directional query results and the set of electrical transient alignment data from the power supply side as joint inputs, a set of frequency window segmentation results is generated through an adaptive frequency window segmentation method.

[0030] Using the frequency window segmentation result set as input, an instantaneous phase trajectory set is generated through an instantaneous phase extraction method;

[0031] Using the instantaneous phase trajectory set as input, a set of frequency point phase coupling relationships is generated through a phase-locked detection method;

[0032] Using the set of frequency-point phase coupling relationships as input, a spectral coupling phase chain is generated through a chain structure construction method.

[0033] Furthermore, the higher-order related item pointing query method includes:

[0034] Using the set of higher-order correlation items in the timing error correlation array as input, the top several higher-order correlation items with the highest mode weight values ​​are selected by the mode weight sorting algorithm to obtain a subset of candidate higher-order correlation items.

[0035] Taking the set of error event indices for each higher-order correlation item in the candidate higher-order correlation item subset as input, the corresponding error event timestamp subsequence and codeword index subsequence are obtained by retrieving from the time-series error correlation array;

[0036] Using the timestamp subsequence of bit error events as input, the implicit period estimate is obtained through time interval statistics and frequency estimation algorithms;

[0037] Using the error event timestamp subsequence, codeword index subsequence, pattern weight value, and implicit period estimation value as joint inputs, a set of high-order related item directional query results is generated through result encapsulation rules.

[0038] Furthermore, the step of constructing a strong-weak coupled causal lattice includes:

[0039] Using a timing error correlation array as input, a set of error mode feature vectors is generated through an error mode feature extraction method.

[0040] Using the spectral coupling phase chain as input, a set of perturbation coupling feature vectors is generated through a perturbation coupling feature extraction method;

[0041] Using the set of error pattern feature vectors and the set of perturbation coupling feature vectors as joint inputs, a set of candidate causal relationship pairs is generated through a cross-domain isomorphic matching method.

[0042] Using a set of candidate causal relationship pairs as input, a strongly and weakly coupled causal lattice is generated through a causal sequential reasoning method, and the strongly and weakly coupled causal lattice is output to S4 for use.

[0043] Furthermore, the causal sequential reasoning method includes:

[0044] Using the candidate causal relationship pair set as input, causal relationship pairs with matching confidence levels lower than a second preset threshold are filtered out by thresholding to obtain a high-confidence candidate causal relationship pair set;

[0045] Using the set of high-confidence candidate causal relationship pairs and the path hierarchy relationship in the spectrum-coupled phase chain as joint input, the causal relationship pairs belonging to the same strong-electric-side coupling path are merged into a set of low-level strong-electric-side frequency point coupling event nodes through the low-level event aggregation algorithm;

[0046] Using the set of high-confidence candidate causal relationship pairs and the higher-order correlation terms of bit error in the timing error correlation array as joint inputs, the top-level event aggregation algorithm merges the causal relationship pairs corresponding to the same bit error mode into a top-level weak current side bit error correlation event node set.

[0047] Taking the causal mapping relationship between the set of frequency coupling event nodes on the bottom-level high-voltage side and the set of bit error association event nodes on the top-level low-voltage side as input, a multi-level intermediate transmission mediation event node set is introduced through a transmission mediation mining algorithm.

[0048] Using the set of frequency-coupled event nodes on the bottom strong electrical side, the set of multi-level conduction mediation event nodes in the middle layer, and the set of bit error association event nodes on the top weak electrical side as joint inputs, a set of partial order relations that satisfies reflexivity, antisymmetry, and transitivity is established through a lattice structure construction algorithm. This set of partial order relations is then organized into a strong-weak electrical coupled causal lattice with supremum and infimum operations.

[0049] Furthermore, the steps of predicting the safety margin decay trajectory and adjusting the linked isolation parameter feedforward include:

[0050] Using a strong-weak coupled causal lattice as input, a set of safety margin decay trajectories is generated through a safety margin decay trajectory prediction method.

[0051] Using the safety margin attenuation trajectory set and the strong-weak electrical coupling causal lattice as joint inputs, the isolation control command sequence after linkage adjustment is generated through the linkage isolation parameter feedforward adjustment method.

[0052] A safety isolation and linkage control system for integrated strong and weak current systems is provided, which is used to implement the aforementioned safety isolation and linkage control method for integrated strong and weak current systems. The system includes:

[0053] Timing error correlation array construction module: used to collect the original differential signal sequence, original ideal codeword sequence and original transient electrical data of the strong current side of the integrated strong and weak current system, and to establish a timing error correlation array through timestamp alignment and event correlation coding;

[0054] Spectrum-coupled phase chain establishment module: Based on the directional query results of the timing error correlation array, it performs adaptive window segmentation and phase coupling identification on the transient data of the high-voltage side to establish a spectrum-coupled phase chain;

[0055] Strong-weak electrical coupling causal lattice construction module: used to perform cross-domain isomorphic matching and causal sequential inference between the timing error correlation array and the spectral coupling phase chain to construct a strong-weak electrical coupling causal lattice;

[0056] Linkage control module: used for safety margin decay trajectory prediction and linkage isolation parameter feedforward adjustment based on strong and weak electrical coupling causal lattice.

[0057] Compared to existing technologies, the advantages of this invention are as follows: This invention solves the core problems of the micro-pattern of weak current communication errors caused by strong current disturbances in integrated strong and weak current systems being masked by protocol error correction mechanisms, and the lack of cross-domain causal mapping in existing monitoring. Through a time-series error correlation array, isolated error events are deeply coupled with strong current transient characteristics, explicitly revealing the micro-patterns of error periodicity, burstiness, and codeword correlation, breaking through the limitations of traditional methods that only count the total error rate, and allowing the implicit characteristics of strong current side state degradation to emerge. The spectral coupling phase chain characterizes the phase coupling path between multiple strong current frequency components, filling the mapping gap between macroscopic spectrum and microscopic interference, achieving precise focusing from weak current error symptoms to the strong current disturbance source frequency band, avoiding redundancy and noise interference from full-band scanning. The strong-weak current coupled causal lattice establishes a strict causal order relationship for cross-domain heterogeneous events, enabling rapid location of the smallest set of disturbance sources causing errors, avoiding the risk of misjudgment from simple correlation analysis. Based on the feedforward regulation of the causal grid, the prediction of the safety margin decay trajectory is combined with the adjustment of linkage parameters to suppress the degradation of isolation performance in advance. This not only accurately identifies early hidden dangers such as insulation aging and contact resistance deterioration on the high-voltage side, but also avoids the disorder of protection action sequence under critical operating conditions. At the same time, it reduces the downtime scope and labor cost of tracing the source of occasional faults, and realizes the paradigm upgrade of the system from sub-healthy operation to proactive pre-emptive control. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart of a safety isolation and linkage control method for integrated strong and weak current systems according to the present invention;

[0060] Figure 2 This is a schematic diagram of the original sequence of the differential signal for weak current communication and the sequence of its physical layer signal inversion points in an embodiment of the present invention;

[0061] Figure 3 This is a schematic diagram of the cluster distribution of bit error events on the periodic phase of a high-voltage switch in an embodiment of the present invention;

[0062] Figure 4 This is a schematic diagram of the spectral coupling phase chain characterizing the energy conduction path between multiple frequency points on the high-voltage side in an embodiment of the present invention;

[0063] Figure 5 This is a schematic diagram of a strong-weak coupling causal lattice illustrating a multi-stage conduction chain from strong electrical disturbance to weak electrical error in an embodiment of the present invention;

[0064] Figure 6 This is a functional block diagram of a safety isolation and linkage control system for integrated strong and weak current systems in this invention. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] Example 1:

[0067] Please see Figure 1 As shown, this embodiment provides a safety isolation and linkage control method for integrated strong and weak current systems, including:

[0068] S1: Collect the original differential signal sequence, original ideal codeword sequence, and original transient electrical data of the integrated strong and weak current system, and establish a timing error correlation array through timestamp alignment and event correlation coding.

[0069] In S1, the original sequence of the differential signal for weak current communication, the original ideal codeword sequence, and the original set of electrical transient data from the strong current side are used as joint inputs. Through layer-by-layer processing using the error monitoring probe decoding and comparison method, the unified time base generation method, the event correlation coding method, and the high-order correlation term construction method, a timing error correlation array is finally constructed. The timing error correlation array and the aligned data set of electrical transient data from the strong current side are then output to S2 for further use. The original sequence of the differential signal for weak current communication refers to the sequence of sampled values ​​of the differential voltage output by the differential transceiver on the weak current communication link, which varies with time. It is continuously acquired by a voltage sampling module set in the bypass of the communication link at a fixed sampling period. For a schematic diagram of the original sequence of the differential signal for weak current communication and its physical layer signal inversion point sequence, please refer to [reference needed]. Figure 2 The original ideal codeword sequence refers to the ideal bit codeword sequence output by the encoding unit inside the weak current side transmitter controller, which has not been transmitted through a physical channel. It is obtained by reading the transmission queue data in the transmitter controller's buffer. The original electrical transient data set on the strong current side refers to the time series set of various electrical quantities obtained by synchronously acquiring key nodes of the strong current circuit within the integrated strong and weak current cabinet. Specifically, it includes the voltage spike waveform sequence of power switching devices, the common-mode current pulse waveform sequence, and the ground plane potential drift sequence. Each waveform sequence is acquired by the corresponding voltage sensor, current sensor, or ground potential probe under a unified sampling period and stored in chronological order.

[0070] Furthermore, S1 includes S11 to S15.

[0071] S11: Based on the original sequence of differential signals in weak current communication and the original ideal codeword sequence, the initial set of error events is obtained by decoding and comparing with the error monitoring probe.

[0072] Specifically, the error monitoring probe decoding and comparison method includes the following three-level refinement steps:

[0073] S111: Using the original sequence of the differential signal for weak current communication as input, a physical layer signal flip-point sequence is extracted using an edge detection algorithm with an adaptive threshold. A physical layer sampling clock sequence is then constructed based on this sequence. The physical layer sampling clock sequence refers to the sequence of time sampling points representing the decision time of each bit within a data frame, calculated through uniform interpolation at the flip-point intervals. The edge detection algorithm with an adaptive threshold dynamically adjusts the judgment threshold based on the mean and standard deviation of the signal amplitude in real time for the original sequence of the differential signal for weak current communication. When the signal level crosses this threshold and the duration is greater than 1 ns, it is determined to be edge-triggered, thus accurately extracting the physical layer signal flip-points.

[0074] S112: Using the physical layer sampling clock sequence obtained in S111 and the original sequence of the differential signal of weak current communication as joint inputs, a decision bit sequence is generated through symbol decision. The symbol decision compares the polarity of the differential voltage at each time point corresponding to the physical layer sampling clock and maps it to a logic bit value, thereby obtaining a decision bit value sequence that corresponds one-to-one with time.

[0075] S113: Using the decision bit sequence obtained in S112 and the original ideal codeword sequence as joint input, an error flag sequence is generated through bit-by-bit comparison and forward error correction flag extraction rules. Each element in the error flag sequence records the comparison result, whether the forward error correction is corrected, and the corresponding codeword index and bit position index.

[0076] S114: Using the error flag sequence obtained in S113 and the physical layer sampling clock sequence obtained in S111 as joint inputs, an initial set of error events is generated through the error event encapsulation rules. The initial set of error events is an event set structure. Each error event unit contains fields such as the error occurrence timestamp, the data frame number to which it belongs, the codeword index to which it belongs, the bit position index within the codeword, and the forward error correction flag.

[0077] By using the error monitoring probe decoding and comparison method in S11, the continuous analog voltage waveform is mapped to a discrete initial set of error events. This enables subsequent fine-grained modeling of errors in both the time and codeword dimensions, solving the problem that traditional methods only count the total error rate and cannot characterize the details of errors. The initial set of error events output by S11 will serve as one of the inputs to S12.

[0078] S12: Based on the initial set of bit error events and the original set of electrical transient data on the high-voltage side obtained in S11, a unified time reference sequence is generated through a unified time base generation method, and a time-aligned set of bit error events and a set of electrical transient data on the high-voltage side are obtained.

[0079] Specifically, the unified time base generation method includes the following three-level refinement steps:

[0080] S121: Using the voltage spike waveform sequence of power switching devices in the original data set of electrical transients on the high-voltage side as input, the high-voltage side switching event time sequence is extracted by the switching transient edge detection algorithm. The high-voltage side switching event time sequence records the timestamp of each turn-on and turn-off moment.

[0081] S122: Using the time sequence of the high-voltage side switching events obtained in S121 and the timestamps of bit error occurrences in the initial set of bit error events output in S11 as joint inputs, a clock offset estimation sequence is obtained through a time drift estimation algorithm. The time drift estimation algorithm estimates the relative offset between the high-voltage and low-voltage side acquisition clocks by calculating the relative phase distribution of bit error events within the high-voltage side switching cycle and comparing it with the theoretical phase distribution. The theoretical phase distribution is obtained by modeling the mathematical relationship between the theoretical period of the high-voltage side switching device drive signal and the communication link clock frequency, and serves as a reference benchmark for clock drift estimation.

[0082] S123: Using the time drift estimate sequence obtained in S122 as input, a unified time reference sequence is generated through smoothing filtering and interpolation. The unified time reference sequence is defined as a monotonically increasing time index sequence, which is used as the sole time reference for various types of strong and weak current data.

[0083] S124: Using the initial set of bit error events output by S11 and the unified time reference sequence generated by S123 as joint inputs, a time-aligned bit error event set is obtained through the timestamp remapping method. Each bit error event unit in the time-aligned bit error event set has its timestamp field replaced with the corresponding value in the unified time reference sequence while keeping the original codeword index and bit position index unchanged.

[0084] S125: Using the original set of electrical transient data on the high-voltage side and the unified time reference sequence generated in S123 as joint inputs, a set of electrical transient aligned data on the high-voltage side is generated through resampling and interpolation methods. Each time sample point in the set of electrical transient aligned data on the high-voltage side corresponds to a time index in the unified time reference sequence and contains multiple components such as the voltage value of the power switching device, the common-mode current value, and the ground plane potential value under that time index.

[0085] The unified time reference generation method in S12 unifies the initial set of bit error events, which were originally sampled independently by different acquisition modules, and the original set of electrical transient data from the power supply side onto the same reference time axis. This avoids phase offset errors caused by inconsistent sampling clocks and provides a foundation for accurately establishing the correspondence between bit errors and power supply disturbances. The time-aligned bit error event set and the power supply side electrical transient aligned data set output by S12 will serve as the input to S13.

[0086] S13: Based on the time-aligned bit error event set and the electrical transient alignment data set of the power supply side obtained in S12, generate the bit error event base matrix through the event association coding method.

[0087] The basic matrix of bit error events is a two-dimensional matrix structure. The row index corresponds to each bit error event unit in the time-aligned bit error event set, and the column index corresponds to multiple feature dimensions, including the unified time reference sequence timestamp, the data frame number to which it belongs, the codeword index to which it belongs, the bit position index within the codeword, the forward error correction mark, and the electrical quantity characteristics of the high-voltage side adjacent to the timestamp.

[0088] Specifically, the event association coding method includes the following three-level refinement steps:

[0089] S131: Using the high-voltage side electrical transient alignment data set obtained in S12 as input, a high-voltage side local transient feature sequence is generated through a sliding time window extraction algorithm. Each element in the high-voltage side local transient feature sequence corresponds to a time window. Within this time window, statistics such as voltage peak extrema, common-mode current pulse peak value, mean and rate of change of ground plane potential drift are calculated. The sliding time window extraction algorithm refers to an algorithm that uses a unified time reference sequence as a benchmark, sets the window size to 1 / 8 of the data frame length, and the sliding step size to 1 / 2 of the window size to calculate the extrema, mean, and rate of change statistics within the window of the high-voltage side electrical transient alignment data.

[0090] S132: Using the time-aligned set of error events obtained in S12 and the high-voltage side local transient feature sequence generated in S131 as joint inputs, a high-voltage side local transient feature unit with the smallest time difference (not exceeding 1 μs) is selected for each error event using the nearest neighbor time window matching method, thereby attaching a high-voltage side electrical feature vector to each error event. The nearest neighbor time window matching method refers to finding the high-voltage side local transient feature window with the smallest time difference (not exceeding 1 μs) based on the error event timestamp, thus achieving matching between the error event and the high-voltage disturbance feature.

[0091] S133: Using the fields in the time-aligned bit error event set obtained in S12 and the electrical feature vector of the high-voltage side added in S132 as joint inputs, construct the basic matrix of bit error events through feature concatenation and standardization operations. Each row vector records the time, codeword position and local high-voltage disturbance features of the corresponding bit error event.

[0092] By using the event correlation coding method in S13, information from both the strong and weak electrical sides is aggregated in the same row vector, so that each bit error event carries the strong electrical disturbance context near its occurrence time, thus laying the foundation for subsequent identification of higher-order correlation patterns and inference of causal relationships. The bit error event basis matrix output by S13 will serve as one of the inputs to S14 and S15.

[0093] S14: Based on the error event matrix obtained in S13, generate a set of higher-order correlation terms using the higher-order correlation term construction method.

[0094] The higher-order correlation term set is a pattern set. Each higher-order correlation term records a set of error event indices and corresponding pattern weight values ​​that have a stable coupling relationship in time and codeword distance. Specifically, the higher-order correlation term construction method includes the following three-level refinement steps:

[0095] S141: Using the timestamp column in the error event matrix obtained in S13 as input, a candidate period set for the error time series is obtained through a periodic candidate frequency search algorithm. The periodic candidate frequency search algorithm is an algorithm that calculates the autocorrelation function of the error time series, selects frequency points with an autocorrelation value greater than 0.7 as candidates, and combines spectral peak extraction to filter out the top five frequencies to form a candidate period set.

[0096] S142: Using the candidate period set obtained in S141 and the timestamp column in the error event base matrix obtained in S13 as joint inputs, the phase distribution sequence under each candidate period is obtained through phase folding operation. The phase distribution sequence projects the position of the error event within a period into a standardized phase value.

[0097] S143: Using the phase distribution sequence obtained in S142 as input, a set of periodic clustering results is generated through phase clustering and burst degree evaluation algorithms. Each periodic clustering result records the set of bit error event indices contained in a phase cluster under a certain candidate period, as well as the concentration index of that phase cluster. Please refer to the schematic diagram of the clustering distribution of bit error events on the periodic phases of the high-voltage switch for more information. Figure 3 The phase clustering and burstiness assessment algorithm refers to an algorithm that uses K-means clustering to group phase distribution sequences, defines burstiness by the ratio of the number of bit error events within a cluster to the time span, and assesses the degree of cluster concentration.

[0098] S144: Using the set of periodic clustering results obtained in S143 as input, the codeword distance distribution entropy value of each phase cluster is obtained through codeword distance statistics and information entropy calculation methods. The codeword distance distribution entropy value is used to measure the degree of dispersion of bit errors on the codeword index axis.

[0099] S145: Using the periodic clustering results obtained in S143 and the codeword distance distribution entropy obtained in S144 as joint inputs, a pattern weight value is generated for each phase cluster according to the pattern weight calculation rule. The pattern weight value comprehensively considers the frequency of recurrence of the phase cluster in time and the concentration in codeword distance, thus forming each higher-order association term. The pattern weight calculation rule refers to the calculation rule that uses the weighted sum of the proportion of time recurrence and the concentration of codeword distance as the pattern weight value.

[0100] S146: Using all the higher-order association terms generated in S145 as input, a set of higher-order association terms is generated through set construction operation and combined with the first preset threshold for filtering. The first preset threshold is obtained by statistically analyzing the cumulative distribution of the pattern weight values ​​of all higher-order association terms in the historical dataset, and taking the 90 percentile as the boundary. Exceeding this value indicates that the pattern is significantly higher than the level of random noise.

[0101] By constructing higher-order correlation terms in S14, the seemingly random and discrete fundamental matrix of bit error events is mined to reveal implicit periodic patterns driven by the high-voltage switching modulation frequency in both time and codeword dimensions. This allows for subsequent quantitative analysis of the impact of these pattern weights on high-voltage disturbances. The set of higher-order correlation terms output from S14 will serve as one of the inputs to S15 and S21.

[0102] S15: Based on the bit error event base matrix obtained in S13 and the set of higher-order correlation terms obtained in S14, construct the timing bit error correlation array through the array fusion generation method, and output the timing bit error correlation array and the set of electrical transient alignment data of the strong power side obtained in S12 to S2 for use.

[0103] Specifically, the array fusion generation method includes the following three-level refinement steps:

[0104] S151: Using the row index of the error event base matrix obtained in S13 and the set of error event indices in the set of higher-order correlation terms obtained in S14 as joint input, the correlation indicator matrix from error events to higher-order correlation terms is obtained through the mapping operation from event to pattern. Each element of the correlation indicator matrix indicates whether the corresponding error event belongs to a certain higher-order correlation term.

[0105] S152: Using the error event base matrix obtained in S13 and the correlation indicator matrix obtained in S151 as joint inputs, a time-series error correlation array is generated through dimension expansion and index encoding operations. In the time-series error correlation array, each error event unit not only retains the original strong and weak electrical characteristics, but also adds a field indicating which higher-order correlation terms the error event belongs to and the corresponding mode weight.

[0106] By constructing the temporal error correlation array in S15, single-point error events and multi-event sequence patterns are stored and indexed in a unified manner. This allows for subsequent analysis of strong and weak electrical coupling relationships at both the error sample level and the pattern level, resolving the contradiction that traditional error statistics cannot simultaneously consider transient characteristics and long-term patterns. The temporal error correlation array output by S15 is one of the core inputs of S2 and S3.

[0107] S2: Based on the directional query results of the timing error correlation array, adaptive window segmentation and phase coupling identification are performed on the transient data of the power supply side to establish a spectrum-coupled phase chain.

[0108] In S2, the timing error correlation array output from S1 and the electrical transient alignment data set from the high-voltage side are used as joint inputs. Through a high-order correlation term directional query method, an adaptive window segmentation method, an instantaneous phase extraction method, and a phase-locked detection method, a spectral coupled phase chain is constructed and output to S3. The spectral coupled phase chain is a directed graph structure where nodes are discrete frequency units of the high-voltage side transient signal, edges represent the detected phase-locked relationships between different frequency units, and edge weights are combinations of phase-locked duration and lock strength, representing the coupling path and strength of energy transfer from low-frequency components to high-frequency components.

[0109] Furthermore, S2 includes S21 to S25.

[0110] S21: Using the timing error correlation array output by S1 as input, generate a set of high-order correlation term pointing query results through the high-order correlation term pointing query method.

[0111] The higher-order association term directional query result set is a result set structure in which each element corresponds to the statistical characteristics of a higher-order association term in the time-series error association array, including the error event timestamp subsequence involved in the higher-order association term, the codeword index subsequence involved, the pattern weight value of the higher-order association term, and the implicit period estimate value estimated based on the time interval sequence.

[0112] Specifically, the higher-order related item pointing query method includes the following three-level refinement steps:

[0113] S211: Taking the set of higher-order correlation items in the timing error correlation array output by S1 as input, the highest number of higher-order correlation items with the highest mode weight values ​​are selected by the mode weight sorting algorithm to obtain a subset of candidate higher-order correlation items.

[0114] S212: Taking the set of error event indices for each higher-order correlation item in the candidate higher-order correlation item subset obtained in S211 as input, the corresponding error event timestamp subsequence and codeword index subsequence are obtained by retrieving them from the time-series error correlation array.

[0115] S213: Using the error event timestamp subsequence obtained in S212 as input, the implicit period estimate is obtained through time interval statistics and frequency estimation algorithms. The implicit period estimate is used to indicate the dominant frequency range of the strong electrical disturbance corresponding to the higher-order correlation term.

[0116] S214: Using the error event timestamp subsequence, codeword index subsequence obtained in S212, the pattern weight value obtained in S211, and the implicit period estimate obtained in S213 as joint inputs, a set of high-order association terminological query results is generated through result encapsulation rules.

[0117] The high-order correlation term directional query method in S21 can prioritize the selection of several modes most sensitive to high-voltage disturbances from a large number of error patterns, and extract their corresponding implicit periodic information. This provides concentrated and targeted frequency candidates for subsequent frequency window selection, avoiding computational redundancy caused by full-band scanning of the high-voltage signal. The set of high-order correlation term directional query results output by S21 will serve as one of the inputs to S22.

[0118] S22: Using the set of high-order correlation term directional query results obtained from S21 and the set of electrical transient alignment data on the strong power side output from S1 as joint inputs, a set of frequency window segmentation results is generated through an adaptive frequency window segmentation method.

[0119] The frequency window segmentation result set is a set structure, where each item records a center frequency, a bandwidth, and the corresponding bandpass signal sequence for that bandwidth. Specifically, the adaptive frequency window segmentation method includes the following three-level refinement steps:

[0120] S221: Using the implicit period estimate from the higher-order correlation term directional query result set obtained in S21 as input, the candidate set of center frequency points is obtained through period-to-frequency conversion operation.

[0121] S222: Using the candidate set of center frequency points obtained in S221 and the set of electrical transient alignment data of the high-voltage side output in S1 as joint inputs, a multi-dimensional spectrum estimation result is generated through a multi-channel joint spectrum estimation algorithm. The multi-dimensional spectrum estimation result includes the amplitude distribution of voltage spikes, common-mode current, and ground plane potential components in the neighborhood of each center frequency point. The multi-channel joint spectrum estimation algorithm refers to performing FFT transformation on the voltage, current, and potential multi-channel data respectively, taking the maximum value of the spectrum amplitude of each channel as the joint amplitude of the corresponding frequency point, and generating the multi-dimensional spectrum estimation result.

[0122] S223: Using the candidate set of center frequencies obtained in S221 and the multi-dimensional spectrum estimation results generated in S222 as joint inputs, a bandwidth is determined for each center frequency through a bandwidth energy concentration assessment and bandwidth adaptive adjustment algorithm. This ensures that the bandwidth includes the main energy components near the frequency while minimizing irrelevant noise components. The bandwidth energy concentration assessment and bandwidth adaptive adjustment algorithm uses the energy percentage within a 3dB bandwidth of the frequency's neighborhood as a concentration index. When the percentage is below 80%, the bandwidth is widened to 5dB; when it is above 95%, it is narrowed to 2dB.

[0123] S224: Using the set of electrical transient alignment data output from S1 and each pair of center frequency points and bandwidths determined by S221 to S223 as joint inputs, the corresponding bandpass signal sequence is obtained through bandpass filtering operation, and the center frequency point, bandwidth and bandpass signal sequence are encapsulated into a frequency window segmentation result unit, thereby forming a frequency window segmentation result set.

[0124] By employing the adaptive window segmentation method in S22, the periodic information implicit in the bit error pattern is used inversely to influence the selection of the high-voltage side frequency band. Compared to a fixed-band analysis strategy, this significantly improves the focusing capability on the frequency band most relevant to the bit error and reduces the interference of irrelevant frequency bands on the phase analysis results. The set of window segmentation results output by S22 will serve as the input to S23.

[0125] S23: Using the set of frequency window segmentation results obtained in S22 as input, generate a set of instantaneous phase trajectories through the instantaneous phase extraction method.

[0126] The instantaneous phase trajectory set is a time series set, where each element is the instantaneous phase sequence corresponding to a certain frequency window, used to describe the phase evolution of the signal within that frequency band on a unified time reference sequence. Specifically, the instantaneous phase extraction method includes the following three-level refinement steps:

[0127] S231: Taking each bandpass signal sequence in the frequency window segmentation result set obtained in S22 as input, a corresponding complex envelope signal sequence is generated through an analytical signal construction algorithm. The real part of the complex envelope signal sequence is the bandpass signal itself, and the imaginary part is the orthogonal component of the same frequency obtained through Hilbert transform. The analytical signal construction algorithm refers to the algorithm that first performs notch filtering on the bandpass signal, and then constructs orthogonal components through Hilbert transform to form a complex envelope signal with the real part being the original signal and the imaginary part being the orthogonal component.

[0128] S232: Using the complex envelope signal sequence obtained in S231 as input, the instantaneous phase sequence is obtained through arctangent function operation. Each sample point of the instantaneous phase sequence represents the phase value of the signal in that frequency band at the corresponding time in the unified time reference sequence.

[0129] S233: Using all the instantaneous phase sequences obtained in S232 as input, a set of continuous instantaneous phase trajectories is obtained through phase expansion and de-jump processing. The set of continuous instantaneous phase trajectories avoids the influence of phase value jumps between periods on subsequent phase difference statistics.

[0130] The instantaneous phase extraction method in S23 restores the instantaneous phase of each frequency band signal on the high-voltage side, enabling subsequent analysis of the synchronization and locking relationship between frequency components directly in the phase space rather than just the amplitude space. This more closely approximates the actual physical process of common-mode disturbances coupling to the low-voltage side via nonlinear paths. The set of instantaneous phase trajectories output by S23 will serve as the input to S24.

[0131] S24: Using the instantaneous phase trajectory set obtained in S23 as input, generate a set of frequency point phase coupling relationships through the phase-locked detection method.

[0132] The frequency-point phase coupling relationship set is a ensemble structure, where each entry records the average phase difference, phase difference fluctuation range, phase-locking duration, and phase-locking strength between a pair of frequency points. Specifically, the phase-locking detection method includes the following three-level refinement steps:

[0133] S241: Using the instantaneous phase sequences corresponding to any two frequency points in the instantaneous phase trajectory set obtained in S23 as joint input, the phase difference time sequence is obtained through phase difference calculation.

[0134] S242: Using the phase difference time series obtained in S241 as input, the average value and variance of the phase difference are calculated through the phase difference statistical analysis algorithm, and the presence of phase locking phenomenon of the frequency pair is determined based on the preset phase locking criterion. The preset phase locking criterion is defined as the phase difference variance being lower than the threshold obtained by adding 3 times the standard deviation of the reference noise variance and the continuous duration exceeding the lower limit of the least common multiple of the power switch period and the communication frame length.

[0135] S243: Using the phase difference time series that satisfies the phase locking criterion in S242 as input, the phase locking duration and phase locking strength are obtained through phase locking duration calculation and phase locking strength measurement algorithm. The phase locking strength measurement algorithm is an algorithm that uses the weighted sum of the reciprocal of the phase difference variance, the locking duration, and the deviation of the mean phase difference as the phase locking strength.

[0136] S244: Using the frequency pairs, average phase difference, variance of phase difference, phase lock duration, and phase lock strength determined by S241 to S243 as joint inputs, a set of frequency point phase coupling relationships is generated through result encapsulation rules.

[0137] The phase-locked detection method in S24 can identify coupling pairs with stable phase relationships between different frequency components on the high-voltage side within a certain time range. This allows us to characterize the timing structure of energy migration from the low-frequency fundamental wave to the high-frequency cluster via modulation and harmonic generation mechanisms. This structure is crucial for tracing how common-mode interference excites the frequency band most sensitive to weak-voltage communication links. The set of frequency point phase coupling relationships output by S24 will serve as the input to S25.

[0138] S25: Using the set of frequency-point phase coupling relationships obtained in S24 as input, a spectrum-coupled phase chain is generated through a chain structure construction method, and the spectrum-coupled phase chain is output to S3 for use.

[0139] Specifically, the chain structure construction method includes the following three levels of refinement steps:

[0140] S251: Using the set of frequency point phase coupling relationships obtained in S24 as input, a set of directed edges is constructed by sorting and filtering according to the frequency point size. Specifically, the sorting and filtering rules are to sort all frequency points from low to high frequency and retain only the phase coupling relationships from low frequency to high frequency, thereby constructing a set of directed edges.

[0141] S252: Using the sorted set of frequency points and the set of directed edges obtained in S251 as joint input, an initial spectrum coupling graph structure is generated through a directed graph construction algorithm. In the initial spectrum coupling graph structure, each node corresponds to a frequency point, and each directed edge corresponds to a frequency point phase coupling relationship.

[0142] S253: Using the initial spectral coupling diagram structure obtained in S252 as input, a spectral coupling phase chain is obtained through path merging and weight aggregation algorithms. In the spectral coupling phase chain, directed edges connecting multiple frequency points are chained together in a monotonically increasing direction to form several directed paths, and a comprehensive weight of the phase-locking strength is calculated for each directed path. The path merging and weight aggregation algorithm refers to merging adjacent paths with continuous frequencies and a phase-locking strength greater than 0.6, using the strength-weighted average of the merged paths (the weight being the proportion of the path length) as the aggregated weight. Please refer to the schematic diagram of the spectral coupling phase chain representing the energy conduction path between multiple frequency points on the high-voltage side. Figure 4 .

[0143] By constructing the spectral coupling phase chain in S25, the energy conduction paths between multiple frequency components within the high-voltage side in the phase space can be formally represented, enabling a transition from traditional amplitude spectrum analysis to phase coupling analysis. This provides an operable structural framework for subsequent causal matching of these coupling paths with the bit error patterns of the low-voltage side. The spectral coupling phase chain output by S25 is one of the core inputs of S3.

[0144] S3: Perform cross-domain isomorphic matching and causal sequential inference between the timing error correlation array and the spectral coupled phase chain to construct a strong-weak coupled causal lattice.

[0145] In S3, the timing error correlation array output from S1 and the spectral coupling phase chain output from S2 are used as joint inputs. Through the collaborative processing of error mode feature extraction method, perturbation coupling feature extraction method, cross-domain isomorphic matching method, and causal sequential inference method, a strong-weak electrical coupled causal lattice is finally generated, and the strong-weak electrical coupled causal lattice is output to S4 for use. The strong-weak electrical coupled causal lattice is a lattice structure that satisfies a partial order relation. Its bottom layer nodes are the set of frequency coupling events on the strong electrical side, the upper layer nodes are the set of error correlation events on the weak electrical side, the middle layer nodes are the set of multi-level conduction mediator events, and the edges are strict causal priority relations, used to represent the multi-level conduction chain from strong electrical perturbation to weak electrical error.

[0146] Furthermore, S3 includes S31 to S34.

[0147] S31: Using the timing error correlation array output by S1 as input, generate a set of error mode feature vectors through the error mode feature extraction method.

[0148] The bit error pattern feature vector set is a set of vectors, where each bit error pattern feature vector corresponds to a higher-order correlation term in the temporal error correlation array. This feature vector is used to quantify the statistical characteristics of the bit error pattern corresponding to that higher-order correlation term in terms of time, frequency, and codeword structure. Specifically, the bit error pattern feature extraction method includes the following three-level refinement steps:

[0149] S311: Taking the set of higher-order correlation items in the timing error correlation array output by S1 as input, the error event index set involved in each higher-order correlation item is read sequentially through the pattern traversal algorithm.

[0150] S312: Taking each set of bit error event indices obtained in S311 as input, the corresponding bit error event timestamp subsequence, codeword index subsequence, and forward error correction mark subsequence are obtained by retrieving them from the timing error correlation array.

[0151] S313: Using the error event timestamp subsequence obtained in S312 as input, the time concentration index and period consistency index of this pattern are obtained through the time concentration and period consistency calculation algorithm. The time concentration and period consistency calculation algorithm refers to the algorithm that defines the time concentration by the standard deviation of the error event within the period, and defines the period consistency by the deviation rate between the actual period and the estimated period (preferably less than or equal to 5%).

[0152] S314: Using the codeword index subsequence obtained in S312 as input, the codeword clustering index of this pattern is obtained through codeword distance statistics and intra-cluster dispersion calculation. The codeword distance statistics and intra-cluster dispersion calculation algorithm refers to the algorithm that uses Hamming distance to statistically analyze codeword differences and defines dispersion by the standard deviation of intra-cluster codeword distance.

[0153] S315: Using the forward error correction marker subsequence obtained in S312 as input, the error correction ratio index in this mode is calculated by the error correction ratio.

[0154] S316: Using the time concentration index and period consistency index obtained in S313, the codeword clustering index obtained in S314, and the error rate index obtained in S315 as joint inputs, error pattern feature vectors are generated through feature concatenation operations, thereby forming a set of error pattern feature vectors.

[0155] The error pattern feature extraction method in S31 can transform the previously difficult-to-compare higher-order correlation terms of bit errors into vector representations within the same feature space, facilitating subsequent isomorphic matching with the disturbance features of the high-voltage side. The set of error pattern feature vectors output by S31 will serve as one of the inputs to S33.

[0156] S32: Using the spectral coupling phase chain output by S2 as input, a set of perturbation coupling feature vectors is generated through the perturbation coupling feature extraction method.

[0157] The perturbation coupling feature vector set is a set of vectors, where each perturbation coupling feature vector corresponds to a directed coupling path in the spectral coupling phase chain, used to quantify the comprehensive characteristics of that path in terms of frequency, phase, and energy transfer. Specifically, the perturbation coupling feature extraction method includes the following three-level refinement steps:

[0158] S321: Taking the spectrum-coupled phase chain output by S2 as input, each directed coupling path is read sequentially through the path traversal algorithm, and the frequency point sequence and the corresponding phase-locked strength weight sequence contained in the directed coupling path are obtained.

[0159] S322: Using the frequency sequence obtained in S321 as input, frequency characteristic indicators such as the starting frequency, ending frequency, and frequency span of the path are obtained through frequency span calculation and frequency level statistical algorithms. The frequency span calculation and frequency level statistical algorithms refer to the algorithm that calculates the frequency span based on the difference between the ending frequency and the starting frequency, and divides the frequency into three levels: ≤1kHz for low level, 1-10kHz for mid level, and >10kHz for high level.

[0160] S323: Using the phase lock strength weight sequence obtained in S321 as input, the average lock strength and bottleneck lock strength indices of the path are obtained through the average lock strength calculation and weakest link identification algorithm. The average lock strength calculation and weakest link identification algorithm refers to the algorithm that takes the arithmetic mean of the coupling strength of each frequency point on the path as the average lock strength, and takes the frequency point corresponding to the minimum strength value as the weakest link.

[0161] S324: Using the statistical characteristics of the high-voltage side current and voltage corresponding to the bottom frequency point in the spectrum-coupled phase chain output by S2 as input, the energy propagation speed index of the path on the time axis is obtained through the path energy transfer rate estimation algorithm. The path energy transfer rate estimation algorithm refers to the algorithm that estimates the energy transfer rate by dividing the energy difference between the starting and ending frequency points of the path by the transmission time.

[0162] S325: Using the starting frequency, ending frequency and frequency span obtained in S322, the average locking strength and bottleneck locking strength obtained in S323, and the energy propagation speed index obtained in S324 as joint inputs, a perturbation coupling feature vector is generated through feature splicing operation, thereby forming a perturbation coupling feature vector set.

[0163] By using the perturbation coupling feature extraction method in S32, the complex frequency-phase coupling paths on the high-voltage side can be abstracted into a set of comparable feature vectors. This allows for the measurement of the potential contribution of different perturbation paths to the bit error mode using a unified feature space during cross-domain matching. The set of perturbation coupling feature vectors output by S32 will serve as one of the inputs to S33.

[0164] S33: Using the set of error pattern feature vectors obtained in S31 and the set of perturbation coupling feature vectors obtained in S32 as joint inputs, a set of candidate causal relationship pairs is generated through the cross-domain isomorphic matching method.

[0165] The candidate causal relationship set is a set of binary pairs, each consisting of a bit error pattern feature vector and a perturbation coupling feature vector, along with a matching confidence index to characterize the probability that the bit error pattern is driven by the perturbation path. Specifically, the cross-domain isomorphic matching method includes the following three-level refinement steps:

[0166] S331: Taking the set of error mode feature vectors obtained in S31 and the set of perturbation coupling feature vectors obtained in S32 as inputs, the components of the two types of feature vectors are scaled to a uniform dimension range through feature normalization and weight allocation algorithms, and different weight coefficients are assigned to time-related features, frequency-related features and intensity-related features.

[0167] S332: Using each bit error mode feature vector obtained in S31 and each disturbance coupling feature vector obtained in S32 as joint input, a partial order inclusion relation test algorithm is used to determine whether the bit error mode feature vector can be included by a certain disturbance coupling feature vector in each major component. That is, it is determined whether the periodic consistency index of the bit error mode is compatible with the frequency span and start and end frequencies of the disturbance path, whether the temporal concentration index of the bit error mode falls within the energy propagation time window of the disturbance path, and whether the intensity correlation index of the bit error mode does not exceed the bottleneck locking strength of the disturbance path. The partial order inclusion relation test algorithm is a step-by-step test algorithm that checks whether the periodicity of the bit error mode is compatible with the frequency span of the disturbance, whether the temporal concentration falls within the energy propagation time window, and whether the intensity index does not exceed the bottleneck locking strength.

[0168] S333: Taking the bit error pattern feature vectors and perturbation coupling feature vector pairs that satisfy the partial order inclusion relationship in S332 as input, a matching confidence index is generated through a matching score calculation algorithm. The matching confidence index can be calculated by comprehensively considering the degree of partial order inclusion, feature difference, and historical statistical verification results. The historical statistical verification results are derived from a database of verified causal relationship pairs, and the lower quartile of their feature vector matching scores is used as the baseline for confidence assessment. The matching score calculation algorithm refers to an algorithm that uses cosine similarity to calculate the feature vector matching degree and then corrects it with the historical matching baseline to obtain the matching confidence.

[0169] S334: Using the error pattern feature vector, perturbation coupling feature vector and matching confidence index obtained from S333 as joint inputs, a set of candidate causal relationship pairs is generated through result encapsulation operation.

[0170] By employing the cross-domain isomorphic matching method in S33, it is possible to determine, from the perspective of physical feature space, which strong electrical disturbance paths can fully explain a certain type of bit error mode without relying on simple temporal sequence relationships. This solves the problem that traditional correlation analysis struggles to eliminate random correlations and common-cause effects. The set of candidate causal pairs output by S33 will serve as the input to S34.

[0171] S34: Using the set of candidate causal relationship pairs obtained in S33 as input, generate a strong-weak electrical coupling causal lattice through the causal sequential reasoning method, and output the strong-weak electrical coupling causal lattice to S4 for use.

[0172] Specifically, the causal sequential reasoning method includes the following three levels of refinement steps:

[0173] S341: Using the candidate causal relationship pair set obtained in S33 as input, causal relationship pairs with matching confidence scores lower than a second preset threshold are filtered out through thresholding to obtain a high-confidence candidate causal relationship pair set. The second preset threshold is determined based on a verified causal relationship pair database by statistically analyzing the lower quartile of their feature vector matching scores, and is used to remove causal relationship pairs with insufficient matching confidence.

[0174] S342: Using the high-confidence candidate causal relationship pair set obtained in S341 and the path hierarchy relationship in the spectral coupled phase chain output in S2 as joint inputs, a low-level event aggregation algorithm is used to merge causal relationship pairs belonging to the same strong-electric coupling path into a low-level strong-electric frequency point coupled event node set. The low-level event aggregation algorithm refers to the algorithm that merges causal relationship pairs belonging to the same strong-electric coupling path with a matching confidence ≥ 0.8, using the path feature vector as the representation of the aggregated nodes.

[0175] S343: Using the high-confidence candidate causal relationship pair set obtained in S341 and the higher-order error correlation terms in the time-series error correlation array output by S1 as joint inputs, a top-level event aggregation algorithm is used to merge causal relationship pairs corresponding to the same error mode into a top-level weak current side error correlation event node set. The top-level event aggregation algorithm refers to an algorithm that merges causal relationship pairs corresponding to the same error mode with a mode weight ≥ 0.7, using the error feature vector as the representation of the aggregated nodes.

[0176] S344: Taking the causal mapping relationship between the bottom-level high-voltage side frequency coupling event node set obtained in S342 and the top-level low-voltage side bit error association event node set obtained in S343 as input, a multi-level intermediate-layer conducted mediation event node set is introduced through a conducted mediation mining algorithm. The multi-level intermediate-layer conducted mediation event node set can represent physical processes such as grounding impedance change and cable shield degradation that are located in time between high-voltage disturbances and bit errors and cannot be directly observed. The conducted mediation mining algorithm refers to an algorithm that mines implicit physical processes (such as grounding impedance change) between high-voltage disturbances and bit errors that satisfy causal transitivity and have an occurrence probability ≥ 0.6 as mediation events.

[0177] S345: Using the set of low-level strong-current side frequency-coupled event nodes obtained in S342, the set of intermediate-level multi-stage conduction mediation event nodes obtained in S344, and the set of high-level weak-current side bit error correlation event nodes obtained in S343 as joint inputs, a set of partial order relations satisfying reflexivity, antisymmetry, and transitivity is established through a lattice structure construction algorithm. This set of partial order relations is then organized into a strong-weak-current coupled causal lattice with supremum and infimum operations. For a schematic diagram of the strong-weak-current coupled causal lattice describing the multi-stage conduction chain from strong-current disturbance to weak-current bit error, please refer to [reference needed]. Figure 5 The lattice structure construction algorithm refers to a lattice structure construction algorithm that establishes a partial order of nodes (cause nodes ≤ result nodes) based on causal mapping relationships, with the supremum as common result nodes and the infimum as common cause nodes.

[0178] By employing the causal sequential reasoning method in S34, a logically consistent strongly- and weakly-coupled causal lattice structure is extracted from the set of candidate causal pairs. This allows the system to quickly locate the smallest possible set of strong electrical disturbance sources when any error mode occurs, and to distinguish between the dominant and secondary paths under multi-path coupling, thus providing an interpretable causal basis for safety margin quantification and control decisions. The strongly- and weakly-coupled causal lattice output by S34 is the core input of S4.

[0179] S4: Prediction of safety margin decay trajectory and feedforward adjustment of linkage isolation parameters based on strong and weak electrical coupling causal lattice.

[0180] In S4, the strong-weak coupling causal grid output from S3 is used as input. Through the cascade processing of the safety margin attenuation trajectory prediction method and the linkage isolation parameter feedforward adjustment method, a linkage-adjusted isolation control command sequence is finally generated. This sequence is then sent to the strong-side switch drive unit and the weak-side communication interface unit for execution. The linkage-adjusted isolation control command sequence is a set of timing control commands. Each command contains parameters such as the drive slope setting value of the target strong-side switch device, the absorption circuit parameter setting value, the differential drive level setting value of the target weak-side communication link, and the filter cutoff frequency setting value. This is used to proactively adjust the strong-weak isolation configuration before the fault becomes apparent.

[0181] Furthermore, S4 includes S41 and S42.

[0182] S41: Using the strong and weak electrical coupling causal lattice output by S3 as input, a set of safety margin decay trajectories is generated through the safety margin decay trajectory prediction method.

[0183] The safety margin decay trajectory set is a collection of time series. Each time series corresponds to the shortest causal chain in the strong-weak coupling causal lattice, from the bottom-level strong-electrical side frequency coupling event node to the top-level weak-electrical side bit error correlation event node. This chain describes the degradation and evolution of the attributes of each node over time. Specifically, the safety margin decay trajectory prediction method includes the following three-level refinement steps:

[0184] S411: Taking the set of strong electrical side frequency coupling event nodes in the strong electrical coupling causal grid output by S3 as input, a set of phase-locking strength historical sequences is constructed through the node historical attribute acquisition algorithm. Each time sample point in the set of phase-locking strength historical sequences records the phase-locking strength of the corresponding frequency coupling event under that time slice.

[0185] S412: Using the historical sequence set of phase-locked strength obtained in S411 as input, a degradation rate estimation sequence for each underlying frequency-point coupling event node is obtained through a degradation rate estimation algorithm. The degradation rate estimation sequence is used to quantify the trend of phase-locked strength decreasing over time. The degradation rate estimation algorithm refers to an algorithm that uses a linear regression model to fit the historical phase-locked strength data, uses the regression coefficients as the degradation rate, and estimates the future degradation trend.

[0186] S413: Using the degradation rate estimate sequence obtained in S412 and the partial order structure of the strong-weak electrical coupled causal lattice output in S3 as joint inputs, the degradation effect of the bottom node is transmitted step by step along the causal edge to the intermediate multi-level transmission intermediate event node set and the top weak electrical side bit error associated event node set through lattice conduction operation, thereby obtaining the future bit error burst probability prediction sequence of each top bit error associated event node.

[0187] S414: Using the strong-weak electrical coupling causal lattice output from S3 and the error burst probability prediction sequence obtained from S413 as joint inputs, the shortest causal chain length sequence from any bottom-level strong electrical frequency coupling event node to a top-level error-related event node is obtained under the conditions of the current time and the future prediction time through the shortest causal chain length calculation algorithm. The shortest causal chain length can be obtained by weighted summation of the edge weights in the strong-weak electrical coupling causal lattice, where the edge weights can be jointly determined by the node degradation degree and the matching confidence. The shortest causal chain length calculation algorithm refers to an algorithm that uses the product of the node degradation degree and the matching confidence as the edge weight and employs Dijkstra's algorithm to search for the shortest path from the bottom layer to the top layer node.

[0188] S415: Using the shortest causal chain length sequence obtained from S414 as input, calculate the safety margin attenuation from the change in insulation impedance on the high-voltage side to the over-limit of the bit error rate in low-voltage communication through the safety margin quantification rule, and organize these safety margin attenuation amounts into a set of safety margin attenuation trajectories on the time axis.

[0189] By using the safety margin decay trajectory prediction method in S41, the sub-healthy state of strong and weak current isolation effectiveness can be identified in advance before the fault manifests as an observable link interruption, and the risk of each causal chain leading to a significant deterioration in bit error rate over a future period can be quantified. The set of safety margin decay trajectories output by S41 will serve as one of the inputs to S42.

[0190] S42: Using the safety margin attenuation trajectory set obtained from S41 and the strong and weak electrical coupling causal lattice output from S3 as joint inputs, the isolation control command sequence after linkage adjustment is generated through the linkage isolation parameter feedforward adjustment method.

[0191] Specifically, the linkage isolation parameter feedforward adjustment method includes the following three-level refinement steps:

[0192] S421: Using the set of safety margin attenuation trajectories obtained in S41 as input, the set of target causal chains whose safety margin attenuation exceeds the preset safety threshold is identified by the threshold comparison algorithm. Each causal chain in the target causal chain set represents a potential dangerous transmission path that requires feedforward intervention. The preset safety threshold is determined based on the tolerance of the difference between the system design isolation performance and the measured isolation performance. Specifically, it is the maximum proportion that the shortest causal chain length can be shortened under the premise of ensuring 99% communication reliability.

[0193] S422: Using the target causal chain set obtained in S421 and the strongly and weakly coupled causal lattice output in S3 as joint inputs, the marginal contribution of each edge in each target causal chain to the overall safety margin attenuation is calculated using a contribution decomposition algorithm, thereby obtaining the set of critical coupling paths with the largest weight contribution in each target causal chain. The contribution decomposition algorithm refers to an algorithm that uses the partial derivative method to calculate the marginal contribution of each edge to the safety margin attenuation, and selects the top 30% after sorting by absolute value as the critical contributing edges.

[0194] S423: Taking the underlying high-voltage side frequency coupling event node corresponding to each path in the key coupling path set obtained in S422 as input, the set of high-voltage side switching devices most affected by the frequency coupling event is determined through a switching device mapping algorithm. Using this set of high-voltage side switching devices and the safety margin attenuation obtained in S41 as joint input, the target drive slope setting value and target absorption circuit parameter setting value for each high-voltage side switching device are calculated through a drive slope and absorption circuit parameter adjustment model. The drive slope and absorption circuit parameter adjustment model refers to a model that optimizes parameters through a linear adjustment model (drive slope = baseline value × (1 + safety margin attenuation × 0.5)) using the safety margin attenuation as input. The switching device mapping algorithm is an algorithm that matches the frequency corresponding to the frequency coupling event with the driving frequency of the switching device, and determines the associated switching device when the matching error is ≤3%.

[0195] S424: Taking the top-level weak current side bit error association event node corresponding to each path in the key coupling path set obtained in S422 as input, the set of weak current communication links participating in this bit error mode is determined through the communication link mapping algorithm. Using this set of weak current communication links and the corresponding bit error burst probability prediction sequence obtained in S41 as joint input, the target differential drive level setting value and target filter cutoff frequency setting value for each weak current communication link are calculated through the differential drive level and filter cutoff frequency adaptive adjustment model. The communication link mapping algorithm refers to an algorithm that maps the codeword index of the bit error mode to the channel number of the communication link, determining the associated link when the mapping accuracy is ≥90%. The differential drive level and filter cutoff frequency adaptive adjustment model refers to an adjustment model that takes the bit error burst probability as input, where the drive level increases linearly with the probability (≤120% of the rated value), and the cutoff frequency is reduced by 10% according to the disturbance frequency.

[0196] S425: Using the target drive slope setting value and target absorption circuit parameter setting value of all high-voltage side switching devices obtained in S423, and the target differential drive level setting value and target filter cutoff frequency setting value of all low-voltage communication links obtained in S424 as joint inputs, the isolation control command sequence after linkage adjustment is generated through the command sequence encapsulation algorithm and output to the execution unit, which then issues it to the high-voltage side switch drive unit and the low-voltage communication interface unit for execution.

[0197] By employing the linked isolation parameter feedforward adjustment method in S42, the system traces back from the top-level error risk to the bottom-level high-voltage disturbance source along the strong-weak electrical coupling causal lattice structure, simultaneously adjusting key parameters on both the strong and weak electrical sides. This enables coordinated control of the main path leading to a decrease in safety margin at the same time, thus avoiding insufficient compensation or excessive suppression caused by adjustments on only one side. The linked and adjusted isolation control command sequence output by S42 can directly act on the field actuators, achieving closed-loop control from sub-health state perception to proactive intervention.

[0198] Overall, S1 to S4 form a closed-loop chain for monitoring, analyzing, and controlling the coupling of strong and weak electrical systems. S1, by constructing a timing error correlation array, explicitly reveals the weak error patterns partially masked by error correction and retransmission mechanisms as higher-order correlation terms. Furthermore, it embeds local transient features of the strong electrical side into each error event, transforming the error from an isolated codeword statistic into a timing event carrying contextual information about the strong electrical disturbance. This provides a foundation for subsequently locating strong electrical interference based on error data. S2, using the higher-order correlation terms in the timing error correlation array as clues, performs adaptive window segmentation and phase-locked analysis on the strong electrical side's electrical transient alignment data set, constructing a spectral coupling phase chain. This visualizes the phase coupling structure and energy transfer paths between multiple frequency components within the strong electrical side, thus bridging the modeling gap between traditional amplitude spectrum analysis and error phenomena. S3 performs cross-domain isomorphic matching between the set of error mode feature vectors and the set of disturbance coupling feature vectors. It then constructs a strong-weak coupling causal lattice between strong-electrical frequency coupling events, conducted mediation events, and weak-electrical bit error correlation events using a causal sequential inference method. This establishes a mapping relationship between strong-electrical disturbances and communication quality degradation at both the feature space and causal structure levels, which is more effective than simple correlation analysis in avoiding misjudgments caused by accidental synchronization and common causes. S4 constructs a set of safety margin attenuation trajectories based on the strong-weak coupling causal lattice and performs feedforward linkage adjustment along the lattice structure. This allows the control strategy to move beyond relying on a single bit error rate or insulation impedance threshold, instead relying on predictive intervention based on multi-level causal chain lengths and degradation rates, thus shifting from post-event alarms to pre-event control.

[0199] For example, in a scenario where a strong and weak current integrated cabinet operates for a long time, the system first continuously collects the original sequence of the weak current communication differential signal and the original ideal codeword sequence through a bit error monitoring probe, and simultaneously collects the original set of electrical transient data on the strong current side. After processing by S1, a timing error correlation array and a strong current side electrical transient alignment data set are obtained. When frequent load switching causes the ground plane potential drift to gradually intensify within a certain period, several higher-order correlation terms with slowly increasing mode weights will appear in the timing error correlation array. These higher-order correlation terms are selected as key modes in the higher-order correlation term directional query method in S21, and the adaptive frequency window segmentation method in S22 drives the frequency window segmentation result set concentrated in a certain resonant frequency band. Then, under the processing of the instantaneous phase extraction method in S23 and the phase lock detection method in S24, a high-weight directed path from the switching fundamental frequency to the resonant frequency band appears in the spectral coupling phase chain. After performing cross-domain isomorphic matching on the set of error mode feature vectors and the set of disturbance coupling feature vectors, S3 incorporates the high-weighted directed path and the corresponding higher-order error correlation terms into the strong-weak coupling causal lattice as high-confidence candidate causal pairs. This forms the shortest causal chain, pointing from a set of power switching device frequency coupling events, through ground impedance change mediation events, to a specific communication link error cluster. When predicting the safety margin attenuation trajectory of the strong-weak coupling causal lattice, S4 discovers that the length of this shortest causal chain will rapidly shorten in the future, exceeding the safety margin attenuation threshold. Therefore, a linked isolation parameter feedforward adjustment method is used to generate a linked adjustment isolation control command sequence. This moderates the drive slope of the relevant power switching devices and optimizes the absorption circuit parameters. Simultaneously, it increases the differential drive level of the affected communication link and adjusts the filter cutoff frequency, suppressing the degradation trend of each node in the causal chain and restoring the shortest causal chain length to the safe range. Thus, without triggering any malfunctioning protection actions, a proactive intervention against potential insulation degradation and electromagnetic coupling risks is completed.

[0200] Example 2:

[0201] This embodiment, based on Embodiment 1, provides a safety isolation and linkage control system for integrated strong and weak current systems, such as... Figure 6 As shown, it includes:

[0202] Timing error correlation array construction module: used to collect the original differential signal sequence, original ideal codeword sequence and original transient electrical data of the strong current side of the integrated strong and weak current system, and to establish a timing error correlation array through timestamp alignment and event correlation coding;

[0203] Spectrum-coupled phase chain establishment module: Based on the directional query results of the timing error correlation array, it performs adaptive window segmentation and phase coupling identification on the transient data of the high-voltage side to establish a spectrum-coupled phase chain;

[0204] Strong-weak electrical coupling causal lattice construction module: used to perform cross-domain isomorphic matching and causal sequential inference between the timing error correlation array and the spectral coupling phase chain to construct a strong-weak electrical coupling causal lattice;

[0205] Linkage control module: used for safety margin decay trajectory prediction and linkage isolation parameter feedforward adjustment based on strong and weak electrical coupling causal lattice.

Claims

1. A safety isolation and linkage control method for integrated strong and weak current systems, characterized in that, The method includes: The original sequence of differential signals of weak current communication, the original ideal codeword sequence of weak current side and the original transient data of electrical system of strong current integrated system are collected, and the timing error code association array is established by timestamp alignment and event association coding. Based on the directional query results of the timing error correlation array, adaptive window segmentation and phase coupling identification are performed on the transient data of the power supply side to establish a spectrum-coupled phase chain. By performing cross-domain isomorphic matching and causal sequential inference on the timing error correlation array and the spectrum-coupled phase chain, a strong-weak electrical coupled causal lattice is constructed. Safety margin decay trajectory prediction and linked isolation parameter feedforward adjustment based on strong and weak electrical coupling causal lattice.

2. The safety isolation and linkage control method for integrated strong and weak current systems according to claim 1, characterized in that, The steps for establishing the timing error correlation array include: Based on the original sequence of differential signals in weak current communication and the original ideal codeword sequence, the initial set of bit error events is obtained by decoding and comparing with the bit error monitoring probe. Based on the initial set of bit error events and the original set of electrical transient data from the high-voltage side, a unified time reference sequence is generated using a unified time base generation method, resulting in a time-aligned set of bit error events and a time-aligned set of electrical transient data from the high-voltage side. Based on the time-aligned bit error event set and the high-voltage side electrical transient alignment data set, a bit error event base matrix is ​​generated through an event correlation coding method; Based on the error event matrix, a set of higher-order correlation terms is generated using a higher-order correlation term construction method. Based on the fundamental matrix of bit error events and the set of higher-order correlation terms, a timing bit error correlation array is constructed through an array fusion generation method, and the timing bit error correlation array and the set of electrical transient alignment data from the power supply side are output.

3. The safety isolation and linkage control method for integrated strong and weak current systems according to claim 2, characterized in that, The error monitoring probe decoding and comparison method includes: Using the original sequence of differential signals for weak current communication as input, the physical layer signal flip point sequence is extracted by an edge detection algorithm with adaptive threshold, and a physical layer sampling clock sequence is constructed based on the physical layer signal flip point sequence. Using the physical layer sampling clock sequence and the original sequence of the weak current communication differential signal as joint inputs, a decision bit sequence is generated through symbol decision; Using the decision bit sequence and the original ideal codeword sequence as joint inputs, an error flag sequence is generated by bit-by-bit comparison and forward error correction flag extraction rules. Using the error flag sequence and the physical layer sampling clock sequence as joint inputs, an initial set of error events is generated through error event encapsulation rules.

4. A safety isolation and linkage control method for integrated strong and weak current systems according to claim 2, characterized in that, The method for constructing higher-order association terms includes: Using the timestamp column in the bit error event matrix as input, a candidate period set for the bit error time series is obtained through a periodic candidate frequency search algorithm; Using the candidate period set and the timestamp column in the error event matrix as joint inputs, the phase distribution sequence under each candidate period is obtained through phase folding operation; Using the phase distribution sequence as input, a set of periodic clustering results is generated through phase clustering and burstiness assessment algorithms; Using the set of periodic clustering results as input, the codeword distance distribution entropy value of each phase cluster is obtained through codeword distance statistics and information entropy calculation methods; Using the periodic clustering result set and codeword distance distribution entropy as joint inputs, a mode weight value is generated for each phase cluster through mode weight calculation rules. Each phase cluster corresponds to a higher-order association term, which records the set of bit error event indexes and the corresponding mode weight value contained in the phase cluster. Using all the generated higher-order association terms as input, a set of higher-order association terms is generated through set construction operation and combined with a first preset threshold for filtering.

5. A safety isolation and linkage control method for integrated strong and weak current systems according to claim 4, characterized in that, The steps for establishing the spectral coupling phase chain include: Using a timing error correlation array as input, a set of high-order correlation term pointing query results is generated through a high-order correlation term pointing query method; Using the set of high-order correlation term directional query results and the set of electrical transient alignment data from the power supply side as joint inputs, a set of frequency window segmentation results is generated through an adaptive frequency window segmentation method. Using the frequency window segmentation result set as input, an instantaneous phase trajectory set is generated through an instantaneous phase extraction method; Using the instantaneous phase trajectory set as input, a set of frequency point phase coupling relationships is generated through a phase-locked detection method; Using the set of frequency-point phase coupling relationships as input, a spectral coupling phase chain is generated through a chain structure construction method.

6. A safety isolation and linkage control method for integrated strong and weak current systems according to claim 5, characterized in that, The higher-order related item pointing query method includes: Using the set of higher-order correlation items in the timing error correlation array as input, the top several higher-order correlation items with the highest mode weight values ​​are selected by the mode weight sorting algorithm to obtain a subset of candidate higher-order correlation items. Taking the set of error event indices for each higher-order correlation item in the candidate higher-order correlation item subset as input, the corresponding error event timestamp subsequence and codeword index subsequence are obtained by retrieving from the time-series error correlation array; Using the timestamp subsequence of bit error events as input, the implicit period estimate is obtained through time interval statistics and frequency estimation algorithms; Using the error event timestamp subsequence, codeword index subsequence, pattern weight value, and implicit period estimation value as joint inputs, a set of high-order related item directional query results is generated through result encapsulation rules.

7. A safety isolation and linkage control method for integrated strong and weak current systems according to claim 6, characterized in that, The steps for constructing a strongly and weakly coupled causal lattice include: Using a timing error correlation array as input, a set of error mode feature vectors is generated through an error mode feature extraction method. Using the spectral coupling phase chain as input, a set of perturbation coupling feature vectors is generated through a perturbation coupling feature extraction method; Using the set of error pattern feature vectors and the set of perturbation coupling feature vectors as joint inputs, a set of candidate causal relationship pairs is generated through a cross-domain isomorphic matching method. Using a set of candidate causal relationship pairs as input, a strongly and weakly coupled causal lattice is generated through a causal sequential reasoning method.

8. A safety isolation and linkage control method for integrated strong and weak current systems according to claim 7, characterized in that, The causal sequential reasoning method includes: Using the candidate causal relationship pair set as input, causal relationship pairs with matching confidence levels lower than a second preset threshold are filtered out by thresholding to obtain a high-confidence candidate causal relationship pair set; Using the set of high-confidence candidate causal relationship pairs and the path hierarchy relationship in the spectrum-coupled phase chain as joint input, the causal relationship pairs belonging to the same strong-electric-side coupling path are merged into a set of low-level strong-electric-side frequency point coupling event nodes through the low-level event aggregation algorithm; Using the set of high-confidence candidate causal relationship pairs and the higher-order correlation terms of bit error in the timing error correlation array as joint inputs, the top-level event aggregation algorithm merges the causal relationship pairs corresponding to the same bit error mode into a top-level weak current side bit error correlation event node set. Taking the causal mapping relationship between the set of frequency coupling event nodes on the bottom-level high-voltage side and the set of bit error association event nodes on the top-level low-voltage side as input, a multi-level intermediate transmission mediation event node set is introduced through a transmission mediation mining algorithm. Using the set of frequency-coupled event nodes on the bottom strong electrical side, the set of multi-level conduction mediation event nodes in the middle layer, and the set of bit error association event nodes on the top weak electrical side as joint inputs, a set of partial order relations that satisfies reflexivity, antisymmetry, and transitivity is established through a lattice structure construction algorithm. This set of partial order relations is then organized into a strong-weak electrical coupled causal lattice with supremum and infimum operations.

9. A safety isolation and linkage control method for integrated strong and weak current systems according to claim 8, characterized in that, The steps for predicting the safety margin attenuation trajectory and adjusting the linked isolation parameter feedforward include: Using a strong-weak coupled causal lattice as input, a set of safety margin decay trajectories is generated through a safety margin decay trajectory prediction method. Using the safety margin attenuation trajectory set and the strong-weak electrical coupling causal lattice as joint inputs, the isolation control command sequence after linkage adjustment is generated through the linkage isolation parameter feedforward adjustment method.

10. A safety isolation and linkage control system for integrated strong and weak current systems, used to implement the safety isolation and linkage control method for integrated strong and weak current systems according to any one of claims 1-9, characterized in that, The system includes: Timing error correlation array construction module: used to collect the original sequence of differential signals of weak current communication in the integrated strong and weak current system, the original ideal codeword sequence of the weak current side and the original transient data of electrical equipment on the strong current side, and to establish a timing error correlation array through timestamp alignment and event correlation coding; Spectrum-coupled phase chain establishment module: Based on the directional query results of the timing error correlation array, it performs adaptive window segmentation and phase coupling identification on the transient data of the high-voltage side to establish a spectrum-coupled phase chain; Strong-weak electrical coupling causal lattice construction module: used to perform cross-domain isomorphic matching and causal sequential inference between the timing error correlation array and the spectral coupling phase chain to construct a strong-weak electrical coupling causal lattice; Linkage control module: used for safety margin decay trajectory prediction and linkage isolation parameter feedforward adjustment based on strong and weak electrical coupling causal lattice.

Images