Multi-channel grounding grid corrosion detection synchronous measurement system and method

Abstract

The application discloses a multi-channel grounding grid corrosion detection synchronous measurement system and method, relates to the technical field of grounding grid corrosion detection, and realizes rapid round patrol of grounding grid port pairs through a preset channel switching strategy, constructs an original port response matrix in real time and extracts an impedance time domain sequence; topological mismatch false corrosion branches are removed by using fingerprint hashing and adjacency similarity; when the observability is insufficient, new measurement points are automatically planned and supplementary measurement is performed; after multi-batch data is fused, a sparse preservation optimization model is used to incrementally reconstruct the grounding grid topology, and consistency is verified by resistance constraint maximum subgraph isomorphism; finally, convergence is determined according to a comprehensive index of three factors of residual error reduction, topology matching and observability, and the final corrosion branch and its confidence whose impedance increment exceeds a preset impedance index are output. The system has high measurement parallelism, low false alarm rate and scalability, and is suitable for online diagnosis and operation and maintenance decision of a large-scale grounding grid at a transformer substation.

Description

Technical Field

[0001] This invention relates to the field of grounding grid corrosion detection technology, and more specifically, to a multi-channel grounding grid corrosion detection synchronous measurement system and method. Background Technology

[0002] Grounding grids are crucial barriers for ensuring the safety of personnel and equipment in substations, power plants, and large industrial sites. Due to long-term exposure to acidic and alkaline soils, stray currents, and welding defects, the conductor cross-section thins year by year. If not detected in time, this can lead to increased contact potential, relay protection malfunctions, and even equipment insulation breakdown. Traditional corrosion detection methods generally employ single-point measurement or manual segmented excavation: the former relies on a small number of fixed measuring points, has low spatial resolution, and is susceptible to electromagnetic interference; the latter, while intuitive, requires power outages for excavation, resulting in high costs, long cycles, and significant impacts on operational safety. Although multi-channel injection and synchronous measuring instruments have emerged in recent years, they still suffer from three major shortcomings: First, the number of measurement channels is limited, differing from the actual number of grounding leads by tens of times, leading to a severely sparse port impedance matrix and an inability to locate localized corrosion; second, construction errors cause discrepancies between the on-site topology and design drawings, and simple threshold comparisons can easily misjudge topology mismatch as corrosion; third, existing systems mostly perform one-time scans, lacking the closed-loop capability to dynamically supplement measuring points based on diagnostic results and improve observability. With the development of smart grids and digital substations, the industry urgently needs a comprehensive solution capable of rapid scanning of large-scale grounding grids, filtering of false corrosion, adaptive addition of measurement points, and incremental topology reconstruction under uninterrupted power supply and minimal human intervention. This solution aims to reduce false alarms and missed alarms and provide quantifiable corrosion assessments. This invention addresses these needs by implementing millisecond-level round-robin measurements through a preset channel switching strategy, using wavelet packet-LSH fingerprinting to filter out topology artifacts, employing integer programming to automatically add measurement points, and introducing sparsity preservation optimization and subgraph isomorphism verification. Ultimately, this results in a list of corrosion branches with confidence levels, overcoming the shortcomings of existing technologies in measurement coverage, topology consistency verification, and diagnostic reliability.

[0003] To address the above problems, this invention proposes a solution. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-channel grounding grid corrosion detection synchronous measurement system and method to solve the problem that the grounding grid topology may deviate from the design drawings during actual construction, resulting in frequent structural misjudgments and false corrosion diagnoses due to topology mismatch. This invention also addresses the lack of a method to quickly improve the observability of suspected corrosion branches by supplementing with minimal measurement points, ensuring stable and reliable inversion diagnosis results, and automatically eliminating false corrosion diagnoses caused by topology mismatch.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A multi-channel grounding grid corrosion detection and synchronous measurement system, characterized in that it includes the following modules executed in sequence:

[0007] The dynamic rotation measurement acquisition module, based on a preset channel switching strategy, performs rotational measurements on port pairs between multiple grounding leads in the substation grounding network; it obtains port voltage signals and port injected current values ​​to construct the original port response data matrix and extracts the impedance time sequence through wavelet packet transform.

[0008] The pseudo-corrosion fingerprint filtering module, based on the original port response data matrix and the historical design topology of the grounding grid, compares the real-time impedance value in the impedance time series with the theoretical impedance value in the historical topology, and identifies the initial abnormal branch set through covariance correlation analysis; it filters pseudo-corrosion branches through the fingerprint similarity of adjacent branches and a preset low-amplitude threshold, and outputs the real suspected corrosion set.

[0009] The observability optimization and expansion module calculates the observability index of a single branch and the global diagnostic quality index for the set of real suspected corrosion branches. When the observability index of a single branch is lower than the preset minimum observability threshold, it is recorded as a low observability abnormal branch. When the global diagnostic quality index is lower than the preset standard observability threshold, the measurement point supplementation process is triggered to perform supplementary measurements.

[0010] The incremental topology reconstruction module integrates the original port response data moments and the supplementary response data, and generates an updated port impedance tensor through a self-weighted iterative Kalman filter; it then uses a sparsity-preserving optimization model to solve for the reconstructed grounding grid topology; the reconstructed grounding grid topology is compared with the historical design topology of the grounding grid using a resistance-constrained maximum subgraph isomorphism algorithm, and outputs the new grounding grid topology after comparison and verification.

[0011] The inversion convergence determination module calculates the comprehensive stability index based on the weighted average of the residual descent factor, topology matching factor, and observability factor. If the comprehensive stability index exceeds the preset threshold, a second least squares inversion is performed on the pseudo-corroded branch set. Branches with impedance increments exceeding three standard deviations of the health benchmark are included in the final corrosion set, and a triplet column of branch number, impedance increment, and confidence level is output.

[0012] As a further aspect of the present invention: the operation of the dynamic rotation measurement acquisition module includes: preloading the channel switching strategy file to the static memory; controlling the relay to close the specified port pair according to the instruction field sequence in the preloaded channel switching strategy; within the signal stability window, injecting current value and acquiring port voltage signal, the number of sampling points being determined by the sampling frequency and window duration; performing triple signal purification on the voltage signal: phase-locked demodulation to separate the fundamental component, FIR filter to suppress high-frequency noise, and statistical outlier removal; generating matrix elements from the purified data according to the impedance conversion formula, and converting the matrix into an impedance time sequence through wavelet packet transformation.

[0013] As a further aspect of the present invention: the determination logic of the pseudo-corrosion branch in the pseudo-corrosion fingerprint filtering module includes: based on the original port response data matrix According to the branch index The impedance observations of the same branch in each measurement frame are extracted into an impedance time-series vector. And record it as the first test sample;

[0014] The topology was designed based on the system's pre-stored grounding grid history. Ideal port impedance tensor obtained through offline calculation using electromagnetic simulation And denoted as the second test sample; where V is the grounding lead port number, and E is the edge set of the grounding grid, i.e. the grounding conductor branch; the deviation matrix ΔZ is obtained by subtracting the mean of the first test sample from the second test sample;

[0015] For each branch e, the deviation matrix is ​​projected onto the inverse covariance space and its magnitude is taken. If the module length Branches exceeding the percentile threshold of the chi-square distribution in the same dimension are listed as initial abnormal branches and included in the set to obtain the initial abnormal branch set. ; for the impedance timing sequence corresponding to the initial abnormal branch Where e is the unique number of the branch; Maximum overlap discrete wavelet packet transform and locality-sensitive hashing are performed to generate a fixed-length bit string. This bit string is considered a structural fingerprint identifier; for each branch... If the two branches are in the historical design topology of the grounding grid If a path shares the same node (i.e., has a common endpoint), it is considered an adjacent branch. The Hamming distance between these branches is calculated and recorded as a fingerprint similarity. The number of similar adjacent branches is then counted. ;

[0016] Traverse all Adjacent branch Where r is the sequence number of the adjacent branch; if and Then count Add 1; and compare the number of similar adjacent branches with the preset minimum neighborhood number threshold;

[0017] like and Then the branch road Add pseudo-corrosion branch set Among them, the pseudo-corrosion branch set ;

[0018] From the initial screening of abnormal branch sets By removing pseudo-corrosion branches, a set of real suspected corrosion branches is generated. .

[0019] As a further aspect of the present invention: the impedance timing sequence corresponding to the initial abnormal branch. Perform maximum overlap discrete wavelet packet transform and locality-sensitive hashing to generate fixed-length bit strings. Specifically, in order to extract features that take into account both the time and frequency domains, the impedance time series corresponding to the initial abnormal branch is analyzed. Where e is the unique number of the branch; perform maximum overlap discrete wavelet packet transform to decompose each impedance time series into multiple sub-bands, and after calculating the variance, kurtosis, and kurtosis of the energy statistics in each sub-band, concatenate them into a fixed-length vector. These fixed-length vectors are first normalized using Z-Score: subtract the sample mean from each dimension and divide by the sample standard deviation; then, the locality-sensitive hashing method is used to normalize the fixed-length vectors. Perform a second hash to generate a bit string of fixed length. This bit string is considered a structural fingerprint identifier and is composed of characteristics such as impedance deviation and frequency domain energy.

[0020] As a further aspect of the present invention: the measurement point supplementation process in the observability optimization and expansion module includes: defining the observability of a single branch as the ratio of the minimum to the maximum singular value of the impedance sequence, and the global diagnostic quality index as the mean observability of all suspected branches; determining the priority of measurement points by ranking them according to node betweenness centrality, and generating a candidate node set by combining the shortest wiring path and cabinet terminal capacity constraints; constructing an integer linear programming model with the goal of minimizing the number of measurement points added, satisfying the three constraints of observability threshold, total wiring length, and number of terminals; inputting the integer programming model into a preset mixed integer linear programming solver to obtain the optimal expansion or rearrangement scheme. The proposed solution includes both adding new ports and rearranging channels.

[0021] As a further aspect of the present invention: the incremental topology reconstruction module uses a sparse-preserving optimization model to solve for the reconstructed grounding grid topology. The specific process includes: defining the original port response data matrix. Number of lines Supplementing the response data matrix Number of lines Zero-padding alignment is performed on the timeline, and the components are stitched together into a three-dimensional tensor. ;in, For three-dimensional tensors Index on the timeline;

[0022] The self-weighted iterative Kalman filter is introduced for auxiliary fusion. The logic of the self-weighted iterative Kalman filter is as follows: take the branch resistance vector. The system state is represented by T, where T is the transpose sign and N is the total number of branches. As the observation equation; where Given the port-branch correlation, q is the number of iterations in the Kalman filter algorithm; As a weight matrix The noise variance of each frame is estimated online; after fusion, the port impedance tensor is updated at the output. ;

[0023] remember The equivalent admittance matrix of the historical topology is determined; a sparsity-preserving incremental optimization model is adopted, based on the formula: ;in, Given a nonlinear mapping function, β is a preset tradeoff coefficient that controls the balance between preserving the original topological sparsity and allowing local modifications;

[0024] The incremental optimization model is solved using the existing iterative reweighted least squares algorithm to obtain the reconstructed admittance matrix. This allows for the direct generation of the reconstructed grounding grid topology. .

[0025] As a further aspect of the present invention: the comprehensive stability index in the inversion convergence determination module is generated as follows: the residual decrease factor is the weighted residual change rate of the most recent two rounds of reconstruction results; the topology matching factor is the final consistency score output by the incremental reconstruction module; the observability factor is the normalized value of the global diagnostic quality index of the optimization and expansion module; the three factors are weighted and summed and compared with the stability threshold. When the comprehensive stability index is greater than or equal to the preset stability threshold, it is determined to meet the standard and the final corrosion confirmation is initiated.

[0026] A synchronous measurement method for corrosion detection of multi-channel grounding grids includes the following steps: S1 Dynamic rotation measurement and acquisition: Load the channel switching strategy file, rotate and measure the port impedance and generate a time-domain response sequence;

[0027] S2 False Corrosion Filtering: Compares real-time data with historical topology and eliminates false corrosion branches through fingerprint similarity analysis;

[0028] S3 observability optimization: When diagnostic indicators are below the threshold, automatically plan a measurement point expansion scheme and perform supplementary measurements;

[0029] S4 Incremental Topology Reconstruction: Reconstructs the grounding grid topology by fusing data from multiple batches, and outputs a new topology structure after isomorphic verification;

[0030] S5 Convergence Determination: Trigger the final erosion branch inversion based on the stability index and output a triplet diagnostic report.

[0031] As a further aspect of the present invention: the final corrosion confirmation in step S5 includes: performing secondary inversion only on branches of the pseudo-corrosion set to eliminate topology mismatch interference; determining branches with impedance increments exceeding the healthy baseline mean + 3 times the standard deviation as true corrosion; generating triplets according to branch number, impedance increment, and confidence level for priority processing by the operation and maintenance system.

[0032] The beneficial effects of this invention are:

[0033] (1) In this invention, the time-domain response sequence is extracted by wavelet packet transform of the dynamic round-robin acquisition module, which effectively suppresses electromagnetic noise and white noise interference on site and significantly improves the identification of impedance fluctuation characteristics; the pseudo-corrosion fingerprint filtering module introduces the structural fingerprint similarity comparison mechanism of adjacent branches to distinguish between real physical corrosion and topology design error from the root, and avoids misjudgment caused by drawing mismatch; the incremental topology reconstruction module integrates multiple batches of data and adopts resistance-constrained subgraph isomorphism verification to ensure that the topology changes caused by corrosion are accurately captured, and the consistency between the reconstruction result and the real grounding grid state is greatly improved.

[0034] (2) In this invention, the observability optimization and expansion module makes autonomous decisions on the expansion of measurement points and channel rearrangement scheme based on the branch diagnosis credibility index, breaking through the blind zone limitation of the traditional fixed measurement point layout; the inversion convergence judgment module realizes the intelligent assessment of topological stability and adaptive control of iteration termination through the synergistic weighting of residual reduction factor, topology matching factor and observability factor, which greatly reduces the number of invalid measurement rounds; finally, the triplet corrosion report branch number / impedance increment / confidence level is output, providing a highly credible basis for operation and maintenance decision-making, and significantly reducing the cost of manual review and downtime maintenance time. Attached Figure Description

[0035] The invention will now be further described with reference to the accompanying drawings.

[0036] Figure 1 This is a schematic diagram of the system framework structure of the multi-channel grounding grid corrosion detection and synchronous measurement system of the present invention;

[0037] Figure 2 This is a schematic diagram of the method flow for monitoring grouting pipelines according to the present invention;

[0038] Figure 3 This is a schematic diagram of the dynamic rotation measurement module in the multi-channel grounding grid corrosion detection synchronous measurement system of the present invention;

[0039] Figure 4 This is a schematic diagram of the pseudo-corrosion fingerprint filtering module in the multi-channel grounding grid corrosion detection and synchronous measurement system of the present invention.

[0040] Figure 5 This is a schematic diagram of the observability optimization and expansion module in the multi-channel grounding grid corrosion detection synchronous measurement system of the present invention. Detailed Implementation

[0041] 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. Example

[0042] Please see Figure 1 As shown, this invention is a multi-channel grounding grid corrosion detection and synchronous measurement system, comprising the following modules:

[0043] like Figure 3 As shown, the dynamic rotation measurement and acquisition module is used to dynamically rotate and acquire the equivalent port resistance values ​​between each grounding lead in the substation grounding network according to the preset channel switching strategy, acquire port voltage and injected current data and construct the original port response data matrix to form a time-domain response sequence that can be used for corrosion identification.

[0044] It should be noted that this invention, by employing a preset channel switching strategy, dynamically measures any port pair between N grounding leads while ensuring that the existing hardware framework remains unchanged; then, it analyzes the measured port voltage signals. With the injected current value Real-time parsing and multi-level filtering are performed to ultimately generate the raw port response data matrix. and the time-domain response sequence derived from the matrix. This is provided for subsequent corrosion identification algorithms to call; the preset channel switching strategy is a set of channel switching instructions, which is used to indicate the port pairs that need to be closed in different time periods and their holding duration; the present invention does not impose any limitations on the generation method of the instruction set;

[0045] Where t is the time index, i is the port index, and j is the port index other than i. , This represents the measurement path from port i to port j, where N is the total number of ports in the grounding network under test.

[0046] The preset channel switching strategy is stored in the form of a pre-generated strategy file, which is abstracted as a Switch_Table data table. The physical format of the Switch_Table includes, but is not limited to, JSON, CSV, or binary byte stream. The strategy file is composed of several Switch_Element elements in sequence. The field structure of each element is defined by the following data structure: InjectID field, SenseID field, Setting field, and Dwell field. Among them, InjectID is the excitation channel number, SenseID is an array of measurement channel numbers whose array length is limited by the maximum number of channels in the system, Setting is the channel hardware parameter configuration, and Dwell is the signal stabilization waiting time.

[0047] The loading process based on the aforementioned preset channel switching strategy includes: after the system is powered on or reset, the controller executes the following sub-steps to obtain the set of channel switching instructions required for subsequent analysis. This step does not limit the generation method of the strategy, but only requires that it can be recognized and invoked by the system. The sub-steps include:

[0048] The controller reads a preset channel switching strategy file Switch_Table via USB, SD card, network interface or on-chip non-volatile memory; the Switch_Table can be in text format such as CSV, JSON, XML or binary format such as BIN, HEX, and this embodiment does not limit it;

[0049] After reading, the controller calls Cyclic Redundancy Check (CRC) or message digest algorithms such as MD5 or SHA-256 to perform an integrity comparison of Switch_Table; if the verification result is true, it enters the subsequent parsing process; if it is false, it triggers an error interrupt and sends a reload command to the host computer.

[0050] Write each entry of Switch_Table into the on-chip static random access memory to form a one-to-one corresponding instruction array Switch_Array[a], where a is a natural number index; at the same time, record the strategy length L = length(Switch_Table);

[0051] Set the channel switching pointer ptr to an initial value of 1; when At that time, the system determines that there are still unexecuted switching instructions; when At this point, the system determines that the current scan has ended;

[0052] When the above operations are completed and there are no errors, the controller sets the run flag RUN_FLAG = 1 so that the main loop of dynamic measurement can detect whether it can enter the measurement stage.

[0053] The dynamic polling acquisition logic relies on a pre-loaded channel switching strategy Switch_Table. It sequentially traverses the strategy elements using the channel switching pointer ptr and controls relay switching and the data acquisition window based on the port combinations and timing parameters within the elements. The specific process is as follows:

[0054] The controller reads Switch_Table[ptr] and drives the matrix switching unit to close the corresponding port via the communication bus; it times according to the Settling instruction field, for example, 10 ms; during this period, it suppresses contact bounce and inrush current; the constant current source is based on the injected current. The output and feedback currents are simultaneously acquired by the ADC; within a Dwell window of 30 ms, the constant current source outputs according to the reference amplitude configured by the system, and the feedback injection current... Synchronous sampling is performed, and the port voltage signal ΔV(t) is acquired in parallel at a set sampling frequency within the Dwell window. The number of sampling points M is determined by the sampling frequency and the Dwell. The port voltage signal ΔV(t) is purified by a triple operation of phase-locked demodulation, FIR filtering, and anomaly rejection to obtain a highly reliable effective voltage value. Assuming a scan executes T frames in total, then the original port response data matrix... The number of rows T = L, the number of columns Globally unique Total combination, and ;

[0055] Row index = instruction sequence number;

[0056] Column index = globally unique Combinations are hashed to ;

[0057] The impedance reduction formula is used to generate a vector after each frame. The impedance conversion formula is as follows:

[0058] Where k is the frame index, i.e., the sequence number, of the dynamic round-robin measurement, and m is the measurement channel index and P is the maximum number of channels allowed by the system;

[0059] The vector will be generated after the conversion is complete. Write raw port response data matrix In the k-th row; if The process continues until the next frame is reached; otherwise, the current scan ends.

[0060] It should be noted that the triple signal purification operation is an existing mature technology, and the process will not be described in detail in this embodiment.

[0061] The specific process of further constructing the original port response data matrix to form a time-domain response sequence that can be used for corrosion identification is as follows: The original port response data matrix... Restored to a three-dimensional tensor ;where PortPair(m) is the reverse lookup from column index to port pair; for each sequence Using a mother wavelet with symmetric orthogonal properties, the decomposition depth depends on the required bandwidth resolution; several energy coefficient vectors of the final layer and their time-varying coefficient sequences are obtained; based on the coefficient sequences, the time-domain response sequence is defined as denoted as... Where E represents the expected value of the energy of all last-layer nodes, and C is the number of wavelet decomposition layers in the maximum overlap discrete wavelet packet transform. It reflects the instantaneous impedance changes of the port pair across multiple frequency bands and has the advantages of suppressing white noise and preserving the slow-change characteristics of corrosion.

[0062] like Figure 4 As shown, the pseudo-corrosion fingerprint filtering module compares the constructed original port response data matrix with the historical design topology of the grounding grid, extracts the initial abnormal branch response features to obtain the structural fingerprint identifier of each corrosion branch, and outputs a set of suspected corrosion branches; the real suspected corrosion branches are pseudo-corrosion branches caused by topological differences.

[0063] Based on the raw port response data matrix obtained from the previous step of dynamic polling. According to the branch index The impedance observations of the same branch in each measurement frame are extracted into an impedance time-series vector. This is recorded as the first test sample. This sample reflects the actual fluctuation of the grounding grid resistance under real-time operating conditions. To facilitate consistent referencing of data from the acquisition layer port and the model from the topology layer node, this embodiment establishes a one-to-one mapping relationship between the two starting from this step, and uniformly indexes them using the unique branch number e.

[0064] The topology was designed based on the system's pre-stored grounding grid history. Ideal port impedance tensor obtained through offline calculation using electromagnetic simulation And denoted as the second test sample; where V is the grounding lead port number and E is the edge set of the grounding grid, i.e. the grounding conductor branch; this sample describes the theoretical value of the branch impedance under the design reference state or healthy state; in order to measure the difference between the first and second test samples, the two are mapped to a corresponding relationship, i.e., the deviation matrix ΔZ is obtained by subtracting the second test sample from the mean of the first test sample.

[0065] To consider the cross-correlation between each port pair, we first need to calculate the covariance matrix Σ over the entire bias matrix ΔZ, and then use the singular value truncation technique to remove the ill-conditioned features from the covariance matrix Σ to obtain a covariance matrix with a suitable condition number.

[0066] Specifically, for each branch e, the deviation matrix is ​​projected onto the inverse covariance space and its magnitude is taken. If the module length Branches exceeding the percentile threshold of the chi-square distribution in the same dimension are listed as initial abnormal branches and included in the set to obtain the initial abnormal branch set. This process is equivalent to the traditional Mahalanobis distance test, but the algorithm is only called as a library function, and the specific matrix operation details are not described in the manual.

[0067] To assess the contribution of a single branch to the overall error, the mean of the absolute deviation of that branch is calculated, and all abnormal branches are sorted in descending order of this value to generate an abnormal impact value sequence table. The branches that appear earlier in the list have larger numerical deviations and more direct impacts, and will receive higher processing priority in subsequent stages.

[0068] To extract features that take both the time and frequency domains into account, the impedance time series corresponding to the initial anomalous branch is analyzed. Where e is the unique number of the branch; perform maximum overlap discrete wavelet packet transform to decompose each impedance time series into multiple sub-bands, and calculate energy statistics such as variance, kurtosis, and kurtosis in each sub-band before concatenating them into a fixed-length vector. These fixed-length vectors are first normalized using Z-Score: subtract the sample mean from each dimension and divide by the sample standard deviation; this operation ensures that subsequent distance metrics reflect only morphological differences and are no longer affected by absolute magnitude; locality-sensitive hashing is then used to normalize the fixed-length vectors. Perform a second hash to generate a bit string of fixed length. This bit string is considered a structural fingerprint identifier and is composed of characteristics such as impedance deviation and frequency domain energy;

[0069] For the initial set of abnormal branches Each branch road Extract its structural fingerprint identifier If there are two branches Historical design topology of grounding grid If a path shares the same node (i.e., has a common endpoint), it is considered an adjacent branch, and its fingerprint similarity is calculated:

[0070] ;in, Fault fingerprint feature vector The Hamming distance value, which is the number of different dimensions between two vectors;

[0071] The calculated fingerprint similarity is compared with a preset similarity threshold, and the abnormal impact values ​​of the two branches are compared with a preset low amplitude threshold.

[0072] when And the abnormal impact values ​​of the two branches All meet When the two branches are found to be abnormally homologous, the abnormality is caused by topological mismatch as shown in the drawing error, rather than actual physical corrosion.

[0073] For each branch road Count the number of similar adjacent branches. ;

[0074] Traverse all Adjacent branch Where r is the sequence number of the adjacent branch; if and Then count Add 1; and compare the number of similar adjacent branches with the preset minimum neighborhood number threshold;

[0075] like and Then the branch road Add pseudo-corrosion branch set Among them, the pseudo-corrosion branch set ;

[0076] From the initial screening of abnormal branch sets By removing pseudo-corrosion branches, a set of real suspected corrosion branches is generated. ;in, Branches in the network that exhibit abnormal isolation or high amplitude characteristics need to be included in the subsequent observability assessment and topology reconstruction process. Branches in the process are marked as design errors, and the corrosion diagnosis process is terminated to save computing power.

[0077] Please see Figure 5 The observability optimization and expansion module obtains the detection index of the suspected corrosion branch through data analysis based on the set of suspected corrosion branches, and compares each detection index with the preset observability threshold. Based on the comparison results, in the corresponding grounding network historical design topology, through betweenness centrality analysis and on-site wiring accessibility constraints, it automatically plans to add new measurement points or adjust the original measurement point wiring scheme, and performs a new round of measurement and acquisition to obtain the supplementary response data matrix.

[0078] Based on the set of real suspected corrosion branches obtained in the previous stage An adaptive optimization method for measurement points based on observability-closed-loop distribution is proposed; its core lies in the fact that the detection quantity index includes:

[0079] First, let's look at the primary detection metric: single-branch observability. The diagnostic reliability of each suspected branch was recorded; among them, 0 ;

[0080] Then, the arithmetic mean of the observability of all suspected branches is taken as the overall diagnostic quality benchmark, using the second detection index. ;

[0081] By using threshold comparison and integer programming, the system automatically generates schemes for adding or rearranging measurement points, thereby obtaining supplementary response data and achieving rapid closed-loop improvement in observability. The following is a detailed processing flow for the six stages: sample definition, index measurement, threshold discrimination, measurement point addition, supplementary measurement, and closed-loop update.

[0082] S1: First detection measure The observability of a single branch is defined as the diagnosability of each suspected branch in this scan under the current measurement point layout. The observability of a single branch is defined as the ratio of the minimum singular value to the maximum singular value of the impedance sequence.

[0083] Second detection quantity indicator This serves as a global diagnostic quality benchmark for candidate node placements, comprehensively considering factors such as node betweenness centrality, shortest wiring path, and number of remaining terminals; both are derived from real-time measurements or on-site constraints, eliminating the need for manual estimation.

[0084] S2: The first detection quantity index The first detection quantity index and the preset minimum observability threshold for each branch in the data. Comparison, if a certain branch < If it is, then it is recorded as a low-observability abnormal branch; and the second detection index is... Compare with a preset standard observability threshold; when When this happens, the measurement point supplementation process is triggered;

[0085] S3: For each branch marked as a low observability anomaly, its endpoint nodes are first sorted according to betweenness centrality. The higher the centrality, the more shortest paths the endpoint node is on, and the greater the potential benefit of adding measurement points. Then, the shortest wiring path from the endpoint node to the nearest measurement cabinet is calculated and compared with field constraints such as maximum wiring length and no-wiring zones to obtain a candidate set of nodes.

[0086] S4: Construct an integer linear programming model with the objective of minimizing the number of additional measurement points:

[0087] The decision variable is whether to add nodes. Where, add distribution = 1, skip = 0;

[0088] Constraint 1: All low-observability branches must satisfy the following after the addition of branches. ;

[0089] Constraint 2: Total wiring length Not exceeding the current total workable length;

[0090] Constraint 3: Number of cabinet terminals Σ The remaining terminal capacity must not be exceeded.

[0091] The integer programming model is input into a mixed-integer linear programming solver to obtain the optimal distribution / rearrangement scheme. This can include both adding new ports and rearranging channels; that is, when existing idle measurement points can be transferred for use, the algorithm automatically regards rearrangement as a lower-cost decision than increasing the number of ports.

[0092] S5: According to the distribution / rearrangement scheme After the on-site wiring is completed, the controller automatically updates the channel switching instruction table to include the newly added measurement points in the round-robin testing; then it performs a complete scan to collect the supplementary response data matrix. The supplementary response data matrix With the original port response data matrix After merging, the observability sequence is recalculated to form the updated first detection measure index. Second detection quantity index ;

[0093] S6: If all low-observable anomalies are in the first detection quantity index The value is greater than or equal to the preset minimum diagnosable threshold. And the second detection quantity indicator When the observability threshold is greater than or equal to the preset standard threshold, the system outputs a signal indicating that the measurement point optimization is complete; otherwise, it will... , As a new detection quantity indicator, return to step S2 to start the next round of measurement point planning; repeat this process, and within a limited number of iterations, the observability of all suspected corrosion branches can be improved to the diagnostic usability range.

[0094] The incremental topology reconstruction module forms an updated port response data matrix based on the supplementary response data and the original measurement data, and then performs incremental reconstruction of the historical design topology of the grounding grid to obtain the reconstructed grounding grid topology. The topology changes caused by corrosion are reflected in the reconstruction results. The maximum subgraph isomorphism algorithm is used to compare and verify the reconstructed grounding grid topology with the historical design topology of the grounding grid, and outputs the verified new grounding grid topology.

[0095] Supplementary response data matrix obtained from observability optimization augmentation module Multi-source data fusion and topology reconstruction are performed, specifically as follows:

[0096] Define the raw port response data matrix Number of lines Supplementing the response data matrix Number of lines First, perform zero-padding alignment on the timeline and then stitch them together into a three-dimensional tensor. ;in, For three-dimensional tensors Index on the timeline;

[0097] To account for the measurement uncertainties of the two batches of data, a self-weighted iterative Kalman filter is introduced for auxiliary fusion. The logic of the self-weighted iterative Kalman filter is as follows:

[0098] Take the branch resistance vector The system state is represented by T, where T is the transpose sign and N is the total number of branches. As the observation equation; where Given the port-branch correlation, q is the number of iterations in the Kalman filter algorithm; As a weight matrix The noise variance of each frame is estimated online; after fusion, the port impedance tensor is updated at the output. ;

[0099] remember The equivalent admittance matrix of the historical topology is determined; a sparsity-preserving incremental optimization model is adopted, based on the formula: ;in, Given a nonlinear mapping function, β is a preset tradeoff coefficient that controls the balance between preserving the original topological sparsity and allowing local modifications;

[0100] The incremental optimization model is solved using the existing iterative reweighted least squares algorithm to obtain the reconstructed admittance matrix. This allows for the direct generation of the reconstructed grounding grid topology. ;

[0101] Based on the improved VF2 algorithm, the grounding grid topology is reconstructed. Historical design topology of grounding grid Perform isomorphic search and extend resistance constraint: only when When edge e is considered to be matchable, the algorithm outputs the set of matching edges. Unmatched edge set The unmatched edge set includes newly added or disconnected edges.

[0102] Furthermore, the present invention also includes the following topology verification process: extracting the historical design topology edge set of the grounding grid. Edge set of the reconstructed grounding grid topology ; Traverse line by line If in If an isomorphic mapping exists, then that branch is included in the matching count. And accumulate its relative impedance error After the traversal is complete, calculate the edge matching rate. The average impedance error is calculated by ratioing the impedance error to the number of branches included in the matching factor. like =0, then take =1; the final consistency score is obtained by calculating the base-side matching rate and average impedance error. According to the formula When the final consistency score Greater than or equal to the preset score threshold When the reconstructed grounding grid topology is deemed to have passed verification, it is output as the new grounding grid topology. Otherwise, it will automatically proceed to measurement point optimization and the next round of supplementary data collection.

[0103] The inversion convergence determination module performs stability judgment on the generated new grounding grid topology. When the new grounding grid topology meets the stability conditions, it re-inverts the resistance state of each pseudo-corrosion branch and generates the final corrosion branch set. The stability conditions refer to: residual decrease factor, topology matching factor, and observability factor.

[0104] The logic for obtaining the stability satisfaction condition is as follows: Based on the two most recent reconstruction results obtained by the incremental topology reconstruction module and performing data analysis on them, the following is obtained: The residual decrease factor is obtained by weighting the residuals of the two most recent reconstruction results obtained by the incremental topology reconstruction module; The final consistency score of the grounding wire used is obtained by statistically analyzing the final consistency score obtained by the incremental topology reconstruction module and used as the topology matching factor; The first detection quantity index obtained by the observability optimization and expansion module is obtained and normalized to 0-1 to obtain the observability factor.

[0105] The comprehensive stability index is obtained by weighted linear combination of the above parameter factors;

[0106] The comprehensive stability index is compared with the preset stability threshold. When the comprehensive stability index is greater than or equal to the preset stability threshold, a stability flag is generated, indicating that the topology is stable and the final corrosion confirmation can proceed.

[0107] Conversely, if the observability optimization and expansion module is not found, the measurement points will be replanned.

[0108] After the stability flag is triggered, only the false corrosion set E_false is subjected to a second least squares inversion. If the impedance increment obtained in this round of inversion is greater than or equal to the mean of branch e under the preset health benchmark plus three times the standard deviation of the resistance value of branch e under the preset health benchmark, then branch e is placed into the final corrosion set E_final. Each branch e that is determined to be real corrosion is packaged and output in the form of a triple. The triple includes: the unique branch number, the impedance increment of the branch relative to the historical benchmark, and the system's confidence level for this judgment. The generated triple list can either appear directly in the diagnostic report or be sorted by confidence by the operation and maintenance system to prioritize on-site verification or repair work.

[0109] In this embodiment, the time-domain response sequence is extracted using wavelet packet transform by the dynamic round-robin acquisition module, effectively suppressing on-site electromagnetic noise and white noise interference, and significantly improving the identification of impedance fluctuation characteristics. The pseudo-corrosion fingerprint filtering module introduces a structural fingerprint similarity comparison mechanism for adjacent branches, fundamentally distinguishing between real physical corrosion and topology design errors, avoiding misjudgments caused by drawing mismatch. The incremental topology reconstruction module integrates multiple batches of data and uses resistance-constrained subgraph isomorphism verification to ensure that topology changes caused by corrosion are accurately captured, resulting in high consistency between the reconstruction results and the actual grounding grid state. The measurement range is significantly improved. The observability optimization module, based on branch diagnostic reliability indicators, autonomously decides on measurement point expansion and channel rearrangement schemes, overcoming the blind zone limitations of traditional fixed measurement point layouts. The inversion convergence judgment module, through the synergistic weighting of residual descent factors, topology matching factors, and observability factors, achieves intelligent assessment of topology stability and adaptive control of iterative termination, significantly reducing invalid measurement rounds. Finally, a triplet corrosion report is output, providing a highly reliable basis for operation and maintenance decisions, significantly reducing manual verification costs and downtime for maintenance.

[0110] Please see Figure 2 A synchronous measurement method for corrosion detection of multi-channel grounding grids includes the following steps: S1: Dynamic round-robin acquisition: Load the channel switching strategy file, measure the port impedance in a round-robin fashion and generate a time-domain response sequence;

[0111] S2: False erosion filtering: Compare real-time data with historical topology and eliminate false erosion branches through fingerprint similarity analysis;

[0112] S3: Observability optimization: When the diagnostic index is below the threshold, automatically plan the distribution of measurement points and perform supplementary measurements;

[0113] S4: Incremental Topology Reconstruction: Reconstructs the grounding grid topology by fusing multiple batches of data, and outputs the new topology structure after isomorphic verification;

[0114] S5.: Convergence determination: Trigger the final erosion branch inversion based on the stability index and output a triplet diagnostic report.

[0115] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0116] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0117] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0118] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0119] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0120] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-channel grounding grid corrosion detection and synchronous measurement system, characterized in that, Includes the following modules executed in the following order: The dynamic rotation measurement acquisition module, based on a preset channel switching strategy, performs rotational measurements on port pairs between multiple grounding leads in the substation grounding network; it obtains port voltage signals and port injected current values ​​to construct the original port response data matrix and extracts the impedance time sequence through wavelet packet transform. The operation of the dynamic rotation measurement and acquisition module includes: preloading the channel switching strategy file to the static memory; controlling the relay to close the specified port pair according to the instruction field sequence in the preloaded channel switching strategy; within the signal stability window, injecting current value and acquiring port voltage signal, the number of sampling points being determined by the sampling frequency and window duration; performing triple signal purification on the voltage signal: phase-locked demodulation to separate the fundamental component, FIR filter to suppress high-frequency noise, and statistical outlier removal; generating matrix elements from the purified data according to the impedance conversion formula, and converting the matrix into an impedance time series through wavelet packet transform; The pseudo-corrosion fingerprint filtering module, based on the original port response data matrix and the historical design topology of the grounding grid, compares the real-time impedance value in the impedance time series with the theoretical impedance value in the historical topology, and identifies the initial abnormal branch set through covariance correlation analysis; it filters pseudo-corrosion branches through the fingerprint similarity of adjacent branches and a preset low amplitude threshold, and outputs the real suspected corrosion branch set. The observability optimization and expansion module calculates the observability index of a single branch and the global diagnostic quality index for the set of real suspected corrosion branches. When the observability index of a single branch is lower than the preset minimum observability threshold, it is recorded as a low observability abnormal branch. When the global diagnostic quality index is lower than the preset standard observability threshold, the measurement point supplementation process is triggered to perform supplementary measurements. The incremental topology reconstruction module integrates the original port response data moments and the supplementary response data, and generates an updated port impedance tensor through a self-weighted iterative Kalman filter; it then uses a sparsity-preserving optimization model to solve for the reconstructed grounding grid topology; the reconstructed grounding grid topology is compared with the historical design topology of the grounding grid using a resistance-constrained maximum subgraph isomorphism algorithm, and outputs the new grounding grid topology after comparison and verification. The inversion convergence determination module calculates the comprehensive stability index based on the weighted average of the residual descent factor, topology matching factor, and observability factor. If the comprehensive stability index exceeds the preset threshold, a second least squares inversion is performed on the pseudo-corroded branch set. Branches with impedance increments exceeding three standard deviations of the health benchmark are included in the final corrosion set, and a list of branch numbers, impedance increments, and confidence level triplets is output.

2. The multi-channel grounding grid corrosion detection and synchronous measurement system according to claim 1, characterized in that, The determination logic for pseudo-corrosion branches in the pseudo-corrosion fingerprint filtering module includes: based on the original port response data matrix. According to the branch index The impedance observations of the same branch in each measurement frame are extracted into an impedance time-series vector. And record it as the first test sample; The topology was designed based on the system's pre-stored historical grounding grid. Ideal port impedance tensor obtained through offline calculation using electromagnetic simulation And denoted as the second test sample; where V is the grounding lead port number, and E is the edge set of the grounding grid, i.e. the grounding conductor branch; the deviation matrix ΔZ is obtained by subtracting the mean of the first test sample from the second test sample; For each branch e, the deviation matrix is ​​projected onto the inverse covariance space and its magnitude is taken. If the module length Branches exceeding the percentile threshold of the chi-square distribution in the same dimension are listed as initial abnormal branches and included in the set to obtain the initial abnormal branch set. ; for the impedance timing sequence corresponding to the initial abnormal branch Where e is the unique number of the branch; Maximum overlap discrete wavelet packet transform and locality-sensitive hashing are performed to generate a fixed-length bit string. This bit string is considered a structural fingerprint identifier; for each branch... If the two branches are in the historical design topology of the grounding grid If a path shares the same node (i.e., has a common endpoint), it is considered an adjacent branch. The Hamming distance between these branches is calculated and recorded as a fingerprint similarity. The number of similar adjacent branches is then counted. ; Traverse all Adjacent branch Where r is the index of the adjacent branch; if the fingerprint similarity and Then count Add 1, The fingerprint similarity is calculated, and the number of similar adjacent branches is compared with a preset minimum neighborhood number threshold. like and Then the branch road Add pseudo-corrosion branch set Among them, the pseudo-corrosion branch set ; From the initial screening of abnormal branch sets By removing pseudo-corrosion branches, a set of real suspected corrosion branches is generated. .

3. The multi-channel grounding grid corrosion detection and synchronous measurement system according to claim 2, characterized in that, The impedance timing sequence corresponding to the initial abnormal branch Perform maximum overlap discrete wavelet packet transform and locality-sensitive hashing to generate fixed-length bit strings. Specifically, in order to extract features that take into account both the time and frequency domains, the impedance time series corresponding to the initial anomalous branch is analyzed. Where e is the unique number of the branch; perform maximum overlap discrete wavelet packet transform to decompose each impedance time series into multiple sub-bands, and after calculating the variance, kurtosis, and kurtosis of the energy statistics in each sub-band, concatenate them into a fixed-length vector. These fixed-length vectors are first normalized using Z-Score: subtract the sample mean from each dimension and divide by the sample standard deviation; then, the locality-sensitive hashing method is used to normalize the fixed-length vectors. Perform a second hash to generate a bit string of fixed length. This bit string is considered a structural fingerprint identifier and is composed of characteristics such as impedance deviation and frequency domain energy.

4. The multi-channel grounding grid corrosion detection and synchronous measurement system according to claim 3, characterized in that, The measurement point addition process in the observability optimization and expansion module includes: defining the observability of a single branch as the ratio of the minimum to the maximum singular value of the impedance sequence; the global diagnostic quality index is the mean observability of all suspected branches; determining the priority of measurement points by ranking them according to node betweenness centrality; generating a candidate node set by combining the shortest wiring path and cabinet terminal capacity constraints; constructing an integer linear programming model with the goal of minimizing the number of added measurement points, satisfying three constraints: observability threshold, total wiring length, and number of terminals; inputting the integer programming model into a preset mixed integer linear programming solver to obtain the optimal expansion or rearrangement scheme. The proposed solution includes both adding new ports and rearranging channels.

5. The multi-channel grounding grid corrosion detection and synchronous measurement system according to claim 1, characterized in that, The incremental topology reconstruction module uses a sparsity-preserving optimization model to solve for the reconstructed grounding grid topology. The specific process includes: defining the original port response data matrix. Number of lines Supplementing the response data matrix Number of lines Zero-padding alignment is performed on the timeline, and the components are stitched together into a three-dimensional tensor. ;in, For three-dimensional tensors Index on the timeline; The self-weighted iterative Kalman filter is introduced for auxiliary fusion. The logic of the self-weighted iterative Kalman filter is as follows: take the branch resistance vector. The system state is represented by T, where T is the transpose sign and N is the total number of branches. As the observation equation; where Given the port-branch correlation, q is the number of iterations in the Kalman filter algorithm; As a weight matrix The noise variance of each frame is estimated online; after fusion, the port impedance tensor is updated at the output. ; remember The equivalent admittance matrix of the historical topology is determined; a sparsity-preserving incremental optimization model is adopted, based on the formula: ;in, Given a nonlinear mapping function, β is a preset tradeoff coefficient that controls the balance between preserving the original topological sparsity and allowing local modifications; The incremental optimization model is solved using the existing iterative reweighted least squares algorithm to obtain the reconstructed admittance matrix. This allows for the direct generation of the reconstructed grounding grid topology. .

6. The multi-channel grounding grid corrosion detection and synchronous measurement system according to claim 1, characterized in that, The comprehensive stability index in the inversion convergence determination module is generated as follows: the residual decrease factor is the weighted residual change rate of the two most recent reconstruction results; the topology matching factor is the final consistency score output by the incremental reconstruction module; the observability factor is the normalized value of the global diagnostic quality index of the optimization and expansion module; the three factors are weighted and summed and compared with the stability threshold. When the comprehensive stability index is greater than or equal to the preset stability threshold, it is determined to meet the standard and the final corrosion confirmation is initiated.

7. A method for synchronous measurement of corrosion detection in multi-channel grounding grids, used to implement the synchronous measurement system for corrosion detection in multi-channel grounding grids as described in any one of claims 1-6, characterized in that, Includes the following steps: S1 Dynamic Round-Robin Acquisition: Loads the channel switching strategy file, round-robin measures port impedance and generates time-domain response sequences; S2 False Corrosion Filtering: Compares real-time data with historical topology and eliminates false corrosion branches through fingerprint similarity analysis; S3 observability optimization: When diagnostic indicators are below the threshold, automatically plan a measurement point expansion scheme and perform supplementary measurements; S4 Incremental Topology Reconstruction: Reconstructs the grounding grid topology by fusing data from multiple batches, and outputs a new topology structure after isomorphic verification; S5 Convergence Determination: Trigger the final erosion branch inversion based on the stability index and output a triplet diagnostic report.

8. The method for synchronous measurement of multi-channel grounding grid corrosion detection according to claim 6, characterized in that, The final corrosion confirmation in step S5 includes: performing secondary inversion only on branches of pseudo-corrosion sets to eliminate topology mismatch interference; determining branches with impedance increments exceeding the healthy baseline mean plus 3 times the standard deviation as true corrosion; and generating triples by branch number, impedance increment, and confidence level for priority processing by the operation and maintenance system.

Images