New energy box-type substation state abnormity diagnosis method and system

Abstract

The invention discloses a new energy box-type substation state abnormity diagnosis method and system, and relates to the technical field of power equipment monitoring. According to the method and the system, through event-triggered high sampling and cross-channel time base self-inspection, multi-source sequence synchronization is ensured, and a consistent time axis is provided for subsequent estimation, so that online analysis under the condition of no power outage is supported; introducing dual-path estimation of RL step fitting and voltage drop robust regression in a transient stage, cooperatively inhibiting a closing peak and outlier data, and obtaining a stable initial value of the contact resistance; in a steady-state stage, adopting Goertzel frequency selection to extract a fundamental wave phasor and performing ohmic purity judgment, and expanding to a resistance-inductance joint model when necessary to remove interference of parasitic inductive reactance on resistance estimation; and then mutual identification and residual constraint are carried out on the temperature and electric power coupling relationship by using a discrete heat RC model, and local write-back correction is carried out on the deviation, so that the consistent closed loop of the electric domain and the heat domain is realized.

Description

Technical Field

[0001] This invention relates to the field of power equipment monitoring technology, and more specifically, to a method and system for diagnosing abnormal conditions in new energy prefabricated substations. Background Technology

[0002] New energy prefabricated substations, especially those used in wind power and photovoltaic applications, often face special operating conditions such as frequent hoisting and short-term paralleling due to their modularity, high integration, and flexible deployment. Under these conditions, the mechanical and thermal stresses of the busbar connectors inside the substation, especially at the busbar contact points, change frequently, easily leading to implicit drift and deterioration of contact resistance. Increased contact resistance can cause localized overheating, accelerating material aging and even triggering serious accidents such as thermal breakdown and fires, severely threatening the safety and reliability of the power grid. This implicit drift in contact resistance is one of the important reasons for abnormal conditions in prefabricated substations.

[0003] Traditional methods for diagnosing busbar contact resistance primarily rely on infrared thermography inspections and contact resistance measurements during scheduled power outages. While infrared thermography can be performed online, it only detects surface temperature and struggles to accurately capture subtle shifts in internal contact resistance, especially early and minor deterioration. Furthermore, it is susceptible to environmental factors such as wind speed and sunlight, leading to the risk of false alarms or missed alarms. Scheduled power outages require interrupting power supply, are time-consuming and labor-intensive, and cannot reflect real-time dynamic changes in busbar contact resistance, failing to provide timely warnings of rapidly deteriorating faults. Moreover, traditional methods are ill-suited for the special operating conditions of frequent hoisting and short-term paralleling in new energy prefabricated substations, failing to provide continuous and reliable monitoring and diagnosis. Currently, there is a lack of a method and system capable of online, quantitative, interpretable, and traceable diagnosis of subtle shifts in busbar contact resistance without power interruption, thus hindering effective diagnosis of anomalies in new energy prefabricated substations. Especially under complex operating conditions, effectively distinguishing between true changes in contact resistance and apparent changes caused by environmental factors and load variations remains a pressing technical challenge.

[0004] To address the above problems, this invention proposes a solution. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method and system for diagnosing abnormal conditions in new energy prefabricated substations, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for diagnosing abnormal conditions in a new energy prefabricated substation includes the following steps: During the closing transient or planned load step trigger, the original sequence of phase current, connector voltage drop, patch temperature and housing acceleration is acquired, high sampling is triggered and cross-channel time alignment and cross-correlation self-check are performed to establish a unified time base and form a synchronization segment. Parallel computation is performed within the rising segment, the total equivalent impedance parameter is fitted based on the RL step model, and the contact ohmic component is estimated based on robust linear regression of voltage drop and current. Path consistency gating is performed on the two results in the denoising segment. The fundamental current and voltage drop phasors are extracted within the steady-state window after the step jump. The reference phase is unified, and the ohmic purity of the impedance is determined. When parasitic inductive reactance exists, it is extended to a joint resistance and inductance model, and the contact resistance is solved. Substitute the contact resistance into the heat dissipation RC model and introduce the ambient temperature input. Perform parameter fitting based on the temperature sequence and calculate the thermal consistency residual. If it fails, correct the contact resistance in the local neighborhood and write it back to the electrical domain. Perform online change point detection on the historical event sequence and update the drift state by combining mechanical prior; construct electric domain gating, thermal domain gating and ohmic purity gating, and output the phase and connector number when the conditions are met and perform evidence chain hashing to solidify.

[0007] In a preferred embodiment, cross-channel time alignment includes: acquiring current and voltage drop at a first sampling rate, and acquiring temperature and acceleration at a second sampling rate; estimating channel offset by searching for peak values ​​within a set hysteresis window through cross-correlation; resampling to the unified time base using linear phase interpolation; determining the start and end of the rising segment using a percentile threshold; and filling in missing packet segments with linear interpolation before participating in subsequent calculations.

[0008] In a preferred embodiment, the RL step fit solves for the inductance and resistance parameters on the rising synchronous segment using least squares; the robust linear regression performs weighted estimation of voltage drop and current and uses Huber loss to suppress outliers; the path consistency gating generates judgment conditions based on the dual-path resistance difference and slope consistency, and fixes the sampling window length and the starting point alignment rule.

[0009] In a preferred embodiment, the phasor extraction within the steady-state window uses the Goertzel algorithm to perform frequency selection calculations on the fundamental wave and the set harmonics. The unified reference phase is based on the specified phase current. The ohmic purity determination is based on the voltage drop and current phase difference threshold to generate a label. If the threshold is not met, the solution is expanded to a joint resistance and inductance model, and the phase calibration record is retained.

[0010] In a preferred embodiment, the heat dissipation RC model employs a first-order or second-order ladder network and estimates thermal parameters online using recursive least squares. The ambient temperature input is obtained through a first-order IIR filter. The thermal consistency residual is calculated by the absolute deviation metric within the window. When the residual exceeds a threshold, the contact resistance is adjusted within a set neighborhood and written back to the electrical domain, while the initial values ​​of the model are updated synchronously.

[0011] In a preferred embodiment, the online change point detection employs a Bayesian online framework, which recursively calculates the change point posterior based on piecewise Gaussian likelihood and hazard function. The mechanical prior is generated by statistical analysis of acceleration spectrum energy and number of impacts and injected with prior weights. The drift state is updated by exponential moving average and used for gating. A maximum running length upper bound and reset strategy are set, and piecewise statistics are cached for verification.

[0012] In a preferred embodiment, the electric domain gating generates a binary indication based on the deviation range between the transient fitting resistor and the steady-state phasor resistor; the thermal domain gating generates a binary indication based on the thermal consistency residual and parameter convergence; and the ohmic purity gating obtains a binary indication based on the phase difference threshold. The three are ANDed to form diagnostic conditions, and the gating details, including the original gating quantities and thresholds of each gating, are output.

[0013] In a preferred embodiment, evidence chain hash solidification includes: structurally encoding the original sequence, the unified time base, fitting parameters and gating indications; calculating fingerprints using SHA-256 and attaching random salts; writing the fingerprints along with the event time and device identifier into an append-only log; the log sequence number and the previous fingerprint forming a chain index; and saving a verification snapshot and generating a batch signature.

[0014] In a preferred embodiment, the output of phase and connector number is accomplished through a topology mapping table, which is generated by a primary wiring and phase configuration and includes connector code, phase identifier and measurement point association relationship; the output simultaneously records the trigger type and window number and associates it with the evidence chain fingerprint for consistent reference, and synchronously generates a number list and writes it into the mapping version number.

[0015] A new energy prefabricated substation status anomaly diagnosis system includes: a data acquisition module, configured to acquire current, voltage drop, temperature and acceleration sequences at high sampling rate when the circuit is closed or a planned load step is triggered, and to perform cross-channel time alignment and cross-correlation self-check; The electrical transient and steady-state analysis module is configured to perform RL step fitting and linear regression in parallel during the rising phase to estimate the contact ohmic component, and extract phasors in the steady-state window to determine the ohmic purity and expand to solve the joint resistance and inductance model when necessary. The thermal domain verification module is configured to substitute the contact resistance into the heat dissipation RC model for parameter fitting and residual calculation, and correct the contact resistance in the local neighborhood and write it back when it fails. The variable point and gating output module is configured to perform online variable point detection on historical events and update the drift state in combination with mechanical priors, construct electric domain gating, thermal domain gating and ohmic purity gating and output the phase and connector number; The evidence chain solidification module is configured to perform hash calculations on key data and parameters and write them to an increment-only log. The technical effects and advantages of the present invention regarding the diagnosis method and system for abnormal conditions in new energy prefabricated substations are as follows: This invention ensures multi-source sequence synchronization through event-triggered high sampling and cross-channel time-based self-checking, providing a consistent time axis for subsequent estimation and supporting online analysis under uninterrupted power supply conditions. In the transient phase, a dual-path estimation method combining RL step fitting and voltage drop robust regression is introduced to collaboratively suppress closing spikes and outlier data, obtaining a stable initial value for contact resistance. In the steady-state phase, Goertzel frequency selection is used to extract the fundamental phasor and perform ohmic purity determination, extending to a joint resistance-inductance model when necessary to eliminate the interference of parasitic inductance on resistance estimation. Subsequently, an off-grid RC model is used to mutually verify and constrain the coupling relationship between temperature and electrical work, and local write-back correction is performed to correct deviations, achieving a consistent closed loop between the electrical and thermal domains. In summary, the scheme achieves online, quantitative, and distinguishable modeling of contact resistance, reducing misjudgments caused by environmental disturbances and harmonic noise, and adapting to complex operating conditions such as frequent paralleling of new energy transformer substations. Attached Figure Description

[0016] Figure 1 This is a flowchart of a method for diagnosing abnormal conditions in a new energy prefabricated substation according to the present invention. Figure 2 This is a structural diagram of a new energy prefabricated substation status anomaly diagnosis system according to the present invention. Detailed Implementation

[0017] 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.

[0018] Example 1, referring to Figure 1 The present invention provides a method for diagnosing abnormal conditions in a new energy prefabricated substation, comprising the following steps: S101: Event Triggering and Time-Based Self-Check; This step is the starting point of the entire diagnostic process, designed to ensure the time synchronization and accuracy of subsequent data analysis. When the system detects a closing transient, such as a substation reconnecting to the grid or reconnecting after disconnection, or a planned load step, such as the commissioning or disconnection of high-power equipment, it immediately triggers a high-sampling mode. In high-sampling mode, the current sensor i_phi(t) and the voltage taps v_drop,phi(t) on both sides of the bus will continuously sample at a frequency of 20kHz for 50ms to capture the details of the transient process; then the sampling frequency drops to 2kHz and continues for 5s to cover the stable process after the transient. At the same time, the contact patch temperature sensor samples at 10Hz and the housing acceleration sensor samples at 1kHz for subsequent thermal cross-validation and mechanical causality analysis.

[0019] To ensure time synchronization of data from different channels, this step involves checking the current. With pressure drop Perform cross-correlation calculations: By finding the time delay corresponding to the peak of the cross-correlation function. This is used to evaluate the time deviation between channels. The criterion is set as follows: That is, the absolute value of the inter-channel delay does not exceed 0.5 milliseconds; at the same time, inter-phase jitter This means that the time delay consistency between different phases is good. If any criterion is not met, it is considered that there is a time synchronization problem in this data acquisition, the current identification is terminated, and the process waits for the next excitation event. This step effectively eliminates spurious parameters caused by sensor trigger misalignment or sampling asynchrony, providing a shared and accurate time reference for subsequent electrothermal mutual verification. In a preferred embodiment, to avoid the saturation effect of turn-on spikes on the voltage tap channel, the range of the voltage tap channel can be increased by 20%. In addition, an 8th-order linear phase FIR filter can be used at the front end for DC removal and trend processing to improve signal quality.

[0020] S102: Electrical transient path A, RL step fitting locks the total equivalent resistance component; after passing the time-based self-test in S101, this step uses the electrical response of the closing transient to estimate the total equivalent resistance of the busbar. During the current rise in the first 20ms after closing, the busbar connector can be equivalent to an RL series circuit. The current response model is as follows: ,in, It is steady-state current. It is a time constant. , It is the equivalent inductance. This is the total equivalent resistance. This step uses a nonlinear least squares method, such as the Gauss-Newton algorithm, to fit the data. and To improve the robustness of the fit, The initial value can be taken as the average value of the last 3ms of the current rise segment. The initial value can be set to 63% of the current. The corresponding time constant. Obtain the fitting parameters. and Then, calculate the total equivalent resistance. .in, This refers to the average voltage drop measured on the connector channel after the transient response during closing reaches a steady state. This is not the nominal busbar voltage, but specifically refers to the steady-state voltage drop across the busbar connector circuit. Furthermore, this is calculated by subtracting the known baseline resistance of the cables and the busbar itself. The baseline resistance is derived from factory calibration or measurements taken during the initial commissioning phase, providing an estimate of the contact resistance. To suppress the impact of closing spikes on the fitting, the residuals are processed using the Huber weighting function. If the fitted time constant... If the time is less than 0.2ms, it indicates that the transient process is too fast, which may indicate model mismatch or data quality issues. In this case, no output is output. Instead, it proceeds to S104 for mold expansion determination.

[0021] S103: Electrical transient path B, voltage drop regression directly estimates the contact ohmic component; this step provides another independent electrical estimation path for contact resistance, forming a dual-path verification with S102. During the current rise phase, to avoid the influence of contact glitches and high-frequency noise, the first N0 sampling points are removed; for example, N0 can be set to several sampling periods depending on the actual situation. Then, a linear relationship model between voltage drop and current is established: ,in It is contact resistance. It could be noise or model error. This step uses the RANSAC (Random Sampling Consistent Linear Regression) algorithm to estimate... The RANSAC algorithm can effectively handle outliers in the data, improving the robustness of regression. The inlier threshold can be set to 3 times the median absolute deviation, i.e., 3 × MAD.

[0022] In obtaining and Then, execute path consistency gating: .in, This is a preset tolerance value, such as 2–5 microohms, whose specific value can be adjusted based on the system noise level and the calibration results of the current transformer CT linear region. If gating fails, i.e., the estimated values ​​of the two independent electrical paths differ too much, the diagnostic result is considered unreliable, and the process returns to S101 to re-acquire data, or a tap self-test is performed to troubleshoot sensor faults. This step applies the same parameter through two independent electrical constraint paths. Intersection locking significantly improves the confidence level of contact resistance estimation.

[0023] S104: Steady-state Ohmic invariant and parasitic term separation; this step aims to distinguish the true ohmic resistance component from parasitic inductive reactance or impedance caused by harmonics, eddy current effects, etc. Within the steady-state segment after the step transition, the fundamental current phasor is extracted using the Goertzel frequency selection algorithm. Phasor of fundamental voltage drop The window length N of the Goertzel frequency selection algorithm is taken as the sampling frequency. With fundamental frequency Integer multiples of the ratio, i.e. To effectively suppress spectral leakage and improve frequency selection accuracy, the estimation method for the fundamental phase angle φ is as follows: the phase information of the fundamental current and voltage drop is directly obtained through the Goertzel algorithm, with the current phasor as the reference phase. Then, a parameter characterizing the impedance amplitude is constructed. In order to evaluate The stability is calculated within three adjacent steady-state windows. coefficient of variation At the same time, calculation Cross-correlation coefficient with power factor angle cosφ .

[0024] when ≤5% and When it is approximately 0, it indicates It is very stable in steady state and independent of the power factor, at which point the parasitic reactance can be considered negligible. This is the actual contact resistance R_c. Otherwise, if >5% or If the value is not close to 0, it indicates the presence of significant parasitic impedance or harmonic effects, requiring model expansion. In this case, the model is expanded to... The contact resistance was simultaneously estimated using the least squares method. and parasitic inductance In this process, it is necessary to... Apply boundary constraints, i.e., | |It must be below the upper limit of the structure caused by vortex effects estimated through geometry, to ensure The physical rationale is established. This step, through the Ohm invariant and frequency-selective phasor method, effectively eliminates non-Ohm components such as harmonics and eddy current effects, ensuring... The physical semantic purity, together with S102 and S103, forms a consistency verification within the electric domain.

[0025] S105: Thermal domain verification, using power consumption and heat generation to drive reverse calibration of the thermal RC model; this step introduces thermal domain information and further solidifies the model through the consistency of the electro-thermal cross-domain equations. The estimated value is used to identify abnormal heat dissipation paths. The contact resistance determined in steps S102–S104 is then used to... Substituting these known quantities into the thermal RC model: .in, It is heat capacity. It's thermal resistance. It is the contact temperature. It is the ambient temperature. Discretize the differential equation as follows: .

[0026] Fitting thermal parameters using the Levenberg–Marquardt algorithm and and impose reasonable boundary constraints, such as , These boundary constraints are based on empirical values ​​of the thermal properties of typical busbar joint materials and structures. After fitting, the thermal consistency residuals are calculated. That is, the measured temperature and the temperature based on The maximum deviation between the predicted temperatures. If Then it is considered that the electric domain estimation is The response is consistent with the thermal domain response, thus passing verification. Otherwise, if the residual is too large, it indicates... The estimated value may be biased or the heat dissipation path may be abnormal. In this case, only... within a small neighborhood Perform a one-dimensional search to minimize and the revised The data is written back to the electrical domain for subsequent analysis. This step, through an electro-thermal coupling model, achieves cross-domain mutual verification and calibration, effectively improving... The estimation is physically reasonable and accurate, and it can identify potential heat dissipation anomalies.

[0027] Regarding the influence of changes in duct or convection coefficients with station operation status, this invention partially considers environmental changes by measuring ambient temperature in real time and incorporating it into the thermal model. Furthermore, thermal resistance... The fitting process adaptively reflects the current heat dissipation conditions. In practical applications, the convective heat transfer coefficient typically ranges from 5 to 25 W / (m²·K), and its variation may lead to… Variation ±1-3℃. This invention allows for certain environmental disturbances by setting a residual threshold of 2℃, while triggering fine-tuning of Rc when the residual exceeds the limit, thereby enhancing the robustness of the model.

[0028] S106: Online Change Point Detection with Mechanical Causal Prior Injection; This step treats mechanical disturbances as prior information and directly integrates them into the contact resistance drift detection mechanism, forming a causal closed loop from event to parameter change. The contact resistance corrected in S105 within the historical event sequence is then analyzed. Perform Bayesian online change point detection. The observation model uses a Gaussian distribution and is equipped with a normal-inverse gamma conjugate prior to derive intra-segment statistics. The key lies in the setting of the danger function H(r): Where H0 is the baseline hazard rate. When in the interval ( [Memory exists in |a(t)|> During a lifting or impact event, a(t) is the shell acceleration collected in S101. The preset impact threshold is used, where γ=3, indicating that mechanical disturbance significantly increases the probability of abrupt changes in contact resistance; otherwise, γ=1. In this way, mechanical disturbance is directly injected as a strong prior into the change point detection model. This step outputs the posterior of the abrupt change. , indicating in The probability of a change point occurring at any given time, and the posterior mean of the drift rate. .

[0029] about The calibration process is based on the analysis of experimental data from actual lifting conditions. Specifically, under typical lifting scenarios, a high-precision triaxial accelerometer (e.g., ±10g range, 0-1kHz bandwidth) is installed near the busbar joint to collect acceleration data from numerous impact events. These data are then statistically analyzed, and the 95th percentile value is used as the baseline. The threshold. For example, in a typical transformer substation hoisting test, the 95th percentile impact acceleration is 5g, then... Set to 5g. This calibration process ensures... It can effectively identify mechanical impacts related to hoisting while avoiding accidental triggering.

[0030] S107: Counterfactual verification to rule out non-contact single-cause explanations; this step aims to enhance the defensibility of the diagnostic conclusions by ruling out non-contact factors, such as load changes, environmental duct changes, and the possibility of these factors explaining the observed phenomena alone. First, a system baseline is selected. This could be the contact resistance value measured during the initial commissioning phase, or the average value over a recent period of stable operation. Then, while maintaining the current at the current moment... Ambient temperature And the thermal resistance obtained by fitting S105 and heat capacity Under the condition that remains unchanged, call the thermal model to generate a counterfactual temperature rise trajectory. This counterfactual trajectory simulates how the contact temperature should change under current load and environmental conditions if the contact resistance remains at the baseline level.

[0031] Next, compare the counterfactual temperature rise trajectory. The actual measured temperature trajectory .like If the maximum deviation between the two exceeds 2 degrees Celsius, and the electrical domain gating S103 and S104 have passed, then the possibility that the observed temperature rise can be explained solely by load or duct changes can be ruled out. This reinforces the causal conclusion that contact resistance degradation is the main cause of abnormal temperature rise, avoiding misjudging temperature changes caused by other factors as contact resistance problems, and thus more accurately diagnosing abnormal conditions in new energy prefabricated substations.

[0032] S108, Constructing Triple Gating: This step is the final output of the diagnostic results. Multiple gating mechanisms ensure the reliability of the diagnostic conclusions and provide quantitative health assessments and remaining life expectancy predictions. Three gating variables are constructed: 1. Electrical domain consistency gating Ce: Ce measures the consistency between the two electrical path estimation results, S102 and S103. The closer the value is to 1, the better the consistency.

[0033] 2. Thermal domain uniformity gating Ch: Ch measures the consistency between the electric domain estimate of Rc and the thermal domain response; the closer the value is to 1, the better the consistency.

[0034] 3. Ohmic purity gated Cκ: Cκ measures the purity of the ohmic component under steady state; the closer the value is to 1, the smaller the influence of parasitic immunity.

[0035] When Ce ≥ 0.7, Ch ≥ 0.7, and Cκ ≥ 0.7, the diagnostic result is considered to have sufficient confidence. In this case, based on the Rc estimate from the most recent N events (e.g., N = 5-10), the Theil-Sen estimator, a linear regression method insensitive to outliers, is used to estimate the drift rate of the contact resistance. Then, the critical value is calculated. Estimated arrival time ,in This is the current Rc value. It is a minimum drift rate, for example This is used to avoid the denominator being zero when the drift rate is zero. It can be set according to the characteristics of the busbar material and the fastening specifications, and can be adjusted in zones according to the ambient temperature.

[0036] At the same time, calculate health status .in, It is the post-mutational a posteriori. The sigmoid function is used to map mutation probabilities to a health score deduction. This health score comprehensively considers the confidence level of the diagnostic result, physical consistency, and mutation risk. Finally, the system output includes phase identification, connector number, evidence chain hash, and... The system provides handling recommendations to guide the handling of abnormal conditions in new energy prefabricated substations. If any gate fails, only the monitoring event is recorded, and no work order is generated, thus avoiding invalid maintenance caused by false positives in a single domain.

[0037] S109, hash solidification; To ensure the verifiability, auditability, and traceability of the diagnostic process, this step hashes and solidifies the key data and parameters involved in S101–S108. Specifically, this includes: the original sensor data used in this diagnostic process, such as current, voltage, temperature, and acceleration; the window index for data acquisition; the residuals of each fitting process; the triple-gated quantities Ce / Ch / Cκ; and the final estimated contact resistance. drift rate Postmutation test And the diagnostic conclusions. This information is organized into a chain of evidence from "data fragments to model version to parameters to conclusions" and written to an increment-only log.

[0038] Regarding the hash algorithm, this invention employs the SHA-256 algorithm to perform hash calculations on data fragments, ensuring data integrity and tamper-proofing. The timestamp source uses high-precision network time protocols PTP or NTP for synchronization, guaranteeing the accuracy and consistency of the timestamps. The storage medium can be a WORM (WriteOnceReadMany) storage device or a distributed ledger based on blockchain technology to further enhance the immutability and traceability of the data. This mechanism ensures that every diagnosis is computably reproducible, providing a complete historical record for subsequent dispute review, model iteration, and fault analysis.

[0039] S110, Baseline Recalibration and Adaptive Backfilling; This step aims to enable the diagnostic model to adaptively evolve in sync with actual field conditions, suppressing systematic drift caused by sensor aging, environmental changes, etc. If thermal consistency residuals appear after K consecutive events, for example, K=3 times... And Ohm's purity coefficient of variation This indicates that there may be a systematic change in the resistance of the cable or busbar itself, causing a baseline change. No longer accurate. At this point, the system triggers a backtracking correction of the "cable + busbar" baseline.

[0040] The calculation method for backtracking correction is as follows: This involves subtracting the currently estimated contact resistance from the total resistance to deduce the baseline resistance between the cable and the busbar. Then, the baseline is updated. Where β is an update weight, for example, β=0.2. This is the median value from the most recent back-tracking correction values ​​to improve robustness. After updating the baseline, the system forces a return to S101 to re-acquire data and re-verify. and Does the baseline decrease? If it does, the baseline correction is effective; otherwise, further investigation of other systemic issues may be necessary. This step ensures that the diagnostic model can operate stably and accurately over the long term, avoiding systemic misjudgments caused by outdated baselines, thereby improving the accuracy of diagnosing anomalies in new energy prefabricated substations.

[0041] The entity executing the above method can be a diagnostic system deployed on the edge industrial control computer of the prefabricated substation, or an online computing platform linked with the station control backend, but is not limited to these.

[0042] Reference Figure 2 The present invention provides a new energy prefabricated substation status anomaly diagnosis system, comprising: The data acquisition module is used to trigger high sampling and perform cross-channel time consistency verification when a closing transient or planned load step occurs, and to collect the current. With pressure drop Raw sensor data, etc. This module corresponds to S101 in the method.

[0043] The electrical transient analysis module is used to fit the current using an equivalent RL model during the initial 20ms rise after closing, locking the total equivalent resistance component. After removing the first N0 sampling points in the rising segment, RANSAC linear regression was used to directly estimate the contact ohmic component. It also performs path consistency gating. This module corresponds to methods S102 and S103.

[0044] The steady-state analysis module is used to obtain the fundamental phasor using Goertzel frequency selection within the steady-state segment after a step jump, and to construct... The coefficient of variation was calculated in three adjacent windows. ,according to Determine whether parasitic resistance is negligible, and then perform model expansion and parameter estimation. This module corresponds to S104 in the method.

[0045] The thermal domain verification module is used to verify the results determined by the electrical transient analysis module and the steady-state analysis module. As known quantities, they are substituted into the thermal RC model, and the thermal parameters are fitted using the Levenberg-Marquardt method to calculate the thermal consistency residuals. and according to Perform the correction of Rc. This module corresponds to S105 in the method.

[0046] The mechanical causal analysis module is used for analyzing event sequences. Perform Bayesian online change point detection, inject mechanical causal prior, and output mutation posterior. Compared with the posterior mean of the drift rate This module corresponds to method S106.

[0047] The counterfact verification module is used to select the system baseline. While maintaining the current , , , Under unchanged conditions, the thermal model is invoked to generate a counterfactual temperature rise trajectory. and according to and The difference excludes non-contact single-cause explanations. This module corresponds to S107 in the method.

[0048] The quantization output module is used to construct triple-gated variables Ce, Ch, and Cκ. When the triple-gated variables meet certain conditions, the drift rate is estimated based on the most recent N events. And calculate the critical value. Estimated arrival time And health level H, and output includes phase, connector number, evidence chain hash and The proposed handling method is S108 in this module.

[0049] The evidence chain solidification module is used to hash and solidify key data and parameters during the diagnostic process, forming an evidence chain, and writing it to an increment-only log. This module corresponds to S109 in the method.

[0050] The baseline recalibration module is used to address situations where K consecutive events occur. If the temperature is >2℃ and ρ_κ>0.05, then the baseline of the "cable + busbar" is backtracked and corrected, and the baseline is updated. This module corresponds to S110 in the method.

[0051] This invention achieves online, interpretable, and traceable diagnosis of implicit drift in busbar contact resistance through an indivisible process: from time consistency to electrical transient dual-path parameter locking to steady-state invariant equations to thermal domain mutual verification to causal change points to counterfactual verification to triple gating to evidence chains to baseline recalibration. This effectively diagnoses anomalies in new energy prefabricated substations, accurately identifying early hotspots and contact degradation trends under unconventional operating conditions such as frequent hoisting and short-term shunt operations, quantifying remaining risk windows, and outputting actionable maintenance priorities. The above steps are interdependent and logically progressive; if any step fails, subsequent conclusions cannot be generated, thus ensuring the overall solution is coherent and indivisible.

[0052] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A new energy box-type substation state abnormality diagnosis method, characterized by, Comprising the following steps: At the moment of closing transient or planned load step triggering, obtain the original sequence of phase current, joint voltage drop, patch temperature and shell acceleration, trigger high sampling and perform cross-channel time alignment and cross-correlation self-check, establish a unified time base and form synchronized segments; In the rising segment, calculate in parallel, based on the RL step model to fit the total equivalent impedance parameters, and based on the voltage drop and current, estimate the contact ohmic component by robust linear regression, and perform path consistency gating on the results of the denoising section; In the steady-state window after the step, extract the fundamental current and voltage drop phasor, unify the reference phase, and determine the ohmic purity of the impedance; when there is a parasitic inductance, expand it to a resistance-inductance combined model and solve the contact resistance; Substitute the contact resistance into the discrete thermal RC model and introduce the environmental temperature input, perform parameter fitting according to the temperature sequence and calculate the thermal consistency residual, if not passed, modify the contact resistance in the local neighborhood and rewrite it to the electrical domain; Perform online change point detection on the historical event sequence and update the drift state combined with mechanical prior; construct electrical domain gating, thermal domain gating and ohmic purity gating, output the phase and joint number when the conditions are met and perform evidence chain hash solidification.

2. The new energy box-type substation state abnormal diagnosis method according to claim 1, characterized in that: Cross-channel time alignment includes: collecting current and voltage drop at a first sampling rate, collecting temperature and acceleration at a second sampling rate; estimate channel offset by searching for peak value in a set lag window through cross-correlation, resample to a unified time base using linear phase interpolation, and determine the start and end of the rising segment with a percentile threshold, and after linear interpolation to fill in the missing packet segment, participate in subsequent calculation.

3. The method according to claim 2, characterized in that: RL step fitting solves inductance and resistance parameters on the rising segment synchronized segment by least squares; robust linear regression estimates voltage drop and current with weights and suppresses outliers using Huber loss; Path consistency gating generates a decision condition based on the consistency of the double-path resistance difference and the slope, and fixes the sampling window length and the start point alignment rule.

4. The method according to claim 1, characterized in that ； The phasor extraction in the steady-state window uses the Goertzel algorithm to calculate the fundamental and set harmonics, and the reference phase is unified based on the specified phase current; the ohmic purity determination generates a label based on the voltage drop and current phase difference threshold, and when not met, expands to a resistance-inductance combined model to solve, and preserves the phase calibration record.

5. The method according to claim 3, characterized in that: The discrete thermal RC model uses a first-order or second-order ladder network and estimates thermal parameters online using recursive least squares; the environmental temperature input is filtered by a first-order IIR filter; The thermal consistency residual is calculated by the absolute deviation in the window, and when the threshold is exceeded, adjust the contact resistance in the set neighborhood and write it back to the electrical domain, and update the model initial value simultaneously.

6. The method according to claim 1, characterized in that: Online change point detection uses a Bayesian online framework, calculates the change point posterior based on piecewise Gaussian likelihood and hazard function, and generates a mechanical prior from acceleration spectrum energy and impact number statistics and injects a prior weight; the drift state is updated by exponential moving average and used for gating, and set the maximum running length upper bound and reset strategy and cache the segmented statistics for review.

7. The method according to claim 5, characterized in that: The electrical domain gate generates a binary indicator based on the deviation interval between the transient fitting resistance and the steady-state phasor resistance, the thermal domain gate generates a binary indicator based on the thermal consistency residual and the parameter convergence degree, and the ohmic purity gate generates a binary indicator based on the phase difference threshold value, and the three are combined by AND operation to form a diagnostic condition, and the gate details are output, including the original amount of each gate and the threshold value.

8. The method according to claim 1, characterized in that: The evidence chain hash solidification includes: structurally encoding the original sequence, the unified time base, the fitting parameters and the gate indicator, calculating the fingerprint using SHA-256 and appending a random salt, writing the event time and device identification together in the append-only log, and the log sequence number and the previous fingerprint constitute a chain index, and save the check snapshot and generate the batch signature.

9. The method according to claim 7, characterized in that: The phase type and joint number output are completed by a topology mapping table, which is generated by the primary wiring and phase type configuration and includes joint coding, phase type identification and measurement point association; The output also records the trigger type and window number and associates the evidence chain fingerprint for consistent reference, and synchronously generates the numbered list and writes the mapping version number.

10. A new energy box-type substation state anomaly diagnosis system for implementing the method of any one of claims 1-9, characterized in that, Comprise: A data acquisition module configured to acquire current, voltage drop, temperature and acceleration sequences at high sampling when closing or planned load step triggers, and perform cross-channel time alignment and cross-correlation self-checking; An electrical transient and steady-state analysis module configured to perform RL step fitting and linear regression in parallel in the rising segment to estimate the contact ohmic component, and extract phasors in the steady-state window to determine ohmic purity and extend to a resistance-inductance joint model solution when needed; A thermal domain mutual verification module configured to substitute the contact resistance into the discrete thermal RC model for parameter fitting and residual calculation, and correct the contact resistance in the local neighborhood when it fails, and rewrite it; A variable point and gate output module configured to perform online variable point detection on historical events and update the drift state combined with mechanical priori, construct electrical domain gate, thermal domain gate and ohmic purity gate, and output phase type and joint number; An evidence chain solidification module configured to hash calculate key data and parameters and write into the only-increase log.

Images