Outlet pressure plate monitoring system and method based on transient identification and adaptive compensation

Abstract

This invention discloses an outlet pressure plate monitoring system and method based on transient identification and adaptive compensation. The system includes: a multi-source high-frequency data acquisition module, a microprocessor module, a diagnostic output module, and an abnormal trend prediction module. The monitoring method includes the following steps: S1: Acquire the ground potential signals on both sides of the outlet pressure plate and the global reference potential signal of the DC bus, and capture the high-frequency transient voltage sequence; S2: Extract time-frequency domain features from the high-frequency transient voltage sequence, calculate the transient entropy feature value and the information entropy of the sample sequence; S3: Calculate the ground bias coefficient of the DC bus and construct a proportional discrimination model; S4: Perform legality verification through the causal chain of the operation sequence, identify and intercept unexpected state changes, and predict and analyze the degradation trend of the pressure plate. This system and method achieve accurate identification and in-depth performance diagnosis of the pressure plate state through edge-side high-frequency waveform feature calculation, global reference frame offset compensation, and a logic interlocking mechanism based on the causal chain.

Description

Technical Field

[0001] This invention relates to the field of power system relay protection and secondary circuit monitoring technology, and in particular to an outlet pressure plate monitoring system and method based on transient identification and adaptive compensation. Background Technology

[0002] As the first line of defense in a power system, relay protection is a crucial guarantee for the safe operation of the power system. The correctness of the relay protection output switch status directly affects the safe and stable operation of the power grid. However, in actual operation, hidden defects such as false switching of switches due to abnormal aging of the switch structure, and incorrect or missed switching of switches due to improper operation and maintenance, often lead to protection failure or maloperation accidents, posing a serious threat to the safe operation of equipment and the power grid.

[0003] Currently, existing monitoring technologies mainly involve adding auxiliary nodes at the pressure plate location or using image recognition technology to monitor the physical state of the pressure plate, or adopting a non-contact current detection mode based on the principle of DC electric field induction.

[0004] However, existing technologies still exhibit significant limitations in complex operating environments. On the one hand, traditional monitoring methods primarily focus on physical location identification, making it difficult to effectively identify hidden faults such as "poor contact" caused by contact oxidation or mechanical wear, and failing to ensure reliable circuit continuity from an electrical performance perspective. On the other hand, substation DC systems are prone to ground potential drift due to insulation degradation, leading to false alarms or missed monitoring by monitoring devices based on fixed voltage thresholds. Furthermore, existing solutions often lack causal logic verification with station-end operation commands, making it difficult to distinguish unexpected displacements caused by accidental human contact or mechanical vibration. In addition, some monitoring systems have complex architectures and high investment costs, and still face significant technical bottlenecks in remote inspection and refined management. Therefore, developing an intelligent pressure plate monitoring device with adaptive compensation capabilities and the ability to deeply identify contact performance is of significant practical importance. Summary of the Invention

[0005] To overcome the above problems, the purpose of this invention is to provide an outlet pressure plate monitoring system and method based on transient identification and adaptive compensation. This monitoring system and method achieve accurate identification of the pressure plate status and in-depth performance diagnosis through edge-side high-frequency waveform feature calculation, global reference frame offset compensation, and a logic interlocking mechanism based on causal chains.

[0006] The technical solution adopted in this invention is:

[0007] An outlet pressure plate monitoring system based on transient identification and adaptive compensation includes:

[0008] The multi-source high-frequency data acquisition module includes four sampling channels. The sampling channels are electrically connected to the output pressure plate circuit in the relay protection cabinet. The first and second sampling channels are respectively connected to the upper and lower ground potential monitoring points of the corresponding output pressure plates to obtain local voltage characteristics. The third and fourth sampling channels are synchronously introduced into the positive and negative ground voltage sampling lines of the DC bus system to obtain the global reference potential characteristics reflecting the background environment of the DC system in real time.

[0009] The microprocessor module receives the raw waveform data stream sent by the multi-source high-frequency data acquisition module and is configured to execute the core solution operator, which includes a normalized adaptive offset compensation operator and a time series transient entropy identification operator.

[0010] The diagnostic output module has an embedded operation sequence causal chain matching engine. It monitors operation command messages in the substation area network in real time through the data bus and establishes a time correlation matrix between physical displacement events and management logic sequences. Based on the probability of the overlap between the mutation time of the proportional factor K and the instruction authorization time window, combined with the transient entropy feature value, it outputs a comprehensive diagnostic message including the engagement / disengagement status, contact health, and operation compliance.

[0011] The abnormal trend prediction module integrates a multi-dimensional time-series feature vector containing historical engagement and disengagement times, transient entropy peak sequence, static operating voltage drop drift, and environmental temperature and humidity correction coefficients. It uses a long short-term memory neural network model to capture the nonlinear evolution law of pressure plate performance from microscopic loss to macroscopic impedance change, and outputs the estimated failure time of the pressure plate contacts and the failure probability curve within a specific future period.

[0012] As a further description of the present invention, the calculation steps of the normalized adaptive offset compensation operator of the microprocessor module are as follows:

[0013] The arithmetic mean of the voltages to ground of the positive and negative poles of the DC bus is calculated in real time, and the ground bias coefficient of the current DC system is extracted.

[0014] A proportional discrimination model is constructed with the real-time full voltage span of the DC bus as the denominator and the absolute value of the instantaneous potential difference between the two ends of the pressure plate as the numerator.

[0015] The electrical state of the pressure plate is mapped to a normalized characteristic space that shifts synchronously with the bus potential in real time. The calculated scaling factor K automatically adapts to DC systems of different voltage levels and eliminates ground potential drift interference. The scaling factor formula is:

[0016] ;

[0017] in, and These are the voltages to ground at the top and bottom of the pressure plate, respectively. and These are the positive and negative voltages of the DC bus to ground, respectively.

[0018] The calculation steps for the time series transient entropy identification operator are as follows:

[0019] Maintain a high-speed data loop buffer with a sampling frequency of not less than 10kHz, and automatically capture a transient voltage sequence X of a preset duration when the voltage switch of the pressure plate is detected.

[0020] The information entropy algorithm is used to quantify the dispersion and complexity of the sequence in the time domain, and to calculate the transient entropy characteristic value reflecting the micro-contact quality of the contacts. This value is then used to identify the hidden contact performance degradation masked by steady-state voltage drop. The calculation formula is:

[0021] ;

[0022] Where X is the extracted transient voltage sequence sample; For the first The probability of a sampling point occurring within a given similarity tolerance range.

[0023] As a further description of the present invention, the transient identification operator has a fault characterization function based on spectral fingerprinting, which identifies and distinguishes the following types of virtual connections by performing time-frequency domain decomposition on the captured high-frequency sequence:

[0024] When low-frequency forced vibration characteristics of 10Hz-500Hz are detected and the transient entropy exceeds the first threshold, it is determined to be a mechanically loose type of loose connection.

[0025] When a broadband random pulse with asymmetric characteristics is detected and the transient entropy exceeds the second threshold, it is determined to be a metal oxide type virtual connection.

[0026] When high-order harmonic energy clusters above 5kHz are identified and accompanied by microscopic discharge characteristics, it is determined to be a fouling-ablation type of loose connection.

[0027] As a further description of the present invention, when the diagnostic output module performs causal chain matching, if it detects that the pressure plate proportional factor K has reversed its state but no legal instruction is found to support it within the corresponding operation instruction authorization time window, it determines that the displacement is an unexpected illegal change, accidental contact, or serious electrical insulation breakdown, and immediately raises the alarm level to the highest level.

[0028] As a further description of the present invention, the diagnostic output module also integrates a comprehensive diagnostic operator. This comprehensive diagnostic operator is based on a multi-source evidence fusion mechanism and supports the input of mechanical contact pressure vectors provided by pressure sensors and local temperature rise vectors provided by infrared sensors. The multi-source evidence fusion mechanism adopts the Durmst-Schaffer (DS) evidence theory.

[0029] .

[0030] in, The confidence level of the fused diagnosis; This represents the degree of conflict between evidence from different sensors.

[0031] By introducing the Durmst-Schaffer (DS) evidence theory, the perceived data from the three dimensions of electrical, mechanical, and thermal environment are weighted, fused, and arbitrated. The complementarity of multi-source data is used to eliminate logical false alarms caused by electromagnetic induction in the secondary circuit.

[0032] As a further description of the present invention, the multi-dimensional time-series feature vector of the abnormal trend prediction module for:

[0033] ;

[0034] Where C represents the cumulative number of throws and withdrawals. This is a sequence of transient entropy peaks. The static pressure drop drift is represented by T and RH, which are correction factors for ambient temperature and humidity.

[0035] The monitoring method for the outlet pressure plate based on transient identification and adaptive compensation includes the following steps:

[0036] S1: The data acquisition module acquires the ground potential signals on both sides of the outlet pressure plate and the global reference potential signal of the DC bus, and captures the high-frequency transient voltage sequence at the moment of pressure plate potential switching.

[0037] S2: Time-frequency domain feature extraction is performed on the high-frequency transient voltage sequence. The transient entropy feature value reflecting the micro-contact quality of the contact is calculated using the information entropy algorithm, and the probability distribution function is used. Calculate the information entropy of the sample sequence Abnormal oscillations during the pressure plate closure process can be identified by using feature vectors.

[0038] S3: Real-time calculation of the DC bus ground offset coefficient in a DC system A proportional discrimination model based on global reference potential is constructed, and the logical engagement / disengagement state of the outlet pressure plate is adaptively identified using a proportional factor. Multidimensional sensing parameters are introduced for weighted fusion arbitration.

[0039] S4: The physical displacement of the outlet pressure plate is verified by operating the causal chain of the sequence, and unexpected state changes are identified and intercepted. Historical operating data is used to predict and analyze the deterioration trend of the pressure plate.

[0040] As a further description of the present invention, the DC bus to ground bias coefficient in S3 The calculation formula is:

[0041] .

[0042] in, and These are the positive and negative voltages of the DC bus to ground, respectively.

[0043] A substation secondary circuit intelligent operation and maintenance terminal is disclosed. The terminal performs complex entropy calculation and causal chain matching locally, and sends the simplified results after logical filtering and deep diagnosis to the digital management platform through a power-specific communication protocol, thereby effectively reducing the load on the station's communication bandwidth.

[0044] As a further description of the present invention, the terminal has a standardized plug-and-play sensor interface, which can flexibly connect auxiliary sensing parameters according to the needs of the field environment, and uses a built-in weighted discrimination model to solve the problem of difficulty in identifying the on / off status caused by excessively high induced voltage of the outlet pressure plate in high voltage level scenarios, thereby improving the absolute accuracy of automatic inspection of relay protection.

[0045] The beneficial effects of this invention are:

[0046] The monitoring system and method for outlet pressure plates based on transient identification and adaptive compensation successfully achieves a technological leap from traditional static level monitoring to deep physical feature monitoring by introducing high-frequency sampling and transient entropy identification algorithms. It can not only accurately identify the normal operation and deactivation status of the pressure plates, but also keenly detect hidden contact defects caused by oxidation, fatigue, or surface contamination, thus providing reliable early warning signals before the fault evolves into a failure to operate, greatly improving the intrinsic safety level of the secondary circuit. Simultaneously, the adaptive logic based on dynamic reference offset compensation completely solves the long-standing problem of false alarms caused by DC grounding interference in the industry. Through a proportional discrimination model, it achieves automatic adaptation to DC systems of all voltage levels, significantly reducing the difficulty of installation, commissioning, and operation and maintenance costs. Furthermore, the command action causal verification logic constructed by the system deeply closes the loop between physical side signals and management side commands, effectively preventing safety risks caused by human error and unexpected actions. Combined with multi-dimensional evidence fusion and trend prediction models, it achieves situational awareness and intelligent operation and maintenance throughout the entire life cycle of the pressure plates, providing key technical support for building highly reliable intelligent substations and significantly reducing the workload of maintenance personnel and the operational risks caused by human negligence. Attached Figure Description

[0047] Figure 1 This is a flowchart of the outlet pressure plate monitoring method based on transient identification and adaptive compensation proposed in this invention.

[0048] Figure 2 This is a flowchart illustrating the status judgment process of the outlet pressure plate monitoring method based on transient identification and adaptive compensation proposed in this invention. Detailed Implementation

[0049] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0050] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0051] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0052] This invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of this invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not adhering to the usual scale. Furthermore, the schematic diagrams are merely examples and should not be construed as limiting the scope of protection of this invention. In actual fabrication, the three-dimensional spatial dimensions of length, width, and depth should be included.

[0053] Furthermore, in the description of this invention, it should be noted that the terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used solely for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. In addition, the terms "first," "second," or "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0054] Unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" in this invention should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; similarly, they can refer to mechanical connections, electrical connections, or direct connections, or indirect connections through an intermediate medium, or internal connections between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0055] like Figures 1-2 As shown, it illustrates a specific embodiment of the present invention:

[0056] Example 1:

[0057] An outlet pressure plate monitoring system based on transient identification and adaptive compensation includes:

[0058] The multi-source high-frequency data acquisition module is electrically and physically connected to the outlet pressure plate circuit and the DC bus circuit. It is configured to synchronously acquire local voltage characteristics reflecting the physical contact state and global reference potential characteristics reflecting the DC background environment, and perform hardware-level noise reduction and range self-adjustment.

[0059] The transient feature extraction module is configured to run a feature analysis operator based on edge computing, and uses the time series transient entropy algorithm to extract micro-nonlinear fluctuation features that reflect the contact stability of the pressure plate.

[0060] The adaptive offset compensation module is configured to calculate the potential offset trajectory of the DC bus in real time, dynamically adjust the discrimination threshold, and realize real-time physical hedging and logical immunity to DC environmental fluctuations.

[0061] The logical causality verification module is configured to integrate operation messages within the substation through the communication interface, perform consistency verification between software instructions and physical behavior, and lock in the causal certainty of monitoring results.

[0062] The abnormal trend prediction module is configured to use a long short-term memory network to model trends in historical state data, predict the probability of physical degradation of the pressure plate, and execute multi-dimensional risk level alarms.

[0063] In this embodiment, the intelligent monitoring system for the outlet pressure plate, equipped with transient identification and adaptive compensation functions, constructs a complete intelligent closed-loop system from bottom-level signal acquisition to top-level trend prediction through the deep collaboration of the five core modules mentioned above. The multi-source high-frequency data acquisition module ensures the breadth and purity of the sensing data; the transient feature extraction and adaptive offset compensation module completes the transition from data to physical information at the edge, solving the two major industry pain points of "environmental interference" and "hidden loose connections"; the logical causal verification module locks in the authenticity of the conclusion by introducing external causal evidence; and the abnormal trend prediction module extends the monitoring time scale to the future, providing solid technical support for the lean operation and maintenance of the digital power grid.

[0064] Specifically, the multi-source high-frequency data acquisition module performs the following details of sensing and preprocessing: It utilizes a high-precision differential arithmetic unit to extract the potential difference between the two ends of the pressure plate relative to the DC common reference point; it designs an anti-aliasing filter network with multi-order steep descent characteristics to suppress high-frequency interference components outside the sampling bandwidth; it employs an ADC chip with synchronous sample-and-hold technology to ensure that the time deviation of physical quantity conversion between channels is controlled within the sub-microsecond range; and it performs mean calibration and noise removal in the digitization process to construct a feature vector matrix reflecting the true physical characteristics of the contacts.

[0065] In this embodiment, meticulous hardware design is the physical foundation for ensuring high confidence in the monitoring results. The differential acquisition architecture effectively cancels common-mode noise generated by complex stray capacitances within the protective enclosure. This "hardware-software combined" protection strategy ensures that every potential fluctuation point captured by the system has a true physical orientation, rather than being a backflow of external environmental interference. Extremely high synchronicity sampling guarantees accurate reconstruction of local signals and global references in the digital domain, which plays an irreplaceable role in calculating microscopic characteristic signals reflecting contact quality.

[0066] Specifically, the transient feature extraction module extracts the micro voltage fluctuation features in the following details: At the microprocessor edge, multi-scale time-frequency decomposition is performed on the captured waveform sequence to calculate the entropy value statistical score reflecting the autocorrelation of the sequence; by comparing the repetition pattern between sub-sequences, a judgment component reflecting whether there is micro-arc discharge or mechanical jumping in the physical contact is obtained, and it is matched with a health model that incorporates real-time environmental weights to output a quantified physical contact performance score.

[0067] In this embodiment, the transient feature extraction module achieves a "precise translation" of the deep physical meaning of electrical waveforms. The application of transient entropy theory enables the system to identify contact instability phenomena hidden behind seemingly normal "stable voltages" displayed on conventional monitoring equipment. Through this vectorized description, the system can not only determine whether the pressure plate is connected, but also further observe the physical degradation trajectory of the contact surface. This refined perception capability provides a quantitative evaluation basis for the intelligent management of power equipment.

[0068] Specifically, the details of the threshold correction performed by the adaptive offset compensation module are as follows: Real-time tracking of the slope of the DC bus voltage change relative to the nominal level; if the accumulated offset exceeds the preset safety dead zone, the threshold reset algorithm is automatically triggered, and the logical judgment benchmark is physically shifted proportionally according to the offset vector; at the same time, the balance offset caused by the current bus insulation state to ground is calculated in real time, and it is added to the judgment threshold as a dynamic margin in real time.

[0069] In this embodiment, the proactive offset hedging logic endows the system with "logic self-healing capability" in response to changes in the DC environment. In actual substation operation, potential fluctuations caused by DC system grounding or large load switching can easily trigger a "storm" of monitoring alarms. By using real-time offset translation, the judgment criteria are ensured to always "float" above the current operating background. This dynamic threshold management method greatly reduces the false alarm frequency and achieves high sensitivity tolerance for complex electrical conditions.

[0070] Specifically, the details of the consistency verification performed by the logical causal verification module are as follows: an event-triggered message parsing engine is established to capture and parse the station area network control sequence containing the pressure plate number, expected action logic, and exact action time in real time; a logical association matrix is ​​constructed to map the voltage change event sensed by the physical layer to the execution timing of the operation command parsed by the protocol layer; under the premise that the two have a high probability of overlap, a final status report reflecting the operation confidence is output in combination with transient characteristics.

[0071] In this embodiment, the logical causal verification module transforms simple electrical parameter detection into analysis with "logical reasoning" capabilities. By constructing a logical correlation matrix, the system can automatically eliminate "false alarms" caused by inductive coupling or DC system interference. This causal comparison not only enhances the authority of the alarm but also provides a detailed chain of evidence for accident tracing, realizing full-chain digital auditing of the "command-behavior-feedback" process of critical circuits in substations.

[0072] Specifically, the abnormal trend prediction module performs trend modeling and early warning in the following details: it retrieves historical transient entropy indices, compensated effective voltage values, and correction coefficients affected by ambient humidity from the local database; it uses the hidden layer memory function of the long short-term memory network to capture the mathematical evolution trajectory of contact performance aging and outputs a time confidence window reflecting the occurrence of severe failure of the pressure plate; when the predicted value deviates from the safe range, it automatically generates a performance degradation alarm message containing waveform evidence summary and pushes it to the mobile operation and maintenance terminal.

[0073] In this embodiment, the anomaly trend prediction module effectively expands the monitoring scope from the "present" to the "future." The degradation of the outlet pressure plate is often a gradual process, progressing from microscopic performance loss to macroscopic impedance abrupt changes. By capturing the nonlinear characteristics of such complex time-series sequences through a long short-term memory network, the system can identify latent precursors of failure, transforming post-failure maintenance into proactive intervention, significantly enhancing the safety margin of power grid operation.

[0074] Example 2:

[0075] The outlet pressure plate monitoring system based on transient identification and adaptive compensation includes the following steps:

[0076] Step S1: The local voltage characteristics reflecting the physical state of the outlet pressure plate and the global reference potential characteristics reflecting the background operating environment of the DC system are acquired in real time through the multi-channel high-frequency synchronous signal acquisition hardware group. The original sampled data stream is subjected to physical-level noise filtering, dynamic range adaptive adjustment and synchronization alignment processing to obtain a digital high-fidelity multi-source synchronous signal sequence with timestamp information.

[0077] Step S2: The microprocessor unit performs edge processing on the digital synchronization signal sequence. The time series transient entropy algorithm is used to extract the nanosecond to millisecond level micro voltage fluctuation characteristics generated by the outlet pressure plate at the moment of the engagement and disengagement action, thereby quantifying the physical health of the contact surface. At the same time, the adaptive offset compensation model is called to dynamically and in real time calibrate the preset logic judgment threshold based on the drift of the DC bus voltage relative to the nominal level and the DC system's balance with ground.

[0078] Step S3: Based on the time series transient entropy characteristic index and the calibrated dynamic judgment threshold, calculate the accurate real-time physical position status of the outlet pressure plate, and combine it with the causal chain of operation instructions with causal correlation obtained through the substation area network to perform a deep logical consistency check. On the basis of determining that the physical displacement is consistent with the instruction, further identify whether there is physical contact performance degradation, and output a diagnostic conclusion including the risk level of the loose connection and the specific fault type.

[0079] Step S4: Store the logic judgment state, transient entropy index, real-time potential difference, and bus operation parameters into the local storage unit to generate a historical operation curve reflecting the evolution of physical performance throughout the entire life cycle of the outlet pressure plate. Use a machine learning model based on long short-term memory network to perform time-series trend modeling on the historical operation curve, predict the probability of anomalies and the estimated failure time of the outlet pressure plate in future operation cycles, and trigger multi-level alarm messages to be sent to the remote operation and maintenance platform according to the severity of the risk.

[0080] In this embodiment, the intelligent monitoring method for the outlet pressure plate with transient identification and adaptive compensation functions achieves coupled monitoring of local voltage fluctuations in the relay protection outlet circuit and the global potential background of the DC system through multi-channel high-frequency synchronous acquisition technology. This eliminates the interference on the accuracy of pressure plate monitoring caused by the complex operating conditions of the DC system, ensuring the high authenticity of the sampled source data. By introducing a time-series transient entropy algorithm, this application can extract nonlinear oscillation characteristics hidden at the microscopic time scale from the seemingly stable potential difference signal. These characteristics directly reflect the oxide layer breakdown process on the pressure plate contact surface, micro-oscillations caused by insufficient pressure, and micro-arc phenomena caused by environmental pollution. This enables early quantitative identification of the pressure plate's loose connection state, solving the problem of transmission The traditional monitoring method has the drawback of only being effective when the circuit is completely disconnected or completely connected. The introduction of the adaptive offset compensation model enables the monitoring system to automatically "correct" its judgment criteria based on the bus voltage level, just like a senior operation and maintenance expert, greatly improving the robustness of the system under extreme DC system conditions. The station area network causal chain logic verification constructs a logical closed loop of "instruction-behavior", and through digital twin-level comparison, it eliminates the influence of "false potential jumps" caused by DC grounding faults on the monitoring conclusions. Finally, by predicting abnormal states through machine learning, it realizes the leap from "passive discovery after fault" to "proactive maintenance before deterioration", which has significant engineering value and technical significance for improving the unattended safety level of smart substations and ensuring the absolute reliability of tripping circuits.

[0081] The proportional discrimination model described in this invention maps the voltage difference across the pressure plate to a normalized feature space. Specifically, the mapping logic of the proportional factor K ensures that the voltage difference across the bus is within a normalized feature space. When fluctuations occur, a threshold is determined. It can satisfy:

[0082] .

[0083] in, Nominal voltage The reference threshold is set below. This allows the judgment benchmark to shift synchronously with the bus voltage, eliminating false alarms caused by bus voltage dips and achieving automatic adaptation to DC systems across all voltage levels at the physical level. At the fault depth identification level, the system uses transient entropy spectral fingerprints to distinguish between different types of loose connections: mechanical loosening exhibits low-frequency forced vibration characteristics of 10Hz-500Hz, metal oxide type presents asymmetric broadband random pulses, and contamination-induced ablation type accumulates high-order harmonic energy above 5kHz. Furthermore, this embodiment constructs a multi-dimensional time-series feature vector integrating historical connection / reconnection counts, transient entropy peak values, static voltage drop drift, and ambient temperature and humidity. It utilizes the memory gating mechanism of the LSTM model to identify the nonlinear evolution of contact performance from microscopic loss to macroscopic impedance abrupt changes, thereby outputting a high-confidence predicted failure time value.

[0084] Specifically, in step S1, the acquisition of features and data preprocessing through the multi-channel high-frequency synchronous signal acquisition hardware group includes: using multiple high-impedance analog front-end acquisition circuits to synchronously access the upper and lower ground potential monitoring points of the outlet pressure plate to sense the micro-potential fluctuations on both sides of the pressure plate; simultaneously accessing the positive and negative ground voltage sampling lines of the DC bus system to obtain the overall bus voltage reference and ground potential offset reference in real time; using an internal high-precision analog-to-digital converter to perform fully synchronous sampling at a sampling rate of no less than 10,000 times per second to obtain the voltage amplitude sequence of each channel at a microsecond time step; using a digital median filter and a moving average window algorithm to filter out high-frequency random electromagnetic pulse interference in the sampling sequence; and dynamically adjusting the signal amplification factor based on the real-time DC bus voltage value to ensure that the maximum effective signal resolution accuracy can be obtained at different voltage levels.

[0085] In this embodiment, the high-impedance analog front-end design is chosen due to the extreme importance of the safe operation of the power system. Its input impedance is designed to be much greater than the inherent impedance of the protection circuit, thereby ensuring that the monitoring device will not cause current diversion or false tripping risks to the original protection tripping logic under any circumstances. The key to synchronous sampling is that the state discrimination of the pressure plate essentially depends on the potential difference between its two ends at the same moment. If the sampling is not synchronized, a huge false voltage drop will be generated under the background of high-frequency interference, leading to false alarms. This application establishes a multi-dimensional coordinate system with the DC bus as the global reference and the two ends of the pressure plate as local monitoring through four channels of fully synchronous real-time sensing. This full-dimensional sensing method can accurately identify the difference between the "false change in voltage to ground" caused by the overall drift of the DC bus potential and the actual displacement of the pressure plate. At the same time, the dynamic range adjustment technology ensures that the system can still maintain an extremely high signal-to-noise ratio when the DC system bus experiences voltage drops due to load fluctuations, thereby providing the purest basic data for subsequent extraction of weak transient features.

[0086] Specifically, in step S2, the extraction of features and dynamic calibration using the time series transient entropy algorithm includes: extracting specific time slices of the potential difference signal at both ends of the pressure plate before and after the voltage change instant; using a sliding window algorithm to statistically analyze the frequency of repetitive patterns within a given similarity tolerance range in the slice sequence; and calculating a transient entropy index that reflects the complexity and determinism of the signal sequence to characterize the microscopic physical stability of the pressure plate contact surface. Simultaneously, the difference between the positive and negative potentials of the DC bus is calculated to obtain the real-time absolute voltage of the bus, and based on the percentage deviation of the absolute voltage from the nominal voltage, the amplitude requirement of the logic judgment threshold is adjusted in real time by using a preset threshold compensation mapping logic.

[0087] In this embodiment, the time-series transient entropy algorithm has a fundamental technical advantage over traditional effective value and average value determination. In actual operation, if a loose connection occurs in the outlet pressure plate (such as loose screws or insulating dust between contacts), the waveform will not immediately show a huge amplitude drop, but will generate a large number of chaotic and irregular micro pulses. As a tool for evaluating the complexity of nonlinear time series, transient entropy can accurately capture this increase in "chaos". When the pressure plate is physically closed well, the voltage sequence has extremely strong determinism and pattern repeatability, and the entropy value is extremely low; conversely, if there is a weak arc at the contact point or unstable contact pressure due to mechanical loosening, the determinism of the signal sequence decreases, and the entropy value increases significantly. This indicator provides direct physical evidence for the risk of loose connection. At the same time, the adaptive offset compensation logic solves the logic failure problem caused by the common "bus low voltage" on the DC side of the power system. Traditional monitoring devices often set fixed thresholds. Once the bus voltage fluctuates normally due to the start of a large load in the station, the fixed threshold is very likely to cause false alarms. This application achieves real-time decoupling between the judgment benchmark and the operating conditions through dynamic calibration, ensuring a perfect balance between monitoring sensitivity and reliability.

[0088] Specifically, in step S3, the determination of the status and type of virtual connection based on the causal chain logic consistency check includes: real-time parsing of operation instruction messages containing operation targets, operation behaviors, and precise execution times through the substation process layer network or station control layer network, and storing the instruction as the expected result in a temporary comparison stack; when the bottom-level sampling detects a change in the pressure plate potential characteristics exceeding the warning threshold, immediately backtracking and comparing whether there is a valid operation record in the instruction comparison stack within a preset time neighborhood; if the physical change and the software instruction are highly consistent in terms of time axis and action type, it is determined to be a normal operation state; otherwise, an alarm is triggered; if both are consistent but the monitored transient entropy index exceeds the health benchmark, the virtual connection is further characterized as a mechanical loosening type, a metal oxidation type, or a contamination isolation type by combining the pressure gradient slope sent by the pressure sensor and the waveform distortion pattern identified by the transient entropy spectrum distribution characteristics. The weighted fusion arbitration follows the DS combination rule for electrical, mechanical, and thermal three-dimensional evidence. , , The confidence assignment after fusion satisfy:

[0089] .

[0090] Among them, K conflict This refers to the conflict factor between pieces of evidence. The initial confidence assignment of the electrical, mechanical, and thermal three-dimensional evidence. Based on the preset trapezoidal membership function, the transient entropy deviation, contact pressure drop gradient and local temperature rise rate are mapped to the [0,1] interval as the basic probability assignment value of the corresponding fault type, thereby realizing the logical transformation of heterogeneous sensing data into mathematical evidence space.

[0091] In this embodiment, the causal chain verification logic is the core component for improving monitoring confidence. In complex power grid secondary circuits, DC single-point grounding faults often generate "illusory voltages" across the pressure plates due to charge redistribution. If judged solely by voltage, the system will inevitably report an error. This embodiment introduces the key dimension of "operational causality," confirming a valid action only when "a human has issued an instruction" and "the physical layer generates a corresponding displacement." This logic verification method greatly filters out spurious displacements caused by natural environmental interference and electrical equipment faults. Furthermore, the further subdivision and diagnosis of loose connection types demonstrates a deep application of intelligence. Different types of loose connections exhibit different "fingerprints" in transient waveforms: for example, waveforms caused by mechanical loosening have obvious mechanical vibration frequency components, while waveforms caused by metal oxidation exhibit typical semiconductor rectification and electrical penetration characteristics. Through in-depth analysis of these microscopic fingerprints, the system can provide clear maintenance recommendations, such as "cleaning the contacts is the symptom" or "tightening the screws is needed," greatly improving the targeting and efficiency of maintenance work.

[0092] Specifically, in step S4, predicting abnormal evolution trends and issuing multi-level alarm signals includes: aligning data on the historical number of times the outlet pressure plate has been engaged and disengaged, the peak transient entropy at each action, and the voltage difference drift during static operation to construct a multi-dimensional time series sample set reflecting the physical performance degradation of a single pressure plate; inputting this sample set into a pre-trained long short-term memory network model, capturing the long-term evolution trend of pressure plate performance fluctuations through the forgetting and remembering mechanisms within the model, and outputting a confidence score reflecting the occurrence of deterministic failures in the future; automatically generating diagnostic reports of different levels such as "normal observation," "performance degradation warning," or "serious fault alarm" based on the threshold range of the score, and using digital signature technology to ensure the integrity and authenticity of the monitoring messages during transmission to the remote backend. The pre-training sample set of the Long Short-Term Memory (LSTM) network comes from the accelerated aging test data of pressure plates simulated in the laboratory environment, as well as the desensitized historical commissioning and decommissioning feature sequences of pressure plates in the substation operation and maintenance database. During the training process, the Z-score normalization method is used to perform dimensional normalization on the heterogeneous data, including temperature, humidity, and voltage drop drift, to ensure the convergence performance of the model when dealing with long-range time series correlations.

[0093] In this embodiment, the introduction of machine learning extends the monitoring system's capabilities from "state awareness" to "lifespan assessment." As a type of mechanical contact equipment, the performance degradation of outlet pressure plates is often a slow process caused by environmental corrosion and wear. Long Short-Term Memory (LSTM) networks are particularly adept at handling this type of time-series data with long-term temporal correlation, capable of identifying subtle decay trends that are undetectable by the naked eye and conventional statistical methods. By calculating the confidence score for future failures, the system can provide maintenance personnel with a basis for "tiered management": pressure plates with low scores are recommended for inclusion in the next power outage maintenance plan; pressure plates with extremely high scores are immediately triggered for emergency handling procedures. This data-driven predictive maintenance not only reduces the risk of unplanned power outages but also significantly extends the service life of substation secondary equipment, reducing overall life-cycle maintenance costs.

[0094] Example 3:

[0095] like Figure 2As shown in the diagram, the state discrimination process based on transient identification and adaptive compensation is as follows: In this process, the system first determines whether the calculation result of the proportional discrimination model is within the logical input range: if yes, it is determined to be in a connected state; if no, it is determined to be in an exited state. For the pressure plate in the connected state, the system further determines whether the transient entropy characteristic value is not greater than the preset safety benchmark: if yes, it outputs a normal input state; if no, it is marked as a suspected loose connection state and triggers multi-dimensional evidence fusion and causal chain verification. If there is no legitimate instruction and there is a logical deviation between the electrical voltage drop and the mechanical pressure, an illegal displacement alarm is issued. If there is a legitimate instruction and the pressure meets the standard, but microscopic arc erosion characteristics are identified, a contamination loose connection alarm is issued.

[0096] In this embodiment, when a positive ground fault occurs in the DC system of the substation, the positive voltage to ground instantly drops to zero. At this time, because the acquisition hardware group is simultaneously connected to the positive voltage to ground sampling line and the negative voltage to ground sampling line reference channel of the DC bus system, the microprocessor unit immediately calculates the offset vector of the global reference potential of the DC bus. The adaptive offset compensation module, based on this offset vector, shifts the potential judgment benchmark at both ends of the pressure plate from the original equilibrium potential to the current ground offset level in real time. Subsequently, when the maintenance personnel normally exit the outlet pressure plate, the system captures the pressure plate potential difference jumping from the microvolt level to the full voltage level, and the sample entropy of this transient waveform is extremely low. At the same time, the causality verification module matches the operation command from the background. The system ultimately determines it as "normal exit operation" and will not generate any erroneous "pressure plate fault" or "abnormal jump" alarms due to the disruption of the bus-to-ground balance.

[0097] In this embodiment, due to prolonged operation and a humid environment, a micro-oxide layer invisible to the naked eye formed on the contact surface of a certain outlet pressure plate, causing nonlinear fluctuations in contact resistance. During static monitoring, the microprocessor detected weak and irregular random fluctuations in the potential difference signal across the pressure plate. The transient entropy calculation showed that the complexity of this sequence significantly exceeded the health baseline range. The diagnostic output module, through causal verification, found that there were currently no operational commands, determining that the fluctuations originated from physical performance degradation. The abnormal state prediction module, by analyzing the growth curve of this entropy value over the past week, used a long short-term memory network to predict that the probability of the pressure plate experiencing false triggering or failure to operate within the next month would exceed the preset safety level, thus issuing a level-two performance warning. Based on the fault type indication in the message, maintenance personnel carried a special cleaning agent to perform targeted maintenance on the contacts, eliminating the potential tripping circuit failure hazard.

[0098] The intelligent monitoring scheme for the outlet pressure plate with transient identification and adaptive compensation functions disclosed in this invention is based on the core technology of acquiring local voltage characteristics reflecting the physical state of the pressure plate and global reference potential characteristics reflecting the background environment of the DC system in real time through a multi-channel high-frequency synchronous signal acquisition hardware group. It utilizes physical-level noise filtering and range self-calibration processing to ensure the high authenticity and accuracy of multi-source signal data. The system uses a microprocessor unit to run a time-series transient entropy algorithm to extract nonlinear features from the micro-voltage fluctuations at the moment of pressure plate action to quantify the physical health of the contact surface. Simultaneously, it incorporates an adaptive offset compensation model based on the real-time potential drift of the DC bus. The system calibrates the judgment threshold stepwise, achieving deep decoupling between the judgment benchmark and complex DC operating conditions. Based on this, the system constructs a proportional discrimination model to accurately identify the true physical location of the pressure plate and introduces a deep verification of the consistency of the causal chain execution logic of the operation sequence. Through weighted fusion arbitration of multi-dimensional sensing parameters, it effectively identifies unexpected displacements and hidden loose connection faults such as mechanical loosening and metal oxidation. Finally, the solution relies on a Long Short-Term Memory (LSTM) network to model the trend of historical operating curves, predicting the probability of abnormality and the time of deterioration failure of the outlet pressure plate in future cycles, thereby assisting maintenance personnel in timely detection of hidden defects and significantly reducing the risk of relay protection circuit failure.

[0099] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

[0100] Many other changes and modifications can be made without departing from the concept and scope of this invention. It should be understood that this invention is not limited to the specific embodiments, and the scope of this invention is defined by the appended claims.

Claims

1. An outlet pressure plate monitoring system based on transient identification and adaptive compensation, characterized in that, include: The multi-source high-frequency data acquisition module includes four sampling channels. The sampling channels are connected to the output pressure plate circuit in the relay protection cabinet through electrical connection. The first and second sampling channels are respectively connected to the upper and lower ground potential monitoring points of the corresponding output pressure plates to obtain local voltage characteristics. The third and fourth sampling channels are synchronously introduced into the positive-to-ground voltage sampling line and the negative-to-ground voltage sampling line of the DC bus system, respectively, to obtain the global reference potential characteristics that reflect the background environment of the DC system in real time. The microprocessor module receives the raw waveform data stream sent by the multi-source high-frequency data acquisition module and is configured to execute the core solution operator, which includes a normalized adaptive offset compensation operator and a time series transient entropy identification operator. The diagnostic output module has an embedded operation sequence causal chain matching engine. It monitors operation command messages in the substation area network in real time through the data bus and establishes a time correlation matrix between physical displacement events and management logic sequences. Based on the probability of the overlap between the mutation time of the proportional factor K and the instruction authorization time window, combined with the transient entropy feature value, it outputs a comprehensive diagnostic message including the engagement / disengagement status, contact health, and operation compliance. The abnormal trend prediction module integrates a multi-dimensional time-series feature vector containing historical engagement and disengagement times, transient entropy peak sequence, static operating voltage drop drift, and environmental temperature and humidity correction coefficients. It uses a long short-term memory neural network model to capture the nonlinear evolution law of pressure plate performance from microscopic loss to macroscopic impedance change, and outputs the estimated failure time of the pressure plate contacts and the failure probability curve within a specific future period.

2. The outlet pressure plate monitoring system based on transient identification and adaptive compensation according to claim 1, characterized in that, The calculation steps of the normalized adaptive offset compensation operator of the microprocessor module are as follows: Calculate the arithmetic mean of the voltages of the positive and negative poles of the DC bus to ground in real time, and extract the current DC system's ground bias coefficient; A proportional discrimination model is constructed with the real-time full voltage span of the DC bus as the denominator and the absolute value of the instantaneous potential difference between the two ends of the pressure plate as the numerator. The electrical state of the pressure plate is mapped to a normalized characteristic space that shifts synchronously with the bus potential in real time. The calculated scaling factor K automatically adapts to DC systems of different voltage levels and eliminates ground potential drift interference. The scaling factor formula is: ； in, and These are the voltages to ground at the top and bottom of the pressure plate, respectively. and These are the positive and negative voltages of the DC bus to ground, respectively. The calculation steps for the time series transient entropy identification operator are as follows: Maintain a high-speed data loop buffer with a sampling frequency of not less than 10kHz, and automatically capture the transient voltage sequence X of a preset duration when the voltage plate switching is detected; The information entropy algorithm is used to quantify the dispersion and complexity of the sequence in the time domain, and to calculate the transient entropy characteristic value reflecting the micro-contact quality of the contacts. This value is then used to identify the hidden contact performance degradation masked by steady-state voltage drop. The calculation formula is: ； Where X is the extracted transient voltage sequence sample; For the first The probability of a sampling point occurring within a given similarity tolerance range.

3. The outlet pressure plate monitoring system based on transient identification and adaptive compensation according to claim 1, characterized in that, The transient identification operator has a fault characterization function based on spectral fingerprinting. By performing time-frequency domain decomposition on the captured high-frequency sequence, it identifies and distinguishes the following types of virtual connections: When low-frequency forced vibration characteristics of 10Hz-500Hz are detected and the transient entropy exceeds the first threshold, it is determined to be a mechanical loosening type of loose connection. When a broadband random pulse with asymmetric characteristics is detected and the transient entropy exceeds the second threshold, it is determined to be a metal oxide type virtual connection. When high-order harmonic energy clusters above 5kHz are identified and accompanied by microscopic discharge characteristics, it is determined to be a fouling-ablation type of loose connection.

4. The outlet pressure plate monitoring system based on transient identification and adaptive compensation according to claim 1, characterized in that, When performing causal chain matching, if the diagnostic output module detects a state reversal of the pressure plate proportional factor K but no valid instruction is found within the corresponding operation instruction authorization time window, it determines that the displacement is an unexpected illegal change, accidental contact, or serious electrical insulation breakdown, and immediately raises the alarm level to the highest level.

5. The outlet pressure plate monitoring system based on transient identification and adaptive compensation according to claim 1, characterized in that, The diagnostic output module also integrates a comprehensive diagnostic operator, which is based on a multi-source evidence fusion mechanism and supports the input of mechanical contact pressure vectors provided by pressure sensors and local temperature rise vectors provided by infrared sensors. The multi-source evidence fusion mechanism adopts the Durmst-Schaffer (DS) evidence theory. ； in, The confidence level of the fused diagnosis; This represents the degree of conflict between evidence from different sensors.

6. The outlet pressure plate monitoring system based on transient identification and adaptive compensation according to claim 1, characterized in that, The multi-dimensional time-series feature vector of the abnormal trend prediction module for: ； Where C represents the cumulative number of throws and withdrawals. This is a sequence of transient entropy peaks. The static pressure drop drift is represented by T and RH, which are correction factors for ambient temperature and humidity.

7. A method for intelligent monitoring of the outlet pressure plate using the system described in any one of claims 1-6, characterized in that, The steps include the following: S1: The data acquisition module acquires the ground potential signals on both sides of the outlet pressure plate and the global reference potential signal of the DC bus, and captures the high-frequency transient voltage sequence at the moment of pressure plate potential switching. S2: Time-frequency domain feature extraction is performed on the high-frequency transient voltage sequence. The transient entropy feature value reflecting the micro-contact quality of the contact is calculated using the information entropy algorithm, and the probability distribution function is used. Calculate the information entropy of the sample sequence Abnormal oscillations during the pressure plate closure process can be identified through feature vectors; S3: Real-time calculation of the DC bus ground offset coefficient in a DC system A proportional discrimination model based on global reference potential is constructed, and the logical engagement / disengagement state of the outlet pressure plate is adaptively identified using a proportional factor. Multidimensional sensing parameters are introduced for weighted fusion arbitration. S4: The physical displacement of the outlet pressure plate is verified by operating the causal chain of the sequence, and unexpected state changes are identified and intercepted. Historical operating data is used to predict and analyze the deterioration trend of the pressure plate.

8. The intelligent monitoring method for an outlet pressure plate according to claim 7, characterized in that: DC bus ground offset coefficient in S3 The calculation formula is: ； in, and These are the positive and negative voltages of the DC bus to ground, respectively.

9. A substation secondary circuit intelligent operation and maintenance terminal, internally embedding the monitoring system according to any one of claims 1-6 and the monitoring method according to any one of claims 7-8, characterized in that, The terminal performs complex entropy calculations and causal chain matching locally, and sends the simplified results after logical filtering and in-depth diagnosis to the digital management platform through a dedicated power communication protocol.

10. A substation secondary circuit intelligent operation and maintenance terminal according to claim 9, characterized in that, The terminal has a standardized plug-and-play sensor interface, which can flexibly connect auxiliary sensing parameters according to the needs of the field environment, and uses a built-in weighted discrimination model to solve the problem of difficulty in identifying the operation and shutdown status caused by excessively high output pressure plate sensing voltage in high voltage level scenarios.

Images