A wind farm booster station safety management system and method based on self-response insulating molecular brush and edge computing

Abstract

This invention relates to the field of safety monitoring and insulation protection technology for wind farm booster stations, and discloses a safety management system and method based on self-responsive insulating molecular brushes and edge computing. The system includes a self-responsive insulating molecular brush coating, a sensor array, an edge computing gateway, and a station-level monitoring backend. After capturing high-energy electrons generated by partial discharge, the coating releases a crosslinking agent and forms a transient insulating gel in situ to suppress the discharge. The edge computing gateway runs a molecular brush state assessment model, extracting the width, decay time, and adjacent pulse interval features of the discharge pulse from the high-frequency partial discharge spectrum. When the pulse width is less than a first threshold, the decay time is less than a second threshold, and there are no subsequent pulses, it is identified as a successful self-healing event, and the coating consumption status is assessed. This invention transforms the material's self-healing behavior into a quantifiable digital quantity, achieving a closed-loop fusion of physical fault suppression and digital early warning.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring and insulation protection technology for wind farm booster stations. Specifically, it relates to a safety management system and method for wind farm booster stations based on self-responsive insulating molecular brushes and edge computing. Background Technology

[0002] The substation of a wind farm plays a crucial role in boosting the low-voltage electricity output from the wind turbines to the grid-connected voltage. The insulation condition of its internal high-voltage equipment, such as transformers, instrument transformers, and gas-insulated switchgear, directly affects the power generation utilization rate and operational safety. Partial discharge is an early sign of insulation degradation. If it is not suppressed in time, it will gradually develop into destructive breakdown, leading to unplanned outages and significant economic losses.

[0003] Currently, wind farm booster stations commonly deploy online monitoring systems, collecting status data through sensors such as partial discharge sensors and dissolved gas sensors in oil, and using locally deployed edge computing gateways or remote cloud platforms for threshold alarms and trend analysis. For example, existing edge computing-based substation equipment status monitoring systems aggregate and perform preliminary analysis of multi-source sensor data through edge gateways, reporting abnormal events to the main station. Such solutions focus on sensing and transmitting fault signals, failing to proactively intervene in the physical development of faults. By the time severe discharge is detected, irreversible insulation damage has already occurred, and protective measures can only passively isolate the fault, unable to suppress it in its nascent stage.

[0004] On the other hand, self-healing insulating materials have been attempted to be introduced into electrical equipment. For example, self-healing microcapsules for cable insulation have been disclosed in the prior art, which release a repair agent to fill cracks when cracks appear in the insulation layer. However, such materials are usually designed for mechanical damage, and the repair process is mostly a one-time, irreversible consumption, unable to cope with continuous and frequent electrical stress impacts such as partial discharge. More importantly, these self-healing behaviors occur inside the material and cannot be detected by external monitoring systems, making it impossible for maintenance personnel to know the location, frequency, and degree of self-healing that has occurred in the insulation, making it difficult to effectively coordinate the self-healing capabilities of the material with digital maintenance systems.

[0005] Therefore, existing technologies suffer from the following problems: In the safety management of wind farm booster stations, physical fault suppression and digital status perception are disconnected. Traditional online monitoring cannot actively reduce discharge energy at the onset of discharge, leading to a passive monitoring and continuous degradation dilemma; while existing self-healing materials lack a perceptible and quantifiable mechanism for repair behavior, making the self-healing process an information blind spot and unable to support predictive maintenance decisions. This is particularly prominent in remote wind farm scenarios with frequent extreme weather and unattended booster stations, where maintenance personnel cannot formulate differentiated maintenance strategies based on the actual self-healing capability and historical consumption status of the insulation, resulting in both over-maintenance and missed inspection risks. Summary of the Invention

[0006] This invention aims to address the disconnect between physical fault suppression and digital status perception in the safety management of existing wind farm booster stations. Specifically, existing online monitoring systems cannot proactively reduce discharge energy at the onset of partial discharge, leading to passive monitoring and continuous degradation. Furthermore, self-healing insulating materials lack external sensing mechanisms for their repair processes, making this a blind spot for maintenance personnel and hindering differentiated and predictive maintenance decisions. Therefore, this invention aims to solve the technical problem of simultaneously achieving proactive discharge energy suppression and perceptible, quantifiable self-healing behavior at the fault initiation stage, thereby integrating the fault self-healing mechanism into a digital operation and maintenance closed loop.

[0007] To address the aforementioned technical problems, this invention provides a wind farm booster station safety management system based on self-responsive insulating molecular brushes and edge computing. The system includes: a self-responsive insulating molecular brush coating applied to key insulation interfaces of the booster station equipment; a sensor array for real-time monitoring of multi-physics characteristic signals generated by the self-responsive insulating molecular brush coating in response to partial discharge; an edge computing gateway deployed locally at the booster station and communicatively connected to the sensor array; and a station-level monitoring backend communicatively connected to the edge computing gateway. The self-responsive insulating molecular brush coating comprises a substrate segment anchored to the surface of an insulating substrate, an electron acceptor molecular cage connected to the substrate segment, and a self-assembled motif located at the end of the molecular brush coating. The electron acceptor molecular cage encapsulates a non-conductive crosslinking agent. The self-responsive insulating molecular brush coating is configured such that upon capturing high-energy electrons generated by partial discharge, the electron acceptor molecular cage ruptures and releases the non-conductive crosslinking agent. The non-conductive crosslinking agent mediates a reversible physical crosslinking between the self-assembled motif and recognition groups on adjacent molecular brush coatings, thereby forming a transient insulating gel in situ at the discharge point to suppress discharge. The edge computing gateway is configured to receive and process the multiphysics feature signals from the sensor group, run the molecular brush state assessment model, extract the width, decay time and adjacent pulse interval features of the discharge pulse from the high-frequency partial discharge spectrum, and when the width of a single discharge pulse is less than a first threshold, the decay time is less than a second threshold and there are no subsequent continuous discharge pulses, the discharge pulse is identified as a successful self-healing event, the coating consumption status is assessed, and the analysis results are sent to the station-level monitoring backend.

[0008] The multi-physics characteristic signals refer to various physical quantity signals generated by the self-responsive insulating molecular brush coating during partial discharge, reflecting the occurrence of self-healing events and changes in coating state. These signals include, but are not limited to: characteristic electromagnetic transient signals generated by the rapid suppression of discharge pulses; leakage current change signals generated by the abrupt change in conductivity of the insulating interface due to gel formation; trace acoustic emission or vibration signals generated by the release of non-conductive crosslinking agents and the gel formation process; and characteristic gas generation rate change signals caused by the suppression of discharge energy. By simultaneously monitoring signals from multiple physical field dimensions, the system can reliably distinguish self-healing events from ordinary interference. The aforementioned multi-physics characteristic signals collectively constitute the data basis for the molecular brush state assessment model to identify self-healing events and assess coating consumption state. Among them, the characteristic electromagnetic transient signals provide information on the width, decay time, and amplitude of the discharge pulse; the leakage current change signals reflect the transient change in conductivity of the insulating interface; the acoustic emission or vibration signals mark the physical timestamps of the release of non-conductive crosslinking agents and gel formation; and the characteristic gas generation rate change signals chemically confirm the cumulative effect of the suppression of discharge energy.

[0009] At the implementation level, the sensor group includes at least a leakage current sensor, a partial discharge sensor, and a dissolved gas sensor in the oil. The partial discharge sensor is an ultra-high frequency partial discharge sensor or a high frequency current sensor; the dissolved gas sensor in the oil is a hydrogen sensor or a methane sensor. Preferably, the sensor group further includes an optical fiber vibration sensor or an acoustic emission sensor for capturing the acoustic waves or vibration signals generated during the release of the non-conductive crosslinking agent and gel formation.

[0010] The core of the self-responsive insulating molecular brush coating lies in integrating sensing-response-recovery functions into a single molecular system. Structurally, a preferred embodiment of the self-responsive insulating molecular brush coating includes: a substrate segment anchored to the surface of an insulating substrate; an electron acceptor molecular cage connected to the substrate segment, encapsulating a releasable non-conductive crosslinking agent; and a self-assembled motif located at the end of the molecular brush coating, configured to form a reversible physical crosslink with a recognition group on an adjacent molecular brush coating after the release of the non-conductive crosslinking agent. The design of the electron acceptor molecular cage is crucial for achieving selectivity: its redox potential window is precisely designed between -1.5V and -2.5V relative to a saturated calomel electrode, allowing it to be reduced and opened only by high-energy electrons generated by partial discharge, typically with energies greater than 3eV, while remaining inert to power frequency electric fields, temperature fluctuations, or background free radicals in oil under normal operating conditions. In a specific example, the electron acceptor molecular cage may be a redox-active cage formed from a triphenylamine derivative; the self-assembled motif and the recognition group may be host-guest inclusion pairs that can form reversibly crosslinked pairs through supramolecular interactions, such as an adamantyl group and a cyclodextrin group, or a cucurbituril group and a methyl viologen group.

[0011] The molecular brush state assessment model is a lightweight temporal anomaly detection model. This model is further configured to: extract a feature vector from the multi-physics characteristic signal for each discharge pulse, wherein the feature vector includes at least the pulse peak amplitude, pulse rise time, pulse half-peak width, decay time constant, time interval with the preceding pulse, discharge phase angle, and sensor channel number; identify discharge pulses with burst-rapid decay characteristics based on the feature vector and recognize them as successful self-healing events; statistically analyze the frequency, amplitude, and phase distribution of the successful self-healing events within a specific time window; issue an insulation degradation warning when the frequency or cumulative energy exceeds a preset threshold; and calculate and output the remaining effective lifetime of the coating based on the total number of self-healing events or the characteristic gas trend. The lightweight temporal anomaly detection model can be one of a single-class anomaly detection network based on TransformerEncoder, an autoencoder based on Long Short-Term Memory network, or an anomaly scoring model based on temporal convolutional network.

[0012] At the methodological level, the wind farm booster station safety management method based on self-responsive insulating molecular brushes and edge computing provided by this invention includes the following steps: In response to partial discharge, the electron acceptor molecular cages in the self-responsive insulating molecular brush coating capture high-energy electrons and then break, releasing the encapsulated non-conductive crosslinking agent. The non-conductive crosslinking agent mediates the formation of reversible physical crosslinks between self-assembled motifs and recognition groups on adjacent molecular brush coatings, forming a transient insulating gel in situ at the discharge point to suppress discharge; a sensor array captures the multiphysics characteristic signals generated by the release and gel formation process in real time and sends them to the edge computing network. The multi-physics characteristic signal is processed by the molecular brush state assessment model in the edge computing gateway. The width, decay time and adjacent pulse interval features of the discharge pulse are extracted from the high-frequency partial discharge spectrum. When the width of a single discharge pulse is less than a first threshold, the decay time is less than a second threshold, and there are no subsequent continuous discharge pulses, the discharge pulse is identified as a successful self-healing event, and the frequency and energy trend are statistically analyzed. When the frequency or cumulative energy exceeds a preset threshold, an early warning message is generated, and the early warning message and assessment results are sent to the station-level monitoring backend. The station-level monitoring backend generates maintenance suggestions by integrating data from multiple devices.

[0013] The evaluation mechanism for coating consumption status is as follows: First, the total encapsulation amount of the effective molecular cages in the self-responsive insulating molecular brush coating is obtained. This total encapsulation amount can be pre-calibrated before the coating leaves the factory or is put into operation through methods such as thermogravimetric analysis. Then, by performing correlation analysis on the total number of historically identified successful self-healing events and the increment of characteristic gas in the oil during the corresponding operating cycle, the average consumption amount of a single self-healing event is obtained. The characteristic gas is, for example, hydrogen. Based on this, the remaining effective lifetime of the coating is calculated based on the ratio of the total encapsulation amount to the average consumption amount. This lifetime prediction result is output in the form of the remaining number of responsive events or equivalent operating time. The first threshold ranges from 10 to 100 microseconds, and the second threshold ranges from 10 to 80 microseconds. In addition, the method also includes estimating the spatial location of insulation weaknesses based on the phase distribution of self-healing events and sensor channel delays, and asynchronously uploading the original characteristic waveform data of self-healing events to the cloud for model iterative training.

[0014] Beneficial effects

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. Achieving a closed-loop fusion of physical fault suppression and digital early warning for the first time. By actively suppressing discharge through a self-responsive insulating molecular brush coating, while the edge computing gateway uses a sensor array to capture and identify the physical characterization signals of the event, the material's self-healing behavior is transformed into a quantifiable digital quantity, solving the problem of the disconnect between self-healing materials and online monitoring systems.

[0016] 2. It can distinguish successful self-healing events from occasional interference signals. The molecular brush state assessment model in the edge computing gateway identifies self-healing events from interference based on the characteristics of discharge pulse width, decay time, and adjacent pulse intervals, avoiding false alarms caused by occasional discharges or external noise, and making the determination of insulation aging trends more accurate.

[0017] 3. Provides dynamic assessment capabilities for coating consumption status and insulation weaknesses. By statistically analyzing the frequency, energy, and phase distribution of self-healing events, and combining this with the ratio of the total effective molecular cage encapsulation of the coating to the average consumption, the edge computing gateway can estimate the remaining effective lifespan of the coating and pinpoint the spatial location of insulation weaknesses, providing precise guidance for maintenance.

[0018] 4. Elevate the safety management of the substation to an intelligent operation and maintenance mode with proactive immunity and continuous self-optimization. This system directly weakens discharge energy in the early stages of a fault, delaying the process of insulation degradation; at the same time, it continuously analyzes the historical trend of self-healing activities, and issues early warnings when the frequency of self-healing in a certain area increases abnormally. Based on this, the station-level monitoring backend generates differentiated maintenance suggestions, reducing unnecessary power outages and the risk of missed inspections. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall system architecture of Embodiment 1 of the present invention.

[0020] Figure 2 This is a schematic diagram of the molecular brush state evaluation model network structure according to Embodiment 1 of the present invention.

[0021] Figure 3 This is a schematic diagram of the workflow of the method in Embodiment 1 of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and multiple embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the scope of protection of the invention.

[0023] Example 1: Adamantane-Cyclodextrin Host-Guest System and Transformer Model

[0024] System Architecture

[0025] Figure 1 This embodiment illustrates a wind farm booster station safety management system based on self-responsive insulating molecular brushes and edge computing. The system consists of an equipment layer, a sensing layer, an edge layer, and an application layer.

[0026] The equipment layer includes high-voltage equipment such as main transformers, high-voltage instrument transformers, and gas-insulated switchgear bays within the substation. At critical insulation interfaces of these devices, such as the winding turn insulation paper of the main transformer, the surface insulation of the equalizing spheres of the high-voltage bushings, and the surface of the basin-type insulators of the gas-insulated switchgear, a self-responsive insulating molecular brush coating with a thickness of 50 to 500 nanometers is applied. This coating is uniformly applied using chemical vapor deposition or dip coating methods.

[0027] The sensing layer includes a sensor array mounted adjacent to the insulating interface. This sensor array comprises at least one ultra-high frequency partial discharge sensor, one high-frequency current sensor, one high-sensitivity leakage current sensor, and one hydrogen sensor using a palladium alloy thin film. Optionally, fiber optic vibration sensors are also mounted in the non-magnetic area of ​​the device housing to capture weak transient mechanical vibration signals. All sensors are communicatively connected to an edge computing gateway via shielded cables or optical fibers, with sampling rates set from 100 kS / s to 10 MS / s to ensure the capture of millisecond-level transient events. These sensors collectively monitor in real time the multiphysics characteristic signals generated by the self-responsive insulating molecular brush coating in response to partial discharge.

[0028] The edge layer comprises an industrial-grade edge computing gateway deployed within the local control cabinet of the booster station. This gateway integrates a time-sensitive network switch, an AI inference acceleration chip, and solid-state storage. The edge computing gateway runs a molecular brush state assessment model.

[0029] The application layer includes a station-level monitoring backend located in the central control room of the substation, which is connected to multiple edge computing gateways via fiber optic Ethernet. The station-level monitoring backend runs data aggregation and visualization software and integrates an interface for a work order management system.

[0030] Molecular brush coating structure and triggering mechanism

[0031] The self-responsive insulating molecular brush coating of this embodiment has the following structure at the microscopic level: after introducing a layer of silane coupling agent on the surface of insulating paper fiber, a flexible molecular brush base layer is formed by covalently grafting poly(hydroxyethyl methacrylate) base segments; an electron acceptor molecular cage is connected to the end of the base segment; and in the outermost layer of the molecular brush coating, the end of each molecular chain is modified with a self-assembled motif.

[0032] The preparation process of the molecular brush coating is as follows: The first step involved the synthesis of an electron acceptor molecular cage. 1.0 mmol of tris(4-ethynylphenyl)amine and 1.5 mmol of a trialdehyde-resorcinol derivative were dissolved in anhydrous tetrahydrofuran. A catalytic amount of trifluoroacetic acid was added, and the reaction was carried out under argon protection at 60 °C with stirring for 48 hours. A cage-like structure was formed through dynamic imine bond self-assembly. After the reaction, sodium cyanoborohydride was added to reduce the imine bonds to stable carbon-nitrogen single bonds. The product was purified by column chromatography to obtain a white powdery solid, which was the electron acceptor molecular cage. Transmission electron microscopy revealed an outer diameter of approximately 2.5 nm and an inner cavity diameter of approximately 1.2 nm. The redox potential of this molecular cage, measured by cyclic voltammetry, was -1.8 V relative to a saturated calomel electrode, indicating a selective response to high-energy electrons.

[0033] The second step involves encapsulating the non-conductive crosslinking agent. The aforementioned electron acceptor molecular cages are dispersed in a perfluoropolyether liquid with a molecular weight of approximately 500 Da. The mixture is treated under ultrasonic and micro-negative pressure conditions for 2 hours, allowing the perfluoropolyether molecules to enter the inner cavity through the cage wall pores. Subsequently, the system temperature is lowered to -20°C, causing the cage wall pores to shrink, achieving physical encapsulation of the non-conductive crosslinking agent. Thermogravimetric analysis shows that the encapsulation rate is 0.15 grams of perfluoropolyether encapsulated per gram of molecular cage. Based on this encapsulation rate and the total grafting amount of molecular cages in the coating, the total effective encapsulation amount of molecular cages per square centimeter of coating can be pre-calibrated.

[0034] The third step involves grafting a molecular brush coating. A layer of 3-aminopropyltriethoxysilane ethanol solution is impregnated onto the surface of the insulating paper fiber and cured at 110°C for 30 minutes to introduce surface amino groups. Subsequently, a poly(hydroxyethyl methacrylate) base segment is grown on the fiber surface via surface-initiated atom transfer radical polymerization for 2 hours, resulting in a flexible molecular brush base layer approximately 50 nm thick. The encapsulated electron acceptor molecular cage is then chemically coupled to the side chain hydroxyl groups of the base segment via EDC / NHS coupling. Finally, adamantyl alkyl groups and cyclodextrin groups are modified at the ends of the molecular brush via esterification, with the modification ratio controlled at 1:1. Each modification step is characterized by X-ray photoelectron spectroscopy and atomic force microscopy to confirm the successful grafting of each functional component. Fourier transform infrared spectroscopy shows that at 1720 cm⁻¹… -1 A stretching vibration peak of the ester carbonyl group appears at 2930 cm⁻¹. -1 The appearance of a CH stretching vibration peak in the adamantane skeleton confirms the successful introduction of the adamantane group.

[0035] During normal operation, the electron acceptor molecular cage is in a closed state, and the adamantyl groups and cyclodextrin groups at the ends of adjacent molecular brushes are freely distributed. When a partial discharge occurs at a certain point on the insulating interface, the high-energy electrons generated by the discharge collide with the electron acceptor molecular cage. If their energy exceeds the reduction potential threshold of the molecular cage, the cage undergoes oxidation, ring opening, and rupture, releasing the encapsulated non-conductive crosslinking agent. The non-conductive crosslinking agent rapidly wets the region, and under intermolecular forces, the adamantyl groups and cyclodextrin groups at the ends of adjacent molecular brushes undergo rapid physical crosslinking through host-guest inclusion, with a binding constant of 1 × 10⁻⁶. 4 M -1 After crosslinking, a transient insulating gel with a thickness of approximately 30 nanometers and a dielectric strength of approximately 80 kV / mm is formed in situ on the surface of the discharge point. This transient insulating gel can effectively increase the surface resistivity to 10. 12 Above Ω, it blocks the electron avalanche path and instantly suppresses discharge. When the discharge stops and the high energy disappears, under the condition of transformer operating temperature of about 80℃, the host-guest inclusion interaction of the adamantyl group and the cyclodextrin group gradually dissociates due to thermal motion. Its dissociation half-life is about 30 seconds. The transient insulating gel disappears automatically, and the molecular brush coating returns to its initial state, completing one self-healing cycle.

[0036] The states of the aforementioned molecular brush coating during normal operation and after responding to partial discharge correspond to the closure and rupture of the electron acceptor molecular cage, the encapsulation and release of the non-conductive crosslinking agent, and the formation and dissociation of reversible physical crosslinks between self-assembled motifs and recognition groups, respectively. Through these structural changes, the coating achieves complete functions of self-sensing, self-inhibition, and self-recovery.

[0037] Molecular brush state assessment model

[0038] Figure 2 The network structure of the molecular brush state evaluation model in this embodiment is shown. This model is a lightweight temporal anomaly detection model based on Transformer Encoder, and its specific structure includes: Input embedding layer: The 7-dimensional feature vector of each discharge pulse is linearly mapped to a 128-dimensional embedding vector. The 7-dimensional feature vector includes the pulse peak amplitude, pulse rise time, pulse half-peak width, decay time constant, time interval with the preceding pulse, discharge phase angle and sensor channel number.

[0039] Position coding layer: Sine position coding is used to add position information to the time series.

[0040] Transformer Encoder stack: Contains 2 layers, each with a 4-head self-attention mechanism and a feedforward network, with a hidden dimension of 128, a feedforward network dimension of 256, and a Dropout rate of 0.1.

[0041] Global average pooling layer: Performs time-dimensional average pooling on the encoder output sequence.

[0042] Output layer: A fully connected layer with Sigmoid activation that outputs a single anomaly score.

[0043] Model training process: A training set of at least 5000 samples is constructed using historical database data on normal operation, known continuous discharge events, and manually verified self-healing events. Binary cross-entropy is used as the loss function, and the Adam optimizer is employed for training. The initial learning rate is 0.001, the batch size is 64, and the training run is repeated for 100 epochs. During training, the discharge pulse sequence is segmented into fixed-length windows of 128 pulses and input into the model.

[0044] System Workflow

[0045] Figure 3 The complete workflow of the system is shown below, with specific steps described as follows: In step S1, when a minor defect appears in the inter-turn insulation paper of a certain phase high-voltage winding of the main transformer, causing an initial partial discharge, this discharge is accompanied by the emission of high-energy electrons. The electron acceptor molecular cage in the self-responsive insulating molecular brush coating at this location captures the high-energy electrons and breaks down, releasing a non-conductive crosslinking agent. The non-conductive crosslinking agent mediates the formation of reversible physical crosslinks between the self-assembled motifs and the recognition groups on the adjacent molecular brush coatings, forming a transient insulating gel in situ, which physically suppresses this partial discharge.

[0046] In step S2, the entire trigger-suppression process is completed within milliseconds. During this process, the ultra-high frequency partial discharge sensor and the high frequency current sensor capture a discharge pulse signal with unique characteristics; the leakage current sensor detects a momentary current drop; and the fiber optic vibration sensor captures a trace acoustic emission signal caused by the release of non-conductive crosslinking agent and gel formation. These multiphysics characteristic signals are synchronously acquired, timestamped, and converged to the edge computing gateway.

[0047] In step S3, the molecular brush state assessment model in the edge computing gateway performs real-time analysis on the aforementioned multi-source signals. The model extracts the width, decay time, and adjacent pulse interval features of each discharge pulse from the high-frequency partial discharge spectrum. The waveform of a successful self-healing event has the following characteristics: its discharge pulse width is narrow, approximately 15 microseconds, less than the first threshold of 50 microseconds; its decay time is short, approximately 20 microseconds, less than the second threshold of 30 microseconds; and there are no subsequent continuous discharge pulses, exhibiting an isolated burst-rapid decay waveform. In contrast, ordinary partial discharge pulses have a larger width, approximately 150 microseconds, decay slowly, and are often accompanied by continuous pulse clusters. The molecular brush state assessment model assigns an output anomaly score of 0.15 to the waveform of a successful self-healing event, which is lower than the judgment threshold of 0.5, thus identifying it as a successful self-healing event; while assigning an output anomaly score of 0.82 to the sequence of ordinary partial discharge pulses, which is higher than the threshold, classifying it as conventional interference or continuous discharge. The comparison of the waveform characteristics of the self-healing event and the ordinary partial discharge pulse clearly demonstrates the basis for the molecular brush state assessment model's identification and differentiation.

[0048] In step S4, the molecular brush state assessment model continuously statistically analyzes the frequency, discharge amplitude, and phase distribution of events identified as successful self-healing events. Simultaneously, the model utilizes the multi-channel arrival time difference of an ultra-high frequency partial discharge sensor, combined with an internal structural model of the transformer, to preliminarily estimate the spatial location of the discharge point. The trace amounts of hydrogen detected by the hydrogen sensor are also recorded and correlated with the self-healing events.

[0049] In step S5, when the frequency of successful self-healing events occurring within the same spatial area within 24 hours exceeds a preset threshold of 5 times, or when the discharge amplitude of a single event shows an increasing trend, the edge computing gateway determines that the coating consumption state of the insulation weakness is intensifying, and the insulation degradation trend is established. The edge computing gateway performs the calculation of the remaining effective life of the coating: based on the pre-calibrated total encapsulation amount of the effective molecular cages in the self-responsive insulating molecular brush coating, and the average consumption amount of a single self-healing event obtained by correlation analysis of the total number of historical self-healing events and the increase in characteristic gases in the oil within the corresponding operating cycle, the ratio of the total encapsulation amount to the average consumption amount is calculated to obtain the remaining effective life of the coating, which is output in the form of the remaining number of responsive events or equivalent operating time. The gateway then packages and sends the generated early warning information, including the fault phase, the estimated fault location, the estimated remaining effective life of the coating, and related characteristic data to the station-level monitoring backend.

[0050] In step S6, the station-level monitoring backend aggregates reports from multiple edge computing gateways and displays the insulation health status and coating consumption of all monitored equipment within the substation in the form of a heatmap on the interface. When multiple devices issue similar warnings, it indicates a potential systemic overvoltage risk. Simultaneously, the backend system automatically generates differentiated maintenance recommendations based on the severity of the warning and the estimated coating life: for locations with frequent self-healing, it recommends scheduling internal inspections or coating replenishment during the next low-wind period; for stable equipment, it automatically extends the maintenance cycle.

[0051] In addition, the edge computing gateway compresses and encrypts the original feature waveform data of each successfully self-healing event and uploads it asynchronously to the cloud model training database of the station-level monitoring backend. This data is used to periodically iterate and optimize the molecular brush state evaluation model and improve its generalization ability.

[0052] Example 2: Cucurbitaureus-Methylvioletin Host-Guest System This embodiment is basically the same as Embodiment 1, except that: The self-assembled motif is a methyl viologen group, and the recognition group is a cucurbituril group [7]. The binding constant between the two is 2 × 10⁻⁶. 5 M -1 The binding constant of adamantane and cyclodextrin in this embodiment is significantly higher than that in Example 1. Therefore, the molecular brush coating of this embodiment exhibits a faster crosslinking response speed (less than 1 millisecond) and higher gel mechanical strength, but the dissociation half-life is also correspondingly extended to approximately 120 seconds. This embodiment is particularly suitable for scenarios with weak but persistent discharge activity, such as the surface of basin insulators in gas-insulated switchgear. The transformer operating temperature in this embodiment is approximately 60°C.

[0053] The molecular brush coating using the cucurbituril-methyl viologen host-guest pair described above has a self-assembly motif of methyl viologen and a recognition group of cucurbituril in its molecular structure. It differs from the adamantane-cyclodextrin system in Example 1 in chemical composition, but is consistent in triggering mechanism and function. That is, after capturing high-energy electrons, it releases a non-conductive crosslinking agent, forms reversible physical crosslinking through host-guest inclusion, and generates transient insulating gel in situ.

[0054] Commonalities in molecular brush coating preparation methods:

[0055] In addition to the two specific embodiments described above, the self-responsive insulating molecular brush coating can also be implemented using the following alternatives: the core structure of the electron acceptor molecular cage can be replaced with a naphthalimide derivative or a phthalimide derivative. Taking a naphthalimide derivative as an example, when a n-butyl group is introduced onto its imide nitrogen atom and a cyano group is introduced at the 2,6 positions of the naphthalene ring, its redox potential relative to the saturated calomel electrode is approximately -1.6V; when a stronger electron-withdrawing group is introduced, its redox potential can be adjusted to -1.0V. The aforementioned redox potential can be continuously adjusted within the range of -1.0V to -2.5V relative to the saturated calomel electrode through the electron-pull effect of the substituents to adapt to the discharge energy characteristics of different device types. The non-conductive crosslinking agent can be replaced with low molecular weight silicone oil with a kinematic viscosity of 50cSt to 500cSt, or fluorosilicone oil with a dielectric strength greater than 35kV / 2.5mm, or liquid oligoisobutylene with a number-average molecular weight of 300Da to 1000Da. The base chain segment can be replaced with a dopamine derivative, which is anchored to the insulating substrate surface through coordination bonds formed by catechol groups; or a phosphate ester group, which is anchored through PO-surface bonding; or different types of silane coupling agents, including trimethoxysilane, triethoxysilane, or chlorosilane, to adapt to the surface chemical properties of different substrate materials such as insulating paper fibers, epoxy resins, and ceramics. The host-guest inclusion pair can also be other systems capable of reversible cross-linking through supramolecular interactions. The architecture of the molecular brush state assessment model can also be replaced with an LSTM-based autoencoder or a temporal convolutional network-based anomaly scoring model, with its input feature vector and training process remaining consistent with Example 1. These alternatives can all achieve the same self-healing event recognition and coating consumption state assessment functions as in Example 1.

[0056] Through the above description of multiple embodiments, this invention deeply integrates the self-healing mechanism in the field of materials with the edge computing intelligent analysis in the field of information, and constructs a closed loop of self-sensing, self-inhibition, self-evaluation and self-optimization for active safety management of insulation, which significantly improves the safety, reliability and maintenance efficiency of wind farm booster stations.

[0057] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A safety management system for wind farm booster stations based on self-responsive insulating molecular brushes and edge computing, characterized in that, include: Self-responsive insulating molecular brush coating is applied to the critical insulation interfaces of the substation equipment. A sensor array is used to monitor in real time the multi-physics characteristic signals generated by the self-responsive insulating molecular brush coating in response to partial discharge. An edge computing gateway, deployed locally at the booster station, communicates with the sensor group. The station-level monitoring backend is communicatively connected to the edge computing gateway. The self-responsive insulating molecular brush coating includes a base chain segment anchored to the surface of an insulating substrate, an electron acceptor molecular cage connected to the base chain segment, and a self-assembled motif located at the end of the molecular brush coating. The electron acceptor molecular cage is encapsulated with a non-conductive crosslinking agent. The self-responsive insulating molecular brush coating is configured such that after capturing high-energy electrons generated by partial discharge, the electron acceptor molecular cage breaks and releases the non-conductive crosslinking agent. The non-conductive crosslinking agent mediates the formation of a reversible physical crosslink between the self-assembled motif and the recognition group on the adjacent molecular brush coating, so as to form a transient insulating gel in situ at the discharge point to suppress discharge. The edge computing gateway is configured to receive and process the multiphysics feature signals from the sensor group, run the molecular brush state assessment model, extract the width, decay time and adjacent pulse interval features of the discharge pulse from the high-frequency partial discharge spectrum, and when the width of a single discharge pulse is less than a first threshold, the decay time is less than a second threshold and there are no subsequent continuous discharge pulses, the discharge pulse is identified as a successful self-healing event, the coating consumption status is assessed, and the analysis results are sent to the station-level monitoring backend. The electron acceptor molecular cage is a redox-active cage formed from a triphenylamine derivative, and its redox potential is between -1.5V and -2.5V relative to the saturated calomel electrode; the self-assembled motif and the recognition group are host-guest inclusion pairs that can form reversibly crosslinked structures through supramolecular interactions. The host-guest inclusion pair is an adamantyl group and a cyclodextrin group, or a cucurbituril group and a methyl viologen group; The molecular brush state evaluation model is a lightweight temporal anomaly detection model, which is further configured as follows: The feature vector of each discharge pulse is extracted from the multi-physics field feature signal. The feature vector includes at least the pulse peak amplitude, pulse rise time, pulse half-peak width, decay time constant, time interval with the preceding pulse, discharge phase angle and sensor channel number. Based on the feature vector, a discharge pulse with a sudden-rapid decay characteristic is identified, and the discharge pulse is identified as a successful self-healing event; The frequency, amplitude, and phase distribution of the successful self-healing events within a specific time window were statistically analyzed. When the frequency or cumulative energy exceeds a preset threshold, an insulation degradation warning is issued; The remaining effective lifespan of the coating is calculated and output based on the total number of successful self-healing events or the characteristic gas trend.

2. The system according to claim 1, characterized in that, The sensor group includes at least a leakage current sensor, a partial discharge sensor, and a dissolved gas sensor in oil; the partial discharge sensor is an ultra-high frequency partial discharge sensor or a high frequency current sensor; the dissolved gas sensor in oil is a hydrogen sensor or a methane sensor.

3. The system according to claim 1, characterized in that, The lightweight temporal anomaly detection model is one of the following: a single-class anomaly detection network based on Transformer Encoder, an autoencoder based on Long Short-Term Memory network, or an anomaly scoring model based on temporal convolutional network.

4. The system according to claim 1, characterized in that, The sensor group also includes an optical fiber vibration sensor or an acoustic emission sensor, which is attached to the device housing or near the insulating paper to capture the sound waves or vibration signals generated during the release of the non-conductive crosslinking agent and the formation of the transient insulating gel.

5. The system according to claim 1, characterized in that, include: In response to partial discharge, the electron acceptor molecular cage in the self-responsive insulating molecular brush coating captures high-energy electrons and then breaks, releasing the encapsulated non-conductive crosslinking agent. The non-conductive crosslinking agent mediates the formation of reversible physical crosslinking between the self-assembled motif and the recognition group on the adjacent molecular brush coating, forming a transient insulating gel in situ at the discharge point to suppress the discharge. The self-responsive insulating molecular brush coating is applied to the key insulating interface of the booster station equipment. The sensor array captures multiphysics characteristic signals generated by the release and gel formation process in real time, and sends the multiphysics characteristic signals to the edge computing gateway, which is deployed locally at the booster station; The multiphysics characteristic signal is processed by the molecular brush state assessment model in the edge computing gateway. The width, decay time and adjacent pulse interval features of the discharge pulse are extracted from the high-frequency partial discharge spectrum. When the width of a single discharge pulse is less than a first threshold, the decay time is less than a second threshold and there are no continuous discharge pulses after it, the discharge pulse is identified as a successful self-healing event. The frequency and energy trend of the self-healing event are statistically analyzed to assess the coating consumption state and insulation degradation trend. When the frequency or cumulative energy of the self-healing event exceeds a preset threshold, the edge computing gateway generates an early warning message and sends the early warning message and evaluation results to the station-level monitoring backend. The station-level monitoring backend integrates data from multiple devices to generate maintenance recommendations.

6. The system according to claim 5, characterized in that, The first threshold ranges from 10 to 100 microseconds, and the second threshold ranges from 10 to 80 microseconds; The assessment of the coating consumption status includes: obtaining the total amount of effective molecular cages encapsulated in the self-responsive insulating molecular brush coating; obtaining the average consumption of a single self-healing event by performing correlation analysis on the total number of historical successful self-healing events and the characteristic gas increment of the corresponding period; and calculating the remaining effective lifetime of the coating based on the ratio of the total encapsulation amount to the average consumption amount.

Images