Fault early warning system for wind power booster station

Abstract

According to the fault early warning system for the wind power booster station provided by the invention, the limitation of traditional single-point monitoring is broken through by fusing the infrared ultrasonic technology and the ultrahigh frequency technology through the non-contact sensor, so that the early fault detection rate is greatly improved; the composite heat dissipation system suppresses the temperature rise of a transformer hot spot based on a dynamic heat management mechanism of a phase change material and a heat pipe, and the energy-saving efficiency is improved; the edge calculation early warning center converts temperature, partial discharge, strain and other multi-source heterogeneous data into a unified health degree index through a weighted fusion algorithm, and in combination with an equipment topology correlation analysis model, the positioning accuracy of a fault source is greatly improved, and the early warning response time delay is greatly reduced.

Description

Technical Field

[0001] This invention relates to the field of fault detection technology for wind power booster stations, and particularly to a fault early warning system for wind power booster stations. Background Technology

[0002] As a key node in the new energy power system, the reliability of wind power substations directly affects the stability of the power grid. With the global installed capacity of wind power exceeding 1200GW, the monitoring technology of substation equipment has undergone three generations of evolution: from single-parameter threshold alarm (early infrared temperature measurement), to multi-sensor independent monitoring (parallel partial discharge detection and temperature monitoring), to basic data fusion (SCADA system integration). The application fields cover onshore / offshore wind power, flexible DC transmission and other scenarios.

[0003] Existing traditional monitoring systems can only achieve simple superposition alarms of discrete parameters, resulting in a false alarm rate as high as 37% (industry average). Especially when transformer winding overheating and cable joint partial discharge occur simultaneously, it is difficult to distinguish the fault source from the conduction anomaly, resulting in a waste of maintenance resources. Summary of the Invention

[0004] In view of this, the present invention provides a wind power booster station fault early warning system to solve the technical defects existing in the prior art.

[0005] Specifically, the present invention provides a fault early warning system for wind power booster stations, including interconnected multi-dimensional monitoring modules, a composite heat dissipation system, and an edge computing early warning center; The multi-dimensional monitoring module collects temperature, partial discharge signals, and strain data from the booster station equipment using non-contact sensors; The composite heat dissipation system receives temperature data from a multi-dimensional monitoring module and actively dissipates heat through the synergistic effect of phase change materials and heat pipes. The edge computing early warning center integrates multi-source data from multi-dimensional monitoring modules to generate device health assessments and fault warnings, and controls the operation mode of the composite heat dissipation system.

[0006] In some implementations, the multi-dimensional monitoring module includes an infrared ultrasonic composite sensor, a partial discharge ultra-high frequency antenna, and a fiber optic strain sensor. Infrared and ultrasonic composite sensors scan the surface of transformers and switchgear to generate surface temperature data, while simultaneously acquiring internal discharge ultrasonic signals. A partial discharge ultra-high frequency antenna scans the busbar compartment space with electromagnetic waves to generate partial discharge signal amplitude; Fiber optic strain sensors measure the strain of high-voltage cables to generate micro-strain data.

[0007] In some implementations, the composite heat dissipation system includes a transformer heat dissipation modification unit, an intelligent ventilation network, and a photovoltaic auxiliary unit; The transformer heat dissipation modification unit transfers heat from the transformer oil tank to the phase change material box via heat pipes. The intelligent ventilation network automatically adjusts the opening and closing of louvers and the direction of fan rotation based on the monitored temperature; The photovoltaic auxiliary unit supplies power to the heat dissipation system.

[0008] In some implementations, the edge computing early warning center includes a data fusion unit, a fault tracing unit, and a display unit; The data fusion unit performs weighted fusion processing on temperature data, ultrasonic signals, and partial discharge signals to generate a health index; The fault tracing unit performs correlation analysis on the health index based on the equipment topology relationship to generate related equipment influencing factors; The display unit converts the analysis results into visualized data.

[0009] In some implementations, the data fusion unit performs weighted fusion processing on temperature data, ultrasonic signals, and partial discharge signals to generate a health index, including: The temperature data collected by the infrared and ultrasonic composite sensor is normalized to generate a standard temperature difference value. The standard temperature difference value is then weighted and squared to generate a temperature influence factor. Logarithmic transformation of the ultrasonic signal intensity is performed to generate signal feature values, and cube root operation is performed on the signal feature values ​​to generate the ultrasonic influence factor. The amplitude of the raw signal acquired by the partial discharge UHF antenna is extracted to generate partial discharge characteristic quantities. The partial discharge characteristic quantities are then subjected to exponential weighted averaging to generate the partial discharge influence factor.

[0010] The health index is calculated based on the temperature influence factor, ultrasound influence factor, and partial radiation influence factor, as well as the preset first calculation formula.

[0011] In some implementations, the first calculation formula includes: Wherein, H is the health index, which is generated by the data fusion unit; The temperature difference value at the i-th temperature monitoring point is derived from the temperature scan result of the infrared-ultrasonic composite sensor. The intensity of the j-th ultrasonic signal comes from the ultrasonic acquisition result of the infrared ultrasonic composite sensor; The amplitude of the k-th partial discharge signal is derived from the electromagnetic wave scanning result of the partial discharge UHF antenna. , β and γ are preset weight coefficients, obtained through training with historical fault data; The total number of temperature monitoring points, This represents the total number of ultrasound monitoring points. This represents the total number of local emission monitoring points.

[0012] In some implementations, the step of generating associated device influence factors by performing correlation analysis on the health index based on device topology relationships includes: Collect electrical connection data between substation equipment to generate a physical topology map reflecting the physical connection status between equipment nodes. The connection data includes the number of connection paths between any two equipment nodes and at least one switch information in each connection path. The switch information indicates whether the switch is closed or closed. When all switches on the connection path between two equipment nodes are closed, the path validity coefficient corresponding to that path is 1; otherwise, it is 0. Construct a correlation matrix between device nodes, wherein the correlation matrix is ​​a square matrix, the correlation matrix includes multiple matrix element values, each matrix element value is calculated based on the number of common paths and common effective coefficient of at least one connection path between two device nodes, the path connectivity of each of the two devices, and a preset third calculation formula, and the common effective coefficient is 1 when the effective coefficient of any connection path between two device nodes is 1; Based on the matrix element values ​​in the correlation matrix, the topological correlation degree corresponding to each matrix element value is calculated according to the preset weights corresponding to each device type and the preset fourth calculation formula. The influence factor of the associated equipment is calculated based on the topological correlation degree.

[0013] In some embodiments, the third calculation formula includes: in, This represents the matrix element values ​​corresponding to device node m and device node q. This represents the number of common paths between device node m and device node q. This represents the degree of device node m. This represents the degree of device node q. represents the common valid coefficient between device node i and device node j, where m and q are natural numbers.

[0014] In some implementations, the fourth calculation formula includes: in, This represents the topological affinity between device node m and device node q, where Q represents the total number of devices. This represents the preset coefficient corresponding to the device type of device node q.

[0015] In some implementations, the second formula for calculating the influence factor of the associated equipment includes: Among them: Let m be the influence factor of the m-th associated device. Let be the partial derivative of the health index with respect to the l-th parameter. The value of the l-th monitoring parameter comes from the raw data of the multi-dimensional monitoring module. The standard deviation of the l-th parameter is obtained through statistical analysis of historical data. Let m be the topological affinity between device node m and device node q. η and η are preset adjustment coefficients, which are manually adjusted according to the equipment type. L is the number of monitoring parameters, and Q is the total number of equipment.

[0016] At least one embodiment of the present invention overcomes the limitations of traditional single-point monitoring by integrating infrared ultrasound and ultra-high frequency technology through non-contact sensors, thereby significantly improving the early fault detection rate. The composite heat dissipation system, based on the dynamic thermal management mechanism of phase change materials and heat pipes, suppresses the temperature rise of transformer hotspots and increases energy efficiency. The edge computing early warning center transforms multi-source heterogeneous data such as temperature, partial discharge, and strain into a unified health index through a weighted fusion algorithm. Combined with the equipment topology correlation analysis model, it significantly improves the accuracy of fault source location and greatly reduces the early warning response delay. Attached Figure Description

[0017] Figure 1 This is a structural block diagram of a wind power booster station fault early warning system provided by the present invention. Detailed Implementation

[0018] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0019] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the one or more embodiments of this specification. The singular forms “a” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items. The modifications “a” and “a plurality” as used in this disclosure are illustrative and not restrictive, and those skilled in the art will understand that they should be understood as “one or more” unless the context clearly indicates otherwise.

[0020] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0021] Traditional wind power substation monitoring technology has long been limited by a crude approach of single-parameter threshold alarms. When faced with complex faults such as transformer overheating and cable partial discharge, it cannot identify multi-physics coupling characteristics and lacks the ability to analyze equipment topology correlations. Analysis of three consecutive years of fault logs from a wind farm revealed that most false alarms stemmed from isolated judgments of temperature and partial discharge signals, and at least one-third of the delayed processing was due to lag in the cooling system response. Therefore, this paper proposes a solution that integrates three key technology modules: infrared ultrasonic composite sensing, phase change material heat pipe synergistic cooling, and edge computing data fusion. First, non-contact sensors simultaneously capture multi-dimensional signals of temperature, ultrasound, and strain. Then, a weighted fusion algorithm is developed to transform heterogeneous data into a health index. Finally, a fault propagation model is established based on equipment topology relationships.

[0022] Specifically, see Figure 1 , Figure 1 The diagram illustrates a structural block diagram of a wind power booster station fault early warning system according to some embodiments of this specification. The system includes interconnected multi-dimensional monitoring modules, a composite heat dissipation system, and an edge computing early warning hub. The multi-dimensional monitoring modules collect temperature, partial discharge signals, and strain data of the booster station equipment using non-contact sensors. The composite heat dissipation system receives the temperature data from the multi-dimensional monitoring modules and actively dissipates heat through the synergistic effect of phase change materials and heat pipes. The edge computing early warning hub integrates multi-source data from the multi-dimensional monitoring modules to generate equipment health assessments and fault early warnings, and controls the operating mode of the composite heat dissipation system.

[0023] Multi-dimensional monitoring modules refer to monitoring units that integrate multiple sensing technologies. They deploy infrared sensors to capture the surface temperature field distribution of equipment, combine this with ultrasonic probes to capture mechanical vibration waves generated by partial discharge, and simultaneously use strain gauges to measure structural deformation, enabling a three-dimensional perception of the equipment's operating status. Non-contact sensors refer to detection devices that do not require physical contact. They use infrared thermal imagers to scan for temperature rise hotspots at electrical joints and capture partial discharge electromagnetic wave signals through a UHF antenna array, avoiding the damage to the equipment's insulation performance caused by traditional contact detection. Partial discharge signals refer to the discharge pulse characteristics generated by insulation degradation. UHF sensors capture nanosecond-level electromagnetic wave pulse sequences, and ultrasonic signal time-frequency analysis algorithms extract discharge type fingerprints, used to identify typical defects such as inter-turn discharge in windings or surface creepage on bushings. Strain data refers to the deformation caused by mechanical stress. Fiber optic grating sensors monitor changes in axial pressure in transformer windings, and vibration accelerometers collect the dynamic response of the housing, providing early warning of cascading faults caused by loose fasteners or structural deformation.

[0024] A composite heat dissipation system can refer to an active thermal management device. When the temperature exceeds a threshold, it activates the phase change material (PCM) heat storage unit to absorb the transient heat load, and simultaneously starts the heat pipe network to conduct heat to the heat dissipation fin assembly, enabling rapid heat transfer at the kilowatt level. PCM refers to a material with latent heat storage characteristics. Organic PCM materials with a melting point of 60 degrees Celsius are encapsulated in the transformer tank wall. When the equipment overheats, they absorb the latent heat of melting, slowing the rate of temperature rise and providing a buffer time for active heat dissipation. Heat pipes can refer to high-efficiency heat transfer elements. A mesh-like heat transfer channel is formed by sintered copper capillary heat pipes, transferring the heat stored in the PCM to a distant air-cooled radiator, solving the problem of uneven heat flux density in traditional oil-immersed heat dissipation.

[0025] Edge computing early warning hub can refer to a localized data processing unit that uses a time-series database to store multi-source monitoring data, runs a weighted fusion algorithm to generate a device health index, and locates the coordinates of the fault source based on a topological association model, enabling multi-parameter collaborative diagnosis at the level of two hundred milliseconds.

[0026] The present invention will be further described below through a detailed embodiment: A coastal wind farm's booster station has deployed the wind power booster station fault early warning system described in this invention. The core of this system comprises a closed-loop protection system consisting of a multi-dimensional monitoring module, a composite heat dissipation system, and an edge computing early warning hub. The multi-dimensional monitoring module adopts a distributed architecture. Its non-contact sensor array includes three functional units: an infrared thermal imager scans the transformer bushing joint temperature field at a sampling rate of 5 frames per second, covering a measurement range from -40°C to 300°C with an accuracy of ±1°C; an ultra-high frequency sensor captures partial discharge signals in the 300 MHz to 3 GHz frequency band using a wideband antenna, and, in conjunction with an ultrasonic probe, receives mechanical vibration waves from 40 kHz to 200 kHz to achieve discharge type spectrum identification; and a fiber optic grating strain sensor monitors cable tray deformation at a density of 50 measuring points per meter, with a range of ±5000 micro-strain. These sensing units establish a real-time data channel with the edge computing early warning hub via an industrial Ethernet network.

[0027] The composite heat dissipation system employs a coupled design of phase change material (PCM) and a heat pipe network. The PCM is an organic mixture with a melting point of 65 degrees Celsius, encapsulated in a 20mm thick aluminum alloy interlayer, covering over 80% of the transformer tank surface area. When an infrared thermal imager detects a local temperature exceeding 70 degrees Celsius, the PCM begins to absorb latent heat of fusion, simultaneously activating the radial heat pipe array. These heat pipes are constructed with a copper powder sintered capillary core structure, with a single pipe capable of transferring 300 watts of heat to the distal heat dissipation fins. The heat dissipation fins are equipped with continuously variable speed fans that automatically adjust their speed based on temperature gradient data from the edge computing early warning center, ensuring that the transformer hotspot temperature rise is consistently kept below 15 Kelvin.

[0028] The edge computing early warning hub is deployed in the station's cabinets. Its data processing flow includes a three-layer architecture: the data acquisition layer synchronizes multi-source signals through a timestamp alignment algorithm to establish a spatiotemporal dataset containing a temperature matrix, partial discharge pulse sequence, and strain curve; the feature fusion layer runs a weighted dynamic entropy algorithm to normalize the temperature gradient variance, partial discharge repetition rate, and strain fluctuation coefficient into a health index of 0 to 100; the decision output layer, based on the device topology map, triggers a three-level early warning mechanism when the health index is below 70—the primary early warning (index 60-70) only records abnormal logs, the intermediate early warning (index 40-60) activates the composite heat dissipation system enhancement mode, and the advanced early warning (index below 40) links the station's circuit breakers to perform protective isolation.

[0029] During system operation, when the UHF sensor detects a typical suspended discharge spectrum accompanied by a local temperature rise rate exceeding 2 degrees Celsius per minute, the edge computing early warning center activates the fault location subroutine. This program compares the time delay difference between adjacent sensor signals and uses triangulation to pinpoint the fault source within a 2-meter accuracy range. Simultaneously, it calls upon the historical database to match similar fault cases, generating a diagnostic suggestion including "87% probability of discharge in the B-phase bushing equalizing ring." Test data shows that this system reduces the fault identification window of traditional monitoring schemes from 72 hours to 45 minutes, and decreases the false alarm rate by 62 percentage points.

[0030] The beneficial effects of one of the embodiments in this specification include at least the following: by fusing infrared ultrasound and ultra-high frequency technology with non-contact sensors, the limitations of traditional single-point monitoring are overcome, and the early fault detection rate is greatly improved; the composite heat dissipation system, based on the dynamic thermal management mechanism of phase change materials and heat pipes, suppresses the temperature rise of transformer hotspots and increases energy efficiency; the edge computing early warning center transforms multi-source heterogeneous data such as temperature, partial discharge, and strain into a unified health index through a weighted fusion algorithm, and combined with the equipment topology correlation analysis model, greatly improves the accuracy of fault source location and significantly reduces early warning response latency.

[0031] In some implementations, the multi-dimensional monitoring module includes an infrared ultrasonic composite sensor, a partial discharge ultra-high frequency antenna, and a fiber optic strain sensor; the infrared ultrasonic composite sensor performs temperature scanning on the surface of the transformer and switchgear to generate equipment surface temperature data, while simultaneously acquiring internal discharge ultrasonic signals; the partial discharge ultra-high frequency antenna performs electromagnetic wave scanning on the busbar compartment space to generate partial discharge signal amplitude; and the fiber optic strain sensor performs strain measurement on the high-voltage cable to generate micro-strain data.

[0032] Infrared-ultrasonic composite sensors refer to integrated probes that combine temperature and acoustic wave detection functions. They simultaneously emit infrared beams to scan the surface thermal radiation intensity of equipment while receiving ultrasonic signals generated by discharges in the 40kHz-200kHz range, enabling parallel diagnosis of equipment overheating and insulation degradation. Partial discharge ultra-high frequency antennas refer to electromagnetic wave detection devices that capture electromagnetic pulses radiated by partial discharges in the 300MHz-3GHz frequency band. Using time-domain reflectometry algorithms, they locate the spatial position of the discharge point and can identify hidden defects such as creepage on bushing surfaces or air gap discharges inside insulators. Fiber optic strain sensors refer to deformation measurement units based on the grating principle. They embed Bragg gratings into the shielding layer of high-voltage cables and invert the mechanical strain distribution through wavelength offset, providing early warning of structural risks caused by loose cable joints or deformed supports.

[0033] Equipment surface temperature data refers to a set of parameters reflecting the thermal state of external insulation. A temperature matrix with a resolution of 640×480 pixels is generated using an infrared thermal imager, and an isotherm model of the equipment surface is established using a 3D reconstruction algorithm, enabling precise identification of the geometric features of overheated areas. Internal discharge ultrasonic signals refer to mechanical vibration waves excited by insulation defects. The time-frequency characteristics of the acoustic emission signals are captured using piezoelectric ceramic sensors, and pattern recognition algorithms distinguish between corona discharge and spark discharge types, allowing for a quantitative assessment of insulation degradation. Busbar compartment space refers to the enclosed area containing the power busbar. An ultra-high frequency antenna array is arranged to form an electromagnetic wave monitoring network. The spatial coordinates of the discharge source are calculated based on the signal arrival time difference, achieving meter-level accuracy in fault location. Partial discharge signal amplitude refers to a quantitative indicator of discharge intensity. The pulse voltage amplitude is recorded using a peak-hold circuit, and oscillation attenuation characteristics are extracted using wavelet transform, used to establish a discharge development trend prediction model. Micro-strain data refers to the quantitative expression of minute deformations. The strain of the cable shielding layer is measured at a resolution of 1με. By establishing a correlation model between strain gradient and current carrying capacity, joint overheating faults can be predicted up to 3 hours in advance.

[0034] Through a multi-physical quantity collaborative sensing mechanism, a three-dimensional monitoring network covering electrical, thermal, and mechanical conditions was constructed. Infrared and ultrasonic fusion detection overcomes the limitations of single-parameter monitoring, ultra-high frequency electromagnetic scanning enables non-destructive diagnosis of insulation defects, and fiber optic strain measurement provides unprecedented visualization capabilities of the mechanical condition of cable systems. The three technologies work together to form a precise profile of the full-dimensional health status of the substation equipment.

[0035] In some implementations, the composite heat dissipation system includes a transformer heat dissipation modification unit, an intelligent ventilation network, and a photovoltaic auxiliary unit; the transformer heat dissipation modification unit conducts heat from the transformer tank to the phase change material box through heat pipes; the intelligent ventilation network automatically adjusts the opening and closing of louvers and the direction of the fan according to the monitored temperature; and the photovoltaic auxiliary unit supplies power to the heat dissipation system.

[0036] The transformer heat dissipation retrofit unit can refer to a transformer thermal management optimization device, which embeds a sintered copper heat pipe array into the side wall of the transformer tank to control the oil temperature below 65 degrees Celsius through heat conduction, reducing energy loss by 15% compared to traditional oil-immersed cooling. The intelligent ventilation network can refer to an adaptive airflow regulation system, which monitors the heat distribution between equipment in real time through a temperature sensor array, and uses a stepper motor to drive the opening and closing angle of louvers and adjust the speed of axial flow fans, achieving balanced temperature control within the substation. The photovoltaic auxiliary unit can refer to a renewable energy power supply module, using a monocrystalline silicon photovoltaic panel array arranged on the building roof, outputting 48V DC power to the cooling fans through an MPPT controller, reducing the system's dependence on the power grid. The phase change material box can refer to a thermal storage and temperature control container, filled with a paraffin-based composite phase change material with a melting point of 60 degrees Celsius. When the heat pipe conducts heat and melts the material, it absorbs 120 kJ / kg of latent heat, used to smooth temperature spikes caused by transformer load fluctuations. The fan direction control function refers to the airflow direction adjustment function. It adopts a dual-rotor axial flow fan design and intelligently switches between exhaust and supply modes based on temperature gradient data, so that the airflow always flows in the direction of the most severely heated area of ​​the equipment.

[0037] A highly efficient and energy-saving temperature control system was constructed through a multi-level thermal regulation mechanism. The transformer modification unit solved the heat dissipation bottleneck of the core equipment, the intelligent ventilation network achieved dynamic balance of the environmental thermal field, and photovoltaic power supply gave the system the ability to operate off-grid. The three work together to form a comprehensive solution that takes into account reliability, economy and environmental protection.

[0038] In some implementations, the edge computing early warning hub includes a data fusion unit, a fault tracing unit, and a display unit; the data fusion unit performs weighted fusion processing on temperature data, ultrasonic signals, and partial discharge signals to generate a health index; the fault tracing unit performs correlation analysis on the health index based on the equipment topology to generate related equipment influence factors; and the display unit converts the analysis results into visualized data.

[0039] The health index refers to a comprehensive evaluation parameter for the degree of equipment deterioration. It uses a linear weighted fusion of features such as temperature gradient variance, ultrasonic signal amplitude, and partial discharge pulse repetition rate to visually reflect the remaining lifespan of the equipment on a percentage basis. Equipment topology refers to the electrical connection logic diagram. It constructs a node connection matrix including transformers, circuit breakers, and busbars, and establishes a model of the mutual influence of equipment states based on impedance parameters, which can predict systemic risks caused by local faults. The related equipment influence factor refers to a fault propagation intensity index. Based on the electrical distance and coupling degree between equipment, it calculates the weight of the impact of abnormal equipment on the state of surrounding equipment, used to formulate differentiated maintenance priority strategies.

[0040] The hierarchical processing architecture enables intelligent transformation of monitoring data, data fusion eliminates the limitations of single sensors, fault tracing reveals hidden associated risks, and visualization significantly lowers the technical decision-making threshold, forming a complete cognitive chain from data collection to operation and maintenance decisions.

[0041] In some implementations, the data fusion unit performs weighted fusion processing on temperature data, ultrasonic signals, and partial discharge signals to generate a health index. This includes: normalizing the temperature data collected by the infrared-ultrasonic composite sensor to generate a standard temperature difference, and performing a weighted sum of squares on the standard temperature difference to generate a temperature influence factor; performing a logarithmic transformation on the ultrasonic signal intensity to generate signal feature values, and performing a cube root operation on the signal feature values ​​to generate an ultrasonic influence factor; and extracting the amplitude of the original signal collected by the partial discharge UHF antenna to generate partial discharge feature quantities, and performing an exponentially weighted average operation on the partial discharge feature quantities to generate a partial discharge influence factor. The health index is then calculated based on the temperature influence factor, ultrasonic influence factor, partial discharge influence factor, and a preset first calculation formula.

[0042] The standard temperature difference refers to the standardized processing result of temperature data. It is used to subtract the reference temperature value from the original temperature reading and then divide by the measurement range. Z-score standardization eliminates interference from ambient temperature fluctuations, accurately reflecting the actual temperature rise of the equipment. The temperature influence factor refers to the contribution of temperature to health status. It uses a weighted sum of squares of the standard temperature difference with a weighting coefficient of 0.3-0.5. By highlighting the contribution weight of high-temperature areas, it enhances the early warning sensitivity of overheating faults. The signal characteristic value refers to the quantitative characterization parameter of the ultrasonic signal. Logarithmic transformation converts the original sound pressure level to decibels. By suppressing background noise interference, it can extract the energy in the 10-50kHz characteristic frequency band reflecting insulation defects. The ultrasonic influence factor refers to the contribution of ultrasound to health status. It uses a cube root operation on the signal characteristic value to compress the dynamic range. Combined with a weighting coefficient of 0.2-0.4, it can balance the signal intensity differences of different discharge types. Partial discharge characteristics refer to the quantification parameters of the discharge pulse. Rise time, repetition frequency, and amplitude envelope features are extracted from ultra-high frequency signals. Through joint time-frequency analysis, surface discharge and internal discharge modes can be distinguished. Partial discharge influence factors refer to the contribution of partial discharge to health status. Historical data is processed using an exponentially weighted average algorithm, assigning higher weight to recent discharge signals, with a coefficient of 0.3-0.5 reflecting the discharge development trend.

[0043] By employing a hierarchical feature extraction and dynamic weighting strategy, efficient integration of multi-source heterogeneous monitoring data was achieved. Standardization of temperature differences eliminated environmental interference, logarithmic transformation of ultrasonic signals enhanced feature recognition, and time-frequency analysis of partial discharge characteristics improved pattern discrimination capabilities. Finally, a health index comprehensively reflecting the equipment status was generated through scientific weighting, providing a reliable basis for intelligent diagnosis.

[0044] In some implementations, the first calculation formula includes: Wherein, H is the health index, which is generated by the data fusion unit; The temperature difference value at the i-th temperature monitoring point is derived from the temperature scan result of the infrared-ultrasonic composite sensor. The intensity of the j-th ultrasonic signal comes from the ultrasonic acquisition result of the infrared ultrasonic composite sensor; The amplitude of the k-th partial discharge signal is derived from the electromagnetic wave scanning result of the partial discharge UHF antenna. , β and γ are preset weight coefficients, obtained through training with historical fault data; The total number of temperature monitoring points, This represents the total number of ultrasound monitoring points. This represents the total number of local emission monitoring points.

[0045] Temperature difference refers to the difference between the equipment surface temperature and the ambient temperature. It is used to subtract the reference ambient temperature value from the raw temperature data collected by the infrared thermal imager. By eliminating the influence of seasonal temperature differences, it can accurately reflect the actual temperature rise status of the equipment during operation. Ultrasonic signal intensity refers to the sound wave energy value generated by insulation defects. The raw signal from the piezoelectric sensor is processed with a 40kHz bandpass filter, and the discharge intensity is quantified by calculating the root mean square value, which can identify partial discharge activity in the range of 0-100dB. Partial discharge signal amplitude refers to the voltage peak value of the electromagnetic pulse. A UHF antenna is used to receive signals in the 300MHz-3GHz frequency band, and the highest point of the pulse waveform is extracted through an envelope detection circuit, reflecting the severity of the discharge with millivolt-level accuracy. Weighting coefficients refer to the adjustment factors of multi-parameter fusion. By analyzing the contribution of each parameter in historical fault samples through machine learning, and using the gradient descent method to optimize the values ​​of wT, α, β, and γ, the health index can maintain a linear correlation with the actual degree of equipment degradation. Temperature monitoring points can refer to critical temperature measurement locations on equipment. Based on thermal simulation analysis, easily overheated areas such as transformer bushings, windings, and cores are identified. Miniature infrared sensor arrays are deployed to achieve millimeter-level resolution monitoring of the surface temperature field with a 5cm spacing. Ultrasonic monitoring points can refer to acoustically sensitive areas. Waterproof piezoelectric sensors are installed at weak points in the equipment insulation. Three detection points are arranged according to the triangulation principle, and the discharge location can be accurately located within ±2cm using the time-difference method. Partial discharge monitoring points can refer to electromagnetic wave acquisition nodes. A wideband Archimedes spiral antenna array is arranged around the high-voltage conductor. The phase comparison method is used for spatial positioning of the radiation source, and a discharge quantity as small as 10pC can be detected.

[0046] A multi-physical quantity coupling algorithm was used to achieve accurate assessment of equipment status. The sum of squares of temperature differences highlights the contribution of hot spots, logarithmic processing of ultrasonic signals enhances the ability to identify weak discharges, and exponential weighting of partial discharge signals reflects the cumulative effect. Finally, a comprehensive index was integrated through dynamic weighting coefficients to provide a quantitative basis for predictive maintenance.

[0047] In some implementations, the step of generating associated equipment influence factors by performing correlation analysis on the health index based on equipment topology relationships includes: collecting electrical connection relationship data between substation equipment, generating a physical topology map reflecting the physical connection status between equipment nodes, wherein the connection relationship data includes the number of connection paths between any two equipment nodes, and at least one switch information in each connection path, the switch information indicating closed or closed, and when all switches on the connection path between two equipment nodes are in the closed state, the path effectiveness coefficient corresponding to that path is 1, otherwise it is 0; constructing a correlation degree matrix between equipment nodes, wherein the correlation degree matrix is ​​a square matrix, the correlation degree matrix includes multiple matrix element values, each matrix element value is calculated based on the number of common paths and common effectiveness coefficients of at least one connection path between two equipment nodes, the path connectivity of each of the two devices, and a preset third calculation formula, and when the effectiveness coefficient of any connection path between two equipment nodes is 1, the common effectiveness coefficient is 1; calculating the topology correlation degree corresponding to each matrix element value based on the matrix element values ​​in the correlation degree matrix, according to the preset weights corresponding to each equipment type, and a preset fourth calculation formula; and calculating the associated equipment influence factors based on the topology correlation degrees.

[0048] Electrical connection relationship data refers to the actual wiring status information between devices. It is used to obtain real-time status signals of devices such as circuit breakers and disconnect switches from the SCADA system. It is parsed into a node-edge network structure through a graphical topology algorithm, which can accurately reflect the current operation mode of the main electrical wiring.

[0049] A physical topology map refers to a visual model of the power grid structure. Using Graphviz, device nodes and connection edges are rendered as a directed graph, with different colors used to label equipment types such as transformers and buses, visually demonstrating the physical paths of fault propagation. The number of paths refers to the total number of electrical pathways between devices. A depth-first search algorithm traverses the topology graph, counting all unique connection combinations between nodes. Considering special wiring methods such as double busbar segmentation, it can identify backup paths under N-1 operating conditions. Switch information refers to the status variables of circuit breakers / disconnectors. The open / closed positions are parsed from GOOSE messages from intelligent terminals, and the switch status is represented by binary encoding. Status estimation verification can eliminate erroneous signal interference. The path effectiveness coefficient refers to the availability index of connection paths. It is set to 1 when all switches on the path are closed and 0 when any switch is open. Combined with real-time updates based on topology analysis, it dynamically reflects changes in network connectivity.

[0050] The correlation matrix can be considered a quantified table of the influence relationships between devices. Constructing an n×n symmetric matrix, the number of shortest paths between nodes is calculated using the Floyd-Warshall algorithm. Combined with normalization of device impedance parameters, it can characterize the fault coupling strength. Matrix element values ​​can be considered quantified indicators of device correlation. A third calculation formula is used to perform a weighted geometric average on parameters such as the number of common paths and the effectiveness coefficient, introducing a device load rate correction factor to dynamically reflect the impact of changes in operating modes. The number of common paths can be considered the number of parallel channels between devices. Searching for the set of non-intersecting paths between two nodes in the topology graph, the number of redundant connections is determined using the maximum flow algorithm, which can assess the ease of fault isolation. The common effectiveness coefficient can be considered a comprehensive indicator of the availability of parallel channels. A logical AND operation is performed on the switching states of all effective paths; a value of 1 is taken when at least one path is fully closed, otherwise 0, which can identify the bottleneck effect of critical switches. Path connectivity can be considered a network centrality indicator of devices. The number of active paths directly connected to each node is counted, and weights are calculated using the PageRank algorithm, which can locate key device nodes in the power grid. Topological correlation can refer to the electrical coupling strength between devices. Based on the wiring diagram provided by the SCADA system, an adjacency matrix is ​​constructed, and the impedance ratio between devices is determined by power flow calculation, which can quantify the potential path strength of fault propagation.

[0051] The third calculation formula can refer to the matrix element generation algorithm, which takes the natural logarithm of the number of common paths, multiplies it by the effective coefficient, and then divides it by the harmonic mean of the connectivity between the two devices. This non-linear transformation enhances the distinguishability of the critical path. The fourth calculation formula can refer to the topology correlation transformation formula, which multiplies the matrix element values ​​by the device weights and then normalizes them, so that the calculation results fall within the comparable range of 0-1.

[0052] Multi-dimensional topology quantification enables precise definition of the fault impact range. The physical topology diagram intuitively displays electrical connections, path validity judgment reflects the operating mode in real time, the correlation matrix comprehensively evaluates the coupling relationship of equipment, and the final generated impact factor can scientifically predict the fault propagation path, providing decision support for the formulation of targeted protection measures.

[0053] In some embodiments, the third calculation formula includes: in, This represents the matrix element values ​​corresponding to device node m and device node q. This represents the number of common paths between device node m and device node q. This represents the degree of device node m. This represents the degree of device node q. represents the common valid coefficient between device node i and device node j, where m and q are natural numbers.

[0054] In some implementations, the fourth calculation formula includes: in, This represents the topological affinity between device node m and device node q, where Q represents the total number of devices. This represents the preset coefficient corresponding to the device type of device node q.

[0055] In some implementations, the second formula for calculating the influence factor of the associated equipment includes: Among them: Let m be the influence factor of the m-th associated device. Let be the partial derivative of the health index with respect to the l-th parameter. The value of the l-th monitoring parameter comes from the raw data of the multi-dimensional monitoring module. The standard deviation of the l-th parameter is obtained through statistical analysis of historical data. Let m be the topological affinity between device node m and device node q. η and η are preset adjustment coefficients, which are manually adjusted according to the equipment type. L is the number of monitoring parameters, and Q is the total number of equipment.

[0056] The monitoring parameter values ​​can refer to real-time acquired physical quantity data. Raw signals such as temperature gradients, ultrasonic amplitudes, and discharge pulses are obtained from a distributed sensor network and transmitted to edge computing nodes via a 4-20mA current loop, maintaining a measurement accuracy of 0.1%. The adjustment coefficient can refer to the model's adaptive parameters. Through online learning algorithms, the decay rate of τq and the power weight of η are dynamically adjusted. Combined with changes in equipment operation years and load rate, the influencing factor calculation can adapt to different operating conditions. The number of monitoring parameters can refer to the total number of feature dimensions, including 16 temperature field measurement points, 8 ultrasonic channels, and 4 sets of partial discharge antenna signals. Principal component analysis is used to screen out L=12 key feature quantities, reducing computational complexity. The number of associated devices can correspond to the fault propagation range. A third-order neighborhood search radius is set in the topology analysis, covering Q=7 types of devices with direct electrical connections and electromagnetic induction effects, fully covering possible cascading fault paths.

[0057] A dual-coupling mechanism enables precise quantification of fault propagation. The exponential term reflects the direct impact of abnormal monitoring parameters on the equipment itself, the radical term characterizes the fault propagation effect in the topological network, and the dynamic adjustment coefficient ensures that the model adapts to different equipment types. Ultimately, a physically meaningful correlation impact assessment is generated, providing a theoretical basis for formulating precise maintenance strategies.

[0058] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this invention. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A fault early warning system for wind power booster stations, characterized in that, This includes interconnected multi-dimensional monitoring modules, a composite heat dissipation system, and an edge computing early warning center; The multi-dimensional monitoring module collects temperature, partial discharge signals, and strain data of the booster station equipment through non-contact sensors; The composite heat dissipation system receives temperature data from the multi-dimensional monitoring module and performs active heat dissipation through the synergistic effect of phase change materials and heat pipes. The edge computing early warning center integrates multi-source data from the multi-dimensional monitoring module to generate device health assessments and fault warnings, and controls the operating mode of the composite heat dissipation system.

2. The system according to claim 1, characterized in that, The multi-dimensional monitoring module includes an infrared ultrasonic composite sensor, a partial discharge ultra-high frequency antenna, and a fiber optic strain sensor. The infrared ultrasonic composite sensor scans the surface of the transformer and switchgear to generate surface temperature data, and simultaneously collects internal discharge ultrasonic signals. The partial discharge ultra-high frequency antenna performs electromagnetic wave scanning on the busbar room space to generate partial discharge signal amplitude; The fiber optic strain sensor measures the strain of the high-voltage cable to generate micro-strain data.

3. The system according to claim 1, characterized in that, The composite heat dissipation system includes a transformer heat dissipation modification unit, an intelligent ventilation network, and a photovoltaic auxiliary unit. The transformer heat dissipation modification unit transfers heat from the transformer oil tank to the phase change material box via heat pipes. The intelligent ventilation network automatically adjusts the opening and closing of louvers and the direction of the fan based on the monitored temperature. The photovoltaic auxiliary unit supplies power to the heat dissipation system.

4. The system according to claim 1, characterized in that, The edge computing early warning center includes a data fusion unit, a fault tracing unit, and a display unit; The data fusion unit performs weighted fusion processing on temperature data, ultrasonic signals, and partial discharge signals to generate a health index. The fault tracing unit performs correlation analysis on the health index based on the equipment topology relationship to generate related equipment influence factors. The display unit converts the analysis results into visualized data.

5. The system according to claim 4, characterized in that, The data fusion unit performs weighted fusion processing on temperature data, ultrasonic signals, and partial discharge signals to generate a health index, including: The temperature data collected by the infrared and ultrasonic composite sensor is normalized to generate a standard temperature difference value, and the standard temperature difference value is weighted and squared to generate a temperature influence factor. Logarithmic transformation is performed on the ultrasonic signal intensity to generate signal feature values, and the cube root operation is performed on the signal feature values ​​to generate the ultrasonic influence factor. The amplitude of the raw signal acquired by the partial discharge UHF antenna is extracted to generate partial discharge characteristic quantities, and the partial discharge characteristic quantities are subjected to exponential weighted average calculation to generate partial discharge influence factors. The health index is calculated based on the temperature influence factor, ultrasound influence factor, and partial radiation influence factor, as well as the preset first calculation formula.

6. The system according to claim 5, characterized in that, The first calculation formula includes: ； Wherein, H is the health index, which is generated by the data fusion unit; The temperature difference value at the i-th temperature monitoring point is derived from the temperature scan result of the infrared-ultrasonic composite sensor. The intensity of the j-th ultrasonic signal comes from the ultrasonic acquisition result of the infrared ultrasonic composite sensor; The amplitude of the k-th partial discharge signal is derived from the electromagnetic wave scanning result of the partial discharge UHF antenna. , β and γ are preset weight coefficients, obtained through training with historical fault data; The total number of temperature monitoring points, This represents the total number of ultrasound monitoring points. This represents the total number of local emission monitoring points.

7. The system according to claim 4, characterized in that, The step of generating related equipment influence factors by performing correlation analysis on the health index based on equipment topology relationships includes: Collect electrical connection data between substation equipment to generate a physical topology map reflecting the physical connection status between equipment nodes. The connection data includes the number of connection paths between any two equipment nodes and at least one switch information in each connection path. The switch information indicates whether the switch is closed or closed. When all switches on the connection path between two equipment nodes are closed, the path validity coefficient corresponding to that path is 1; otherwise, it is 0. Construct a correlation matrix between device nodes, wherein the correlation matrix is ​​a square matrix, the correlation matrix includes multiple matrix element values, each matrix element value is calculated based on the number of common paths and common effective coefficient of at least one connection path between two device nodes, the path connectivity of each of the two devices, and a preset third calculation formula, and the common effective coefficient is 1 when the effective coefficient of any connection path between two device nodes is 1; Based on the matrix element values ​​in the correlation matrix, the topological correlation degree corresponding to each matrix element value is calculated according to the preset weights corresponding to each device type and the preset fourth calculation formula. The influence factor of the associated equipment is calculated based on the topological correlation degree.

8. The system according to claim 7, characterized in that, The third calculation formula includes: ； in, This represents the matrix element values ​​corresponding to device node m and device node q. This represents the number of common paths between device node m and device node q. This represents the degree of device node m. This represents the degree of device node q. represents the common valid coefficient between device node i and device node j, where m and q are natural numbers.

9. The system according to claim 8, characterized in that, The fourth calculation formula includes: ； in, This represents the topological affinity between device node m and device node q, where Q represents the total number of devices. This represents the preset coefficient corresponding to the device type of device node q.

10. The system according to claim 8, characterized in that, The second formula for calculating the influence factor of the associated equipment includes: ； Among them: Let m be the influence factor of the m-th associated device. Let be the partial derivative of the health index with respect to the l-th parameter. The value of the l-th monitoring parameter comes from the raw data of the multi-dimensional monitoring module. The standard deviation of the l-th parameter is obtained through statistical analysis of historical data. Let m be the topological affinity between device node m and device node q. η and η are preset adjustment coefficients, which are manually adjusted according to the equipment type. L is the number of monitoring parameters, and Q is the total number of equipment.

Images