Transformer winding deformation online monitoring system and method based on microwave resonance principle

Abstract

The invention relates to a transformer winding deformation online monitoring system based on a microwave resonance principle. The system adopts a layered architecture and specifically comprises a monitored entity, a sensor array in an oil tank, an edge computing and communication node and a cloud edge collaborative intelligent platform. The monitored entity is an oil-immersed transformer winding; the sensing array in the oil tank is fixed on the inner wall of the oil tank according to a preset space matrix based on a specifically designed microwave resonator and is used for sensing the displacement of a winding; the edge calculation and communication node integrates a radio frequency measurement unit, a multi-path switching unit, a core processing unit and a communication interface unit, and is responsible for quickly scanning the sensing array and extracting resonance characteristic parameters in real time; and the cloud edge collaborative intelligent platform is deployed in a remote server, integrates a deformation diagnosis model and a three-dimensional visualization engine, and is used for receiving the characteristic parameters and performing quantitative evaluation, positioning and visual early warning of deformation. According to the invention, in-service, real-time and non-contact online monitoring and intelligent diagnosis of the mechanical deformation of the transformer winding are realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power equipment state monitoring, and relates to a transformer winding deformation online monitoring system and method based on a microwave resonance principle. BACKGROUND

[0002] As the core equipment of power grid energy conversion and transmission, the operation reliability of oil-immersed power transformers is directly related to the safety and stability of the power system. According to statistics, mechanical failure of windings accounts for more than 40% of power outages caused by transformer failures worldwide, and insulation breakdown and short circuit faults caused by mechanical deformation are particularly prominent. As the "heart" component of the transformer, the winding is wound by multiple layers of insulated wires, which needs to withstand multiple physical field coupling effects during long-term operation: in terms of electromagnetic force, the electrodynamic force generated by short-circuit current can reach hundreds of kilo-newtons, causing axial compression or radial expansion of the winding; in terms of thermal stress, copper loss and iron loss caused by load fluctuations cause periodic changes in winding temperature, and insulation materials gradually age in thermal expansion and contraction; in terms of mechanical impact, vibration during transportation, mechanical stress during installation, and sudden events such as earthquakes can cause irreversible damage to the winding structure. These factors together cause the winding to gradually deform, such as radial bulging, axial displacement, and inter-pie displacement, and when the deformation exceeds the tolerance limit of the insulation material, it will cause local discharge, inter-turn short circuit and other malignant faults, seriously threatening the safety of the transformer itself and the stable operation of the power grid.

[0003] Traditional winding mechanical state detection technology has significant limitations. The frequency response analysis method (FRA) commonly used in the industry compares the impedance characteristic changes of the winding in a specific frequency range to diagnose deformation, and its detection accuracy can reach millimeter level, but it needs to apply a test signal after the transformer is shut down, and cannot reflect the dynamic changes during operation. Although the low-voltage short-circuit impedance method can be tested under voltage, it can only detect the overall deformation of the winding, and its sensitivity to local small deformation is insufficient, and the signal-to-noise ratio is difficult to guarantee when the test voltage is too low. More importantly, both methods are post-diagnosis means and cannot issue early warnings in the early stages of deformation, resulting in a long-term passive mode of "fault-driven" maintenance strategy. According to statistics of power equipment operation and maintenance data, the mean time between failures (MTBF) of transformers using traditional detection methods is about 8-10 years, and sudden failures caused by failure to detect timely account for as high as 65%, causing huge economic losses and social impact.

[0004] The emergence of microwave resonant technology has provided a revolutionary solution for winding deformation monitoring. Based on electromagnetic field theory, this technology causes a shift in resonant frequency when the distance between the resonator and the object being measured changes. By establishing a mathematical model of frequency shift and displacement, precise measurement of micrometer-level displacement can be achieved. Compared to traditional detection methods, microwave resonant technology has three significant advantages: First, its non-contact measurement characteristic eliminates the need for sensors on the winding, avoiding insulation risks associated with electrical connections; second, the low attenuation coefficient of millimeter-wave (30-300GHz) electromagnetic waves in transformer oil allows for deep monitoring by penetrating insulating paperboard; and third, it exhibits strong resistance to electromagnetic interference, maintaining measurement stability even in strong electric field environments. Experimental data shows that in a 25# transformer oil environment, a 24GHz resonant sensor achieves a detection sensitivity of 0.5MHz / μm for 10μm-level displacement, with a signal-to-noise ratio exceeding 40dB, fully meeting the requirements for early monitoring of winding deformation.

[0005] The online monitoring system based on microwave resonance employs a distributed sensing architecture, consisting of a resonant sensor array deployed on the inner wall of the fuel tank, an edge computing unit, and a cloud analysis platform. The sensor array is fabricated using flexible printed circuit (FPC) technology, allowing it to be mounted flush with the curved surface of the fuel tank. Each sensor operates independently without interference, forming a monitoring network covering the entire height of the winding. The edge computing unit integrates a high-speed sampling module and lightweight AI algorithms, enabling real-time processing of sensor data and extraction of feature parameters. By comparing with a digital twin model, it predicts the winding deformation trend. The cloud platform utilizes big data analytics, combining historical equipment data with the operating patterns of similar models to achieve fault warnings and optimized maintenance decisions. This system achieves a triple transformation: from "offline detection" to "online monitoring," from "overall assessment" to "local location," and from "passive maintenance" to "proactive prevention," improving operational efficiency by over 70%.

[0006] The online monitoring system based on the microwave resonance principle can sense subtle changes in the electromagnetic coupling parameters between the winding and the sensor in real time by deploying a sensor array and an edge intelligent processing unit on the inner wall of the oil tank. Combined with digital twin and cloud-edge collaborative analysis, it realizes the transformation from "periodic power outage inspection" to "continuous state perception and predictive maintenance", which significantly improves the intelligence level of transformer operation and maintenance and the safe operation capability of power grid assets. Summary of the Invention

[0007] In view of this, the purpose of the present invention is to provide an online monitoring system and method for transformer winding deformation based on the principle of microwave resonance, so as to solve the technical problem that the existing detection methods must be operated by power off and cannot achieve real-time online monitoring and early deformation warning, and realize high sensitivity, quantification and in-service real-time monitoring of winding mechanical deformation.

[0008] To achieve the above objectives, the present invention provides the following technical solution: An online monitoring system for transformer winding deformation based on the microwave resonance principle, comprising an in-tank sensor array, edge computing and communication nodes, and a cloud-edge collaborative intelligent platform; The internal sensing array is fixed to the inner wall of the transformer tank and is used to sense the mechanical displacement of the windings. The internal sensing array includes a microwave resonator array, a mechanical mounting structure, signal leads, and a sealing connector. The microwave resonator array consists of multiple planar resonators arranged in a preset spatial matrix. The mechanical mounting structure is used to rigidly fix and insulate the resonator array. The signal leads transmit the radio frequency signal of the resonator array to the outside of the tank through the sealing connector. The edge computing and communication node is deployed locally on the transformer and electrically connected to the sensor array inside the oil tank. The edge computing and communication node includes a core processing unit, a radio frequency measurement unit, a multiplexing and signal conditioning unit, an environmental sensor, a communication and interface unit, and a power supply and auxiliary unit. The core processing unit cyclically selects each resonator channel in the sensor array inside the oil tank through the multiplexing and signal conditioning unit, and controls the radio frequency measurement unit to measure the reflection coefficient S11 parameter of the selected channel. The core processing unit has a built-in algorithm to extract the current values ​​of the resonant frequency f and the quality factor Q from the S11 parameters, and calculates their offset (Δf, ΔQ) relative to a preset reference value to generate deformation characteristic parameters. The core processing unit also reads the operating condition data collected by the environmental sensor and uploads the deformation characteristic parameters and the operating condition data together via the communication and interface unit. The cloud-edge collaborative intelligent platform is deployed on a remote server and communicates with the edge computing and communication nodes. The cloud-edge collaborative intelligent platform includes an intelligent diagnosis and visualization subsystem and a data management and operation and maintenance subsystem. The intelligent diagnosis and visualization subsystem receives uploaded data, analyzes it through a deformation diagnosis AI model, outputs the deformation index, risk level and location information of the winding, and drives a 3D visualization engine for display. The data management and operation and maintenance subsystem is used for data storage, management and report generation.

[0009] Furthermore, the substrate of the microwave resonator array is made of Rogers RO4350B material, and its resonator unit is a rectangular open-ring structure with a target center frequency of 2.45GHz±50MHz; the mechanical mounting structure is made of polytetrafluoroethylene material and the resonator array is fixed in a three-dimensional matrix manner with 3 axial layers and 4 radial points; the signal lead is a coaxial cable with a characteristic impedance of 50Ω and is led out through an SMA type sealed connector with a withstand voltage rating of greater than 10kV.

[0010] Furthermore, the radio frequency measurement unit is a vector network analysis module built on the ADf4351 frequency synthesizer and ADL5380 quadrature demodulator chipset; the scanning frequency band of the radio frequency measurement unit is 2.2GHz to 2.6GHz, the single-point measurement time is less than 300 milliseconds, and it is controlled by the core processing unit through the SPI bus.

[0011] Furthermore, the multiplexing and signal conditioning unit includes an SP12T electromechanical RF switch and a low-noise amplifier; the RF switch has a channel switching time of less than 15 milliseconds and an isolation of better than 60dB at the 2.4GHz frequency point; the low-noise amplifier is used to amplify and condition the transmission signal between the RF measurement unit and the sensor array inside the fuel tank.

[0012] Furthermore, the core processing unit adopts an embedded industrial control computer based on the x86 architecture, which runs the Linux operating system and is used to execute the S11 parameter processing algorithm, task scheduling and communication control.

[0013] Furthermore, the communication and interface unit includes a 4G wireless communication module and an isolated wired communication interface; the 4G wireless communication module supports the MQTT protocol and is used for data transmission with the cloud-edge collaborative intelligent platform; the isolated wired communication interface is used for local data interaction.

[0014] Furthermore, the power supply and auxiliary unit includes a wide voltage input AC / DC power module and a backup power management circuit, used to power each unit within the edge computing and communication node and provide continuous power when the main power supply is abnormal.

[0015] Furthermore, the environmental sensors include an oil temperature sensor, a vibration sensor, and a load current sensor; the oil temperature sensor uses a platinum resistance temperature measurement scheme, the vibration sensor uses an accelerometer with an IEPE interface, and the load current sensor uses the Hall effect principle.

[0016] This invention also provides an online monitoring method for transformer winding deformation based on the microwave resonance principle, applied to the above-mentioned system, comprising the following steps: S1. System initialization and benchmark establishment: Under the condition that the transformer is in a healthy state and under typical operating conditions, the radio frequency measurement unit is controlled by edge computing and communication nodes to scan all resonator channels of the sensor array in the oil tank, measure and store the reference resonant frequency f0 and reference quality factor Q0 of each channel. S2, Periodic Online Scan: After the system enters the monitoring state, the core processing unit cycles through the multiplexer channel and the RF switch in the signal conditioning unit according to a preset period, and instructs the RF measurement unit to perform a frequency scan on the current channel to obtain its reflection coefficient S11 parameter. S3. Edge side feature extraction and processing: The core processing unit processes the S11 parameters in real time, extracts the current resonant frequency f and quality factor Q through algorithm fitting, calculates the relative offset Δf / f0 and ΔQ / Q0 between them and the corresponding reference value, and simultaneously reads the oil temperature, vibration and load current data collected by the environmental sensor. S4. Data encapsulation and transmission: The core processing unit packages the processed feature parameters (Δf / f0, ΔQ / Q0) with multi-source environmental data and uploads them to the cloud-edge collaborative intelligent platform in the form of encrypted MQTT messages through the 4G DTU in the communication and interface unit. S5. Cloud-based intelligent diagnosis and fusion analysis: After receiving data, the cloud platform stores it in a time-series database; the deformation diagnosis AI model in the intelligent diagnosis and visualization subsystem calls the digital twin model library, integrates real-time feature data, historical trends and multi-physics simulation results, and performs quantitative calculation of deformation degree, location positioning and risk level assessment. S6. Results Visualization and Early Warning Decision-Making: Diagnostic results drive the 3D visualization engine to generate a winding deformation heat map, which is dynamically displayed on the data dashboard. When the assessed deformation index or risk level exceeds the preset threshold, the system automatically triggers a graded early warning, pushes early warning information through the platform interface and API, and generates a diagnostic report containing location and maintenance suggestions, completing the closed loop from status perception to decision support.

[0017] Furthermore, step S5 also includes: S51. Data fusion and compensation steps: use real-time oil temperature data to compensate for frequency drift. S52. Call the digital twin model library to map the compensated frequency drift to the actual physical deformation of the winding. S53. Based on the spatial data pattern differences of the monitoring points, determine the phase and axial position of the deformation. S54. Combine the anomaly detection model trained with historical fault cases to assess the comprehensive risk level and output the deformation index.

[0018] The beneficial effects of this invention are as follows: This invention constructs a complete online monitoring system by designing a non-contact sensing array based on the microwave resonance principle, a high-performance edge computing node, and a cloud-edge collaborative intelligent platform. This system enables real-time, online, and high-precision monitoring of the mechanical deformation of oil-immersed transformer windings and early fault warning, effectively improving the transformer's condition perception capability and the level of intelligent operation and maintenance.

[0019] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0020] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a schematic diagram of the monitored entity and the sensor array inside the fuel tank in an embodiment of the present invention; Figure 2 This is a schematic diagram of edge computing and communication nodes in an embodiment of the present invention; Figure 3 This is a schematic diagram of the cloud-edge collaborative intelligent platform in an embodiment of the present invention; Figure 4 This is a flowchart illustrating the software system in an embodiment of the present invention. Detailed Implementation

[0021] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings.

[0022] The online monitoring system for transformer winding deformation based on the microwave resonance principle provided by this invention mainly includes an in-tank sensor array, edge computing and communication nodes, and a cloud-edge collaborative intelligent platform. The technical solution will be described in detail below through specific embodiments.

[0023] Example 1: This embodiment provides a specific implementation plan for an online monitoring system for transformer winding deformation based on the microwave resonance principle. The system adopts a hierarchical distributed architecture, consisting of four parts: the monitored entity, an in-tank sensor array, edge computing and communication nodes, and a cloud-edge collaborative intelligent platform. This forms a complete monitoring closed loop from physical sensing and edge processing to cloud-based intelligent decision-making. Its hardware and logic composition is similar to... Figure 1 , Figure 2 and Figure 3 The diagram shows a perfect correspondence.

[0024] 1. The monitored entity and the sensor array inside the fuel tank like Figure 1As shown above, the monitoring objects of this system are the windings of oil-immersed power transformers, including the high-voltage winding (panel type), the low-voltage winding (spiral type), and the voltage regulating winding, all of which are immersed in mineral insulating oil such as Karoon KN4010. The insulation distance between the high-voltage winding and ground is designed to be greater than 100mm, and the insulation distance between the low-voltage winding and ground is greater than 30mm. This spatial parameter is the basis for designing the sensor array layout and sensitivity calibration. Through electromagnetic-mechanical coupling simulation and experimental calibration, the radial deformation sensitivity of the high-voltage winding is approximately Δf / f0~0.05% / mm, and that of the low-voltage winding is approximately Δf / f0~0.03% / mm. This relationship is the physical basis for subsequent quantitative inversion of deformation.

[0025] As shown in the lower part of the figure, the sensor array inside the oil tank, serving as the system's direct sensing unit, is fixedly installed on the inner wall of the transformer oil tank and operates under long-term oil immersion. Its core is a microwave resonator array: utilizing Rogers RO4350B high-frequency board material with excellent dielectric constant and temperature stability (…). Using a substrate with r=3.66±0.05, a rectangular open-loop resonator was designed and fabricated. The target resonant frequency of a single resonator in oil medium was designed to be 2.45GHz ± 50MHz, with an unloaded quality factor (Q value) greater than 150 to ensure high sensitivity to minute displacements. The array adopts a three-dimensional matrix layout of "3 layers axially and 4 points radially", forming a total of 12 independent monitoring points to achieve all-round coverage of the spatial deformation of the winding.

[0026] The mechanical mounting structure is precision-machined from polytetrafluoroethylene (PTFE), utilizing its excellent insulation, chemical corrosion resistance, and low coefficient of thermal expansion. The resonator is securely fixed to the bracket using M4 304 stainless steel countersunk bolts. All mating surfaces are sealed with fluororubber O-rings, which have a long-term temperature resistance exceeding 150°C, ensuring the long-term sealing reliability of the transformer tank.

[0027] The signal leads use RG-316 coaxial cables with a characteristic impedance of 50Ω±2Ω, a uniform length of 1.5 meters (with an error of ±5 cm), and are terminated with SMA straight male connectors. All cables from all channels are ultimately converged and passed through a custom-designed multi-channel high-voltage sealed pass-through. This pass-through is a critical component, ensuring that each signal channel can withstand a power frequency withstand voltage greater than 10kV AC between the core wire and the housing, preventing internal high voltage leakage along the signal lines, while simultaneously guaranteeing an RF signal insertion loss of less than 0.5dB in the 2.4GHz band, achieving high-fidelity signal extraction.

[0028] 2. Edge computing and communication nodes like Figure 2 As shown, the edge computing and communication node is an intelligent, integrated processing unit deployed locally on the transformer, installed in an industrial enclosure with an IP65 protection rating. Its internal structure and operational coordination are as follows: Power Supply and Auxiliary Unit: Utilizing the Mean Well MDR-60-24 industrial-grade AC / DC switching power supply module, it supports a wide input range of AC85-264V and provides a stable DC 24V / 2.5A output. Internally integrated surge protection, overvoltage protection, and overcurrent protection circuitry. This unit also manages a supercapacitor backup power supply (UPS), which can continuously power critical components during short-term mains power outages (within seconds), ensuring uninterrupted monitoring and data security.

[0029] Core Processing Unit: The core of the control and computing system is an Advantech ARK-3520P embedded industrial computer. It is equipped with an Intel Celeron J1900 quad-core processor, 4GB DDR3L memory, and a 64GB mSATA solid-state drive, running the Ubuntu Core operating system. It provides a stable Linux environment and sufficient computing power for running real-time signal processing algorithms, task scheduling, and communication management.

[0030] Multiplexing and Signal Conditioning Unit: This unit is controlled by the GPIO of the core processing unit. Its core is a Mini-Circuits ZASWA-2-50DR+ SP12T (single-pole twelve-throw) electromechanical RF switch, used for cyclic switching between 12 sensing channels. The switching time is less than 15 milliseconds, and the channel isolation is better than 60dB at 2.4GHz, effectively preventing crosstalk. The switch output is connected to an ERA-3SM+ low-noise amplifier with a gain of approximately 20dB and a noise figure of approximately 3dB, used to compensate for cable and switch insertion losses, improving the signal-to-noise ratio and dynamic range of the signal sent to the measurement unit.

[0031] RF Measurement Unit: A vector network analysis module customized for this system, it is the core for achieving high-precision frequency measurement. It is built upon the ADf4351 broadband frequency synthesizer chip and the ADL5380 quadrature demodulator chipset. The core processing unit configures the ADf4351 via the SPI bus to generate a finely stepped scan signal from 2.2GHz to 2.6GHz, which is amplified and then used to excite the selected resonator. The reflected signal and the reference signal are mixed in the ADL5380, down-converted to obtain the baseband I / Q signal, converted by an ADC, and read by the core processing unit to reconstruct the complex reflection coefficient S11 at each frequency point. This module has a single-point measurement speed of less than 300 milliseconds and a dynamic range greater than 80dB.

[0032] Environmental sensor module: Used to synchronously collect operating condition information that affects diagnosis.

[0033] Oil temperature measurement: A PT100 three-wire platinum resistance temperature sensor is used, which is connected to the MAX31865 dedicated high-precision temperature measurement chip. The latter provides digital temperature values ​​directly to the core processing unit through the SPI interface, with an accuracy of 0.1°C.

[0034] Vibration measurement: A PCB 352C33 model IEPE (integrated piezoelectric) accelerometer is used, which outputs a low-impedance voltage signal proportional to the vibration acceleration and is connected to the analog input port of the core processing unit.

[0035] Load current measurement: The LEM Lf 510-S Hall effect current sensor is used. The primary current is isolated and converted into a proportional voltage signal output through the magnetic balance principle, which is also connected to the analog input port of the core processing unit.

[0036] Communication and Interface Unit: Remote wireless communication: Employs the Renrenlian USR-G781 industrial-grade 4G DTU. The core processing unit is configured using the AT command set via a serial port (through a USB-to-serial chip) to establish a TLS encrypted connection with the cloud platform as an MQTT client, enabling the uploading of monitoring data and the downlink reception of cloud commands.

[0037] Local wired interface: An isolated RS-485 interface based on the ADM2483 high isolation chip is used to connect this node to the substation’s existing local monitoring network to achieve data sharing.

[0038] 3. Cloud-Edge Collaborative Intelligent Platform like Figure 3 As shown, the cloud-edge collaborative intelligent platform is deployed on a remote server cluster, forming the "intelligent brain" of the system. It adopts a microservice architecture and mainly includes two subsystems: Intelligent Diagnosis and Visualization Subsystem: Data access and storage: Deploy a high-performance MQTT Broker cluster (such as EMQX) to receive massive amounts of data from edge nodes. Real-time streaming data is written to the InfluxDB time-series database for fast querying and real-time trend display; structured data such as device assets, user information, diagnostic reports, and maintenance work orders are stored in a PostgreSQL relational database.

[0039] Deformation Diagnosis AI Model: This is the core of the software. The model obtains real-time and historical feature parameters (Δf, ΔQ) and environmental data (oil temperature, vibration, load) from the database. The diagnostic process includes: First, data fusion and compensation, such as using real-time oil temperature data to compensate for frequency drift; Second, physical quantity inversion, calling a digital twin model library strictly corresponding to the transformer model (this library contains multiphysics simulation results based on accurate 3D models and material properties), mapping the compensated frequency drift to the actual physical deformation of the winding (millimeter level); Third, comprehensive assessment and location, using machine learning algorithms (such as anomaly detection models trained based on historical fault cases) to assess the comprehensive risk level of the current state, and based on the spatial data pattern differences of 12 monitoring points, intelligently determining the phase and axial position where deformation is most likely to occur (e.g., "middle of phase C of high-voltage winding").

[0040] 3D Visualization Engine and Data Dashboard: Developed based on WebGL technology (such as the Three.js framework). The engine reads the transformer's 3D model and, based on the deformation location and degree data output by the diagnostic engine, renders a 3D deformation heatmap with color gradient changes in real time, overlaying it onto the corresponding locations on the model to provide an extremely intuitive status display. Simultaneously, a multi-dimensional data dashboard displays the Δf trends of each channel, environmental parameter curves, historical risk levels, etc., in rich chart formats.

[0041] Data Management and Operation Subsystem: It provides complete asset ledger management, user access control (RBAC model), and operation audit logs.

[0042] Alarm Management Module: Based on the diagnosed risk level (such as alert, abnormal, dangerous), it automatically triggers a graded early warning mechanism and notifies the corresponding level of maintenance personnel through various means such as platform interface, SMS, and mobile application push.

[0043] Reporting and Work Order Engine: Automatically generates standardized diagnostic reports (supports PDF / Word export) that include deformation location, severity, trend analysis, and maintenance recommendations. It can also automatically create or associate maintenance work orders and push them to the enterprise's asset management system, forming a closed loop of "monitoring-diagnosis-decision-maintenance".

[0044] System Workflow Referring to the accompanying drawings and system architecture, the specific implementation of the workflow of this system is as follows, which corresponds to and refines the steps described in claim 9: System Initialization and Baseline Establishment (S1): During the initial installation or commissioning of the system after major overhaul, under typical operating conditions where the transformer is confirmed to be in a healthy state and the load is stable, the core processing unit of the edge node executes the baseline learning program. Its control RF switch sequentially selects channels 1 to 12. For each channel, the RF measurement unit is instructed to perform a full-band (2.2-2.6GHz) scan to acquire a high-precision baseline S11 curve. The built-in algorithm accurately extracts the baseline resonant frequency f0 and baseline quality factor Q0 for each channel from the curve, and stores this information, along with operating condition information such as oil temperature, in encrypted form on the local solid-state drive, establishing a unique "health fingerprint" database for the transformer.

[0045] Periodic online scanning and edge processing (S2, S3, S4): After the system enters normal monitoring mode, the edge nodes automatically execute the following cycle according to a preset period (e.g., 1 minute): Channel selection and scanning (S2): The core processing unit controls the RF switch through GPIO to select the Nth (N increases cyclically) sensing channel.

[0046] Fast measurement: Configure and trigger the RF measurement unit via the SPI bus to perform a fast scan (<300ms) on the current channel and obtain the latest S11 parameter array.

[0047] Real-time edge feature extraction (S3): The core processing unit runs a real-time fitting algorithm (such as the phase zero-crossing method) to process the S11 data and quickly calculates the resonant frequency f and quality factor Q of the current channel. Then, the core feature quantities are calculated—normalized frequency offset Δf / f0 and normalized quality factor change ΔQ / Q0.

[0048] Environmental data synchronous acquisition: Within the same time slice, the oil temperature value of MAX31865 is read via SPI, and the voltage values ​​of vibration and load current sensors are obtained by ADC sampling and converted into engineering values.

[0049] Looping and Packaging: After completing one channel, switch to the next channel and repeat the above process until all 12 channels have been scanned. Then, all 12 sets of feature parameters (Δf / f0, ΔQ / Q0), synchronized environmental data, precise timestamps, and device IDs generated in this cycle are packaged into a unified JSON format data packet.

[0050] Data Encryption Transmission (S4): The core processing unit sends data packets to the 4G DTU via serial port. The DTU then publishes the data packets to the topic specified by the cloud platform using the MQTT protocol (QoS level 1) through the established TLS encrypted tunnel, completing the uplink transmission.

[0051] Cloud-based intelligent diagnosis and fusion analysis (S5): After receiving the data, the cloud platform message middleware persists it to the time-series database and triggers the intelligent diagnosis engine.

[0052] The diagnostic engine extracts real-time data streams and historical trend data of the device from the database.

[0053] The digital twin model corresponding to the device is invoked to perform multi-source information fusion. First, compensation and normalization processing are performed for the effects of operating conditions such as temperature and load.

[0054] The deformation diagnosis AI model works based on fused data: by matching and analyzing the monitored spatially distributed frequency offset patterns with the simulated fault maps in the digital twin model library, the deformation of the winding (unit: millimeters) is quantitatively calculated, the current overall risk level (such as the deformation index of 0-100) is assessed, and the most likely area of ​​deformation is located (such as "high voltage winding phase B, between the 5th and 6th wires").

[0055] Results visualization and early warning decision-making (S6): Visualization: The diagnostic results drive the 3D visualization engine in real time, dynamically rendering a deformation heatmap on the transformer's 3D model in the web user interface, with color depth representing the severity of deformation. The data dashboard updates the curves of each parameter synchronously.

[0056] Intelligent Early Warning and Decision Support: The diagnostic engine compares the calculated deformation index with preset multi-level thresholds. Once a threshold is exceeded, the system immediately generates an early warning event and stores it in the business database. The alarm management module initiates a tiered early warning process based on the risk level (attention, abnormal, dangerous), pushing alarm information through various methods such as platform messages, SMS, and voice notifications. Simultaneously, the reporting engine automatically generates standardized diagnostic reports containing detailed location information, severity analysis, historical comparisons, and maintenance priority suggestions. It can also automatically create maintenance work orders and push them to the operation and maintenance management system, forming a complete decision support closed loop.

[0057] like Figure 4 The diagram illustrating the system workflow and data communication of this invention fully demonstrates the closed-loop operation process from edge perception to cloud-based intelligent decision-making. The following is a detailed description of the steps illustrated: 1. System Startup, Self-Test, and Initialization: After power-on, the system first executes a self-startup and self-test process to diagnose the status of the RF measurement unit, multiplexer, environmental sensors, and communication modules within the edge computing and communication nodes. After successful self-test, the system enters the initialization configuration and baseline learning phase. During this phase, the edge node loads a pre-installed digital certificate and establishes a secure encrypted tunnel based on the TLS 1.3 protocol with the cloud MQTT Broker via the 4G DTU, completing two-way authentication and negotiating communication parameters (such as heartbeat interval and QoS level). Subsequently, when the transformer is under no-load or steady-state healthy operation, the system automatically performs baseline learning: the core processing unit controls the Mini-Circuits ZASWA-2-50DR+ RF switch to sequentially select the 12 resonator channels of the sensor array within the tank, and instructs a custom VNA module based on the ADf4351+ADL5380 chipset to perform a full-band (2.2-2.6GHz) scan on each channel. The edge algorithm (running on Ubuntu Core system) analyzes the S11 parameters obtained from the scan in real time, accurately extracts and stores the reference resonant frequency f0 and reference quality factor Q0 of each channel locally, and establishes a digital health baseline profile for the device.

[0058] 2. Periodic Main Detection Loop and Edge Feature Extraction: After the system has been running normally, it enters the main detection loop. Within each monitoring cycle, the core processing unit performs the following sequence of operations: Channel selection: The RF switch is controlled by GPIO to cyclically select one resonator channel (channel N).

[0059] Signal scanning and acquisition: The VNA module is triggered via the SPI bus to perform a fast point-frequency scan (single point <300ms) on the current channel to obtain its latest reflection coefficient S11 parameter array. Simultaneously, the digital value of the PT100 oil temperature acquired by the MAX31865 chip is read via the I2C interface, and analog signals from the PCB 352C33 vibration sensor and the LEM Lf 510-S current sensor are simultaneously acquired via the ADC.

[0060] Real-time signal processing: The built-in edge algorithm performs phase fitting and bandwidth analysis on the S11 data, calculates the resonant frequency f and quality factor Q of the current channel in real time, and then calculates the core characteristic quantities—normalized frequency offset Δf / f0 and normalized quality factor change ΔQ / Q0.

[0061] Looping and Data Fusion: After processing one channel, check if the scanning and calculation of all 12 channels have been completed. If not, return to the channel selection step. Once all channels are complete, package and fuse the characteristic parameter arrays of all channels in this cycle, the synchronously acquired oil temperature, vibration, and load current data, along with the device ID and precise timestamp, into a structured JSON data packet.

[0062] 3. Secure data transmission: Packaged data is uploaded via the Yourenren USR-G781 4G DTU. Data is published to the designated topic on the cloud platform via an established TLS secure tunnel using the MQTT protocol (QoS level set to 1, ensuring at least one delivery). The system has a robust communication guarantee mechanism: maintaining periodic heartbeats to keep the connection alive; configuring will messages so the cloud can immediately detect abnormal device offline situations; and possessing breakpoint resumption capabilities, caching data locally during temporary network interruptions and automatically resuming transmission once the connection is restored.

[0063] 4. Cloud-based Intelligent Diagnosis and Analysis Decision-Making: During the data reception and storage phase, the cloud platform performs integrity verification, timestamp continuity checks, and format parsing on uploaded data packets. Data that passes verification is stored in the InfluxDB time-series database. Subsequently, the cloud-based intelligent diagnosis platform initiates a deep analysis process: Model Invocation and Data Fusion: The diagnostic engine invokes simulation models from the digital twin model library that perfectly correspond to the transformer model, winding structure, and sensor array layout. Multi-source data fusion and AI diagnostics are then initiated. The deformation diagnostic AI model deeply integrates and cross-analyzes real-time characteristic parameters (Δf / f0, ΔQ / Q0), environmental data, historical trend sequences, and multi-physics simulation results from the digital twin.

[0064] Quantitative Calculation and Risk Assessment: The model quantitatively calculates the degree and location of deformation, outputting the physical deformation of the winding (in millimeters) and its location information in three-dimensional space (e.g., "high voltage winding B phase 2-3 wire disc"). Simultaneously, it performs risk assessment and trend prediction, generating a comprehensive deformation index and future evolution trend.

[0065] Decision-making and early warning generation: The system compares the assessment results with preset multi-level safety thresholds to determine whether the thresholds have been exceeded. If the thresholds are not exceeded, the health status database is updated only. If the thresholds are exceeded, an early warning event and report are immediately generated, and a tiered early warning mechanism is triggered (pushed through multiple channels such as the platform, SMS, and APP according to the risk level).

[0066] 5. Visualized Interaction and Closed-Loop Operation and Maintenance: All diagnostic and early warning information drives the visualization and decision support interface. 3D visualization: The WebGL-based 3D visualization engine uses diagnostic results to render a 3D deformation heat map of the transformer winding in real time in the browser, intuitively mapping the location and severity of the fault.

[0067] Comprehensive data display: The data dashboard dynamically displays real-time characteristic parameter curves, environmental parameters, and historical curve comparison views for each channel.

[0068] Operations and Maintenance Decision Management: All activated alerts are aggregated in an alert list and integrated into the work order management system. Operations and maintenance personnel can confirm alerts, dispatch work orders, and record processing information through the interface.

[0069] Closed-loop feedback optimization: After on-site handling is completed, the results can be fed back to the system. This data, verified on-site, can be used to optimize the parameters of the digital twin model corresponding to the equipment or adjust the thresholds of the diagnostic algorithm, thereby realizing a complete intelligent operation and maintenance closed loop of system self-learning and continuous optimization.

[0070] Example 2: This example provides an online monitoring system for transformer winding deformation based on the microwave resonance principle. Its specific composition, hardware connection relationship, workflow, and technical effects are exactly the same as those described in Example 1. This system achieves online monitoring and intelligent diagnosis of transformer winding deformation through the collaborative work of an in-tank sensor array, edge computing and communication nodes, and a cloud-edge collaborative intelligent platform.

[0071] Example 3: This example provides a non-transitory computer-readable storage medium storing computer program instructions. When the computer program instructions are executed by the processor of an electronic device, the electronic device is able to perform the online monitoring method for transformer winding deformation based on the microwave resonance principle as described in Example 1 and claim 9.

[0072] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0073] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0074] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0075] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0076] Those skilled in the art will understand that all or part of the steps in the above facts and methods can be implemented by a program instructing related hardware. The program or the program described therein can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps: at this time, the corresponding method steps are introduced. The storage medium can be ROM / RAM, magnetic disk, optical disk, etc.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications should be covered within the scope of the claims of the present invention.

Claims

1. A transformer winding deformation on-line monitoring system based on the principle of microwave resonance, characterized in that: The system comprises an in-tank sensor array, an edge computing and communication node, and a cloud-edge collaborative intelligent platform; The in-tank sensor array is fixed to the inner wall of the transformer tank and is used for sensing mechanical displacement of the winding; the in-tank sensor array comprises a microwave resonator array, a mechanical mounting structure, a signal lead and a sealing through connector; the microwave resonator array is composed of a plurality of planar resonators arranged in a preset spatial matrix; the mechanical mounting structure is used for rigidly fixing and insulating sealing of the resonator array; the signal lead is led out to the outside of the tank through the sealing through connector. The edge computing and communication node is deployed locally to the transformer and is electrically connected to the in-tank sensor array; the edge computing and communication node comprises a core processing unit, a radio frequency measurement unit, a multi-channel switching and signal conditioning unit, an environmental sensor, a communication and interface unit, and a power supply and auxiliary unit; the core processing unit cyclically selects each resonator channel in the in-tank sensor array through the multi-channel switching and signal conditioning unit, and controls the radio frequency measurement unit to measure the reflection coefficient S11 parameter of the selected channel; the core processing unit extracts the current values of the resonant frequency f and the quality factor Q from the S11 parameter, and calculates the offset (Δf, ΔQ) thereof relative to the preset reference value to generate a deformation characteristic parameter; the core processing unit also reads the working condition data collected by the environmental sensor, and uploads the deformation characteristic parameter and the working condition data together via the communication and interface unit; The cloud-edge collaborative intelligent platform is deployed on a remote server and is communicatively connected to the edge computing and communication node; the cloud-edge collaborative intelligent platform comprises an intelligent diagnosis and visualization subsystem and a data management and operation and maintenance subsystem; the intelligent diagnosis and visualization subsystem receives the uploaded data, analyzes the data through a deformation diagnosis AI model, outputs the deformation index, risk level and positioning information of the winding, and drives a three-dimensional visualization engine to display; the data management and operation and maintenance subsystem is used for data storage, management and report generation.

2. The transformer winding shape on-line monitoring system based on the principle of microwave resonance according to claim 1, characterized in that: The substrate of the microwave resonator array is made of Rogers RO4350B material, the resonator unit is a rectangular open loop structure, and the target center frequency is 2.45GHz±50MHz; the mechanical mounting structure is made of polytetrafluoroethylene material and fixes the resonator array in a three-dimensional matrix manner with three axial layers and four radial points; the signal lead is a coaxial cable with a characteristic impedance of 50Ω, and is led out through an SMA type sealing through connector with a voltage resistance level greater than 10kV.

3. The transformer winding shape on-line monitoring system based on the principle of microwave resonance according to claim 2, characterized in that: The radio frequency measurement unit is a vector network analysis module constructed based on ADf4351 frequency synthesizer and ADL5380 quadrature demodulator chip set; the scanning frequency range of the radio frequency measurement unit is 2.2GHz to 2.6GHz, the single-point measurement time is less than 300 milliseconds, and the core processing unit controls the radio frequency measurement unit through the SPI bus.

4. The transformer winding shape on-line monitoring system based on the principle of microwave resonance according to claim 3, characterized in that: The multi-path switching and signal conditioning unit comprises an SP12T electromechanical radio frequency switch and a low-noise amplifier; the channel switching time of the radio frequency switch is less than 15 milliseconds, and the isolation at a frequency of 2.4 GHz is better than 60 dB; and the low-noise amplifier is used for amplifying and conditioning the transmission signals between the radio frequency measurement unit and the sensor array in the oil tank.

5. The transformer winding shape on-line monitoring system based on the principle of microwave resonance according to claim 4, characterized in that: The core processing unit adopts an embedded industrial computer based on an x86 architecture, which runs a Linux operating system and is used for executing the S11 parameter processing algorithm, task scheduling and communication control.

6. The transformer winding shape on-line monitoring system based on the principle of microwave resonance according to claim 5, characterized in that: The communication and interface unit comprises a 4G wireless communication module and an isolated wired communication interface; the 4G wireless communication module supports the MQTT protocol and is used for data transmission with the cloud-edge collaborative intelligent platform; and the isolated wired communication interface is used for local data interaction.

7. The transformer winding shape on-line monitoring system based on the principle of microwave resonance according to claim 6, characterized in that: The power supply and auxiliary unit comprises an AC / DC power supply module with a wide voltage input and a backup power management circuit, which are used for supplying power to each unit in the edge computing and communication node and providing continuous power when the main power supply is abnormal.

8. The transformer winding form online monitoring system based on the principle of microwave resonance according to claim 7, characterized in that: The environmental sensors comprise an oil temperature sensor, a vibration sensor and a load current sensor; the oil temperature sensor adopts a platinum resistance temperature measurement scheme, the vibration sensor adopts an IEPE interface accelerometer, and the load current sensor adopts a Hall effect principle.

9. A transformer winding deformation on-line monitoring method based on the principle of microwave resonance, applied to the system of any one of claims 1 to 8, characterized in that: The method comprises the following steps: S1, system initialization and reference establishment: when the transformer is in a healthy state and a typical working condition, the edge computing and communication node controls the radio frequency measurement unit to scan all resonator channels of the sensor array in the oil tank, measures and stores the reference resonant frequency f0 and the reference quality factor Q0 of each channel; S2, periodic online scanning: after the system enters the monitoring state, the core processing unit cyclically selects each resonator channel through the radio frequency switch in the multi-path switching and signal conditioning unit according to a preset period, and instructs the radio frequency measurement unit to perform frequency scanning on the current channel to obtain the reflection coefficient S11 parameter; S3, edge side feature extraction and processing: the core processing unit processes the S11 parameter in real time, extracts the current resonant frequency f and the quality factor Q through algorithm fitting, calculates the relative offset Δf / f0 and ΔQ / Q0 of the current resonant frequency f and the quality factor Q with respect to the corresponding reference values, and synchronously reads the oil temperature, vibration and load current data collected by the environmental sensors; S4, data packaging and transmission: the core processing unit packages the feature parameters (Δf / f0, ΔQ / Q0) obtained by processing and the multi-source environmental data, and uploads them to the cloud-edge collaborative intelligent platform in the form of encrypted MQTT messages through the 4G DTU in the communication and interface unit; S5, cloud-side intelligent diagnosis and fusion analysis: after the cloud platform receives the data, the data is stored in a time series database; the deformation diagnosis AI model in the intelligent diagnosis and visualization subsystem calls a digital twin model library, fuses real-time feature data, historical trends and multi-physical field simulation results, and performs deformation degree quantitative calculation, position positioning and risk level assessment. S6, result visualization and early warning decision: the diagnostic result drives the three-dimensional visualization engine to generate winding deformation thermodynamic diagram, and dynamically displays on the data board; when the evaluated deformation index or risk level exceeds the preset threshold, the system automatically triggers a hierarchical early warning, pushes the early warning information through the platform interface, and generates a diagnostic report containing positioning and maintenance suggestions, completing the closed loop from state perception to decision support.

10. The transformer winding form on-line monitoring method based on the principle of microwave resonance according to claim 9, characterized in that: The step S5 further comprises: S51, data fusion and compensation step, using real-time oil temperature data to compensate for temperature drift of frequency offset; S52, calling a digital twin model library, mapping the compensated frequency offset to the actual physical deformation of the winding; S53, based on the spatial data mode difference of the monitoring points, judging the phase and axial position of the deformation; S54, combining the abnormal detection model trained by the historical fault cases, evaluating the comprehensive risk level and outputting the deformation index.

Images