Integrated intelligent monitoring and protection system for power electronic components of intermediate frequency power supply

Abstract

The application discloses an integrated intelligent monitoring and protection system for power electronic components of a medium-frequency power supply, which comprises a distributed intelligent measurement layer, a high-speed data aggregation layer and a collaborative analysis and evaluation layer.The distributed intelligent measurement layer comprises a plurality of intelligent measurement nodes, which are used for sampling electrical operating parameters of the components, calculating local characteristic parameters and generating primary abnormality identification when an abnormality occurs.The high-speed data aggregation layer interconnects the intelligent measurement nodes through a deterministic communication network, realizes time alignment and aggregation upload of data, and the collaborative analysis and evaluation layer fuses data of all nodes, performs electric-thermal-aging collaborative analysis at the component level, and outputs health state evaluation and predictive maintenance information.Through operation and thermal coupling and life prediction models, real-time junction temperature, aging damage and residual life are calculated, and health grading and fault early warning are performed.The system realizes intelligent online monitoring, accurate state evaluation and predictive protection of the power electronic components, and effectively improves the reliability and maintainability of the medium-frequency power supply.

Description

Technical Field

[0001] This invention belongs to the field of power electronics technology, and specifically discloses an integrated intelligent monitoring and protection system for power electronic components in medium-frequency power supplies. Background Technology

[0002] Medium-frequency power supplies are widely used in critical fields such as aviation, shipbuilding, special industrial heating, and radar. Their core converter unit consists of multiple power electronic switching devices connected in parallel or series. Existing technologies for monitoring and protecting these components have the following shortcomings:

[0003] Information isolation: Parameters such as voltage, current, and temperature are usually collected by discrete sensors, resulting in poor data synchronization and difficulty in accurately reflecting the dynamic operating conditions of devices.

[0004] Protection lag: Protection strategies are mostly based on hardware threshold comparison (such as desaturation detection), which only take effect after a fault occurs. This is "post-event protection" and cannot provide early warning or suppress the occurrence of faults.

[0005] Lack of health management: It is impossible to assess the junction temperature fluctuations and aging status of devices in real time (such as bonding wire fatigue and substrate solder layer aging). Maintenance relies on periodic inspections or replacement after failure, which is inefficient.

[0006] Poor traceability: When a failure occurs, there is a lack of multi-dimensional data records from hundreds of milliseconds to several seconds before the failure, which is not conducive to root cause analysis.

[0007] Therefore, it is necessary to invent an integrated intelligent monitoring and protection system for power electronic components in medium-frequency power supplies to solve the above problems. Summary of the Invention

[0008] To overcome the aforementioned deficiencies in the prior art, this invention provides an integrated intelligent monitoring and protection system for power electronic components in medium-frequency power supplies. The distributed intelligent measurement layer comprises multiple intelligent measurement nodes for sampling the electrical operating parameters of the components, calculating local characteristic parameters, and generating primary anomaly identifiers when anomalies occur. A high-speed data aggregation layer interconnects the intelligent measurement nodes via a deterministic communication network, achieving time alignment and aggregation of data uploads. A collaborative analysis and evaluation layer integrates data from all nodes, performs component-level electro-thermal-aging collaborative analysis, and outputs health status assessments and predictive maintenance information. By running an electro-thermal coupling and lifespan prediction model, the system calculates real-time junction temperature, aging damage, and remaining lifespan, and performs health grading and fault warnings, effectively solving the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution: an integrated intelligent monitoring and protection system for power electronic components in a medium-frequency power supply, comprising:

[0010] The distributed intelligent measurement layer includes multiple intelligent measurement nodes that are set up one-to-one with each power electronic component. These nodes are used to synchronously and frequently sample the multi-dimensional electrical operating parameters of the corresponding components, calculate local characteristic parameters based on the sampled data, and generate a primary anomaly flag when the local characteristic parameters exceed a local threshold.

[0011] A high-speed data aggregation layer is used to interconnect the multiple intelligent measurement nodes in a deterministic communication network topology, enabling time alignment and aggregation of local feature parameters of each node for uploading.

[0012] The collaborative analysis and evaluation layer, connected to the high-speed data aggregation layer, is used to receive and fuse local feature parameters from all intelligent measurement nodes, perform component-level electro-thermal-aging collaborative analysis, and output health status assessment results and predictive maintenance information.

[0013] The collaborative analysis and evaluation layer includes:

[0014] The condition assessment and diagnosis unit, implemented based on software algorithms, is used to run the electrothermal coupling model and the lifetime prediction model. Based on the local characteristic parameters, it calculates the real-time junction temperature, junction temperature fluctuation, power loss, cumulative aging damage and remaining useful life of each component, and performs health status classification and fault warning based on preset rules.

[0015] The data recording and case library management unit is used to continuously record system operation data, and when the primary anomaly flag is triggered or the health status level reaches the warning level, it automatically saves the original sampling data and local feature parameters of all relevant intelligent measurement nodes within 30 seconds before and after the event, forming a traceable diagnostic case library.

[0016] The power electronic components include IGBTs, MOSFETs, or thyristors.

[0017] Preferably, the intelligent measurement node includes:

[0018] The current measurement unit is used to measure the load current flowing through power electronic components;

[0019] A voltage measurement unit is used to simultaneously measure the on-state voltage drop and off-state transient voltage of the components;

[0020] The junction temperature estimation unit is used to back-calculate the transient junction temperature of the component in real time based on the on-state voltage drop and through a pre-calibrated voltage drop-temperature relationship curve.

[0021] An edge processing unit, connected to the current measurement unit, voltage measurement unit, and junction temperature estimation unit respectively, is used to perform the synchronous sampling and calculation to obtain the local characteristic parameters including peak current, current change rate, on-state voltage drop, voltage change rate, instantaneous power consumption, and transient junction temperature.

[0022] Preferably, the junction temperature estimation unit injects a constant measuring current into the power electronic component that does not affect the operation of the main circuit, detects its on-state voltage drop under this current, and back-calculates the transient junction temperature based on the pre-calibrated voltage drop-temperature relationship.

[0023] Preferably, the high-speed data aggregation layer adopts a topology based on time-sensitive networking or daisy-chain fiber optic Ethernet, and the intelligent measurement node is embedded in the topology as a data acquisition terminal.

[0024] Preferably, the collaborative analysis and evaluation layer further includes a protection logic execution unit, implemented by a field-programmable gate array, for receiving the health status evaluation result or the primary anomaly identifier, and generating system-level protection control instructions according to preset logic.

[0025] Preferably, the condition assessment and diagnostic unit is used for:

[0026] Based on the instantaneous power consumption and transient junction temperature in the local characteristic parameters, the junction temperature fluctuation spectrum and thermal stress are calculated through the electrothermal coupling model;

[0027] Based on the thermal stress, the cumulative lifetime damage is calculated using the lifetime prediction model, and the remaining useful lifetime is estimated.

[0028] Based on long-term historical data of the local characteristic parameters, the performance degradation trend of the components is analyzed.

[0029] Preferably, when the performance degradation trend indicates that the linear rate of increase of the on-state saturation voltage drop of the component continues to exceed a specified proportion of its initial value, or the junction-shell thermal resistance estimated based on the electrothermal model shows a continuous increasing trend, an early warning message indicating aging of the bonding wire or substrate solder layer is generated.

[0030] Preferably, the data recording and case library management unit is configured to: store the original sampling data, local feature parameters and system operating condition snapshots in a standardized format; and store the diagnostic case library according to event type and component type, supporting case retrieval and optimized invocation of life prediction models.

[0031] Preferably, the collaborative analysis and evaluation layer also includes a cloud platform interface for uploading the health status evaluation and prediction results and condensed feature data to a cloud server.

[0032] Preferably, the edge processing unit further includes a self-calibration module and a self-diagnosis module. The self-calibration module is used to periodically perform zero-point and gain calibration on the current measurement unit and the voltage measurement unit; the self-diagnosis module is used to determine the working status of the intelligent measurement node itself by analyzing the rationality of the measurement data.

[0033] The technical effects and advantages of this invention are as follows:

[0034] 1. By equipping each key power electronic component with an independent intelligent measurement node, synchronous, high-frequency sampling of key electrical parameters such as load current, on-state voltage drop, and turn-off voltage is achieved, and transient junction temperature is inferred in real time based on pre-calibrated curves. This distributed direct measurement method overcomes the shortcomings of low accuracy and large delay of traditional indirect estimation or single-point monitoring, and can accurately capture the actual operating status and transient stress of components, providing a high-precision data foundation for early anomaly identification;

[0035] 2. The system ensures strict time alignment of all node data through a high-speed data aggregation layer, enabling the collaborative analysis layer to perform electro-thermal-aging collaborative analysis based on fused system-wide data. By running electro-thermal coupling models and lifetime prediction models, it can not only calculate real-time junction temperature and fluctuations, but also quantify cumulative aging damage and predict remaining useful life. This multi-physics, long-term comprehensive analysis can more accurately assess the health status of components, identify potential degradation trends such as bond wire aging and solder layer fatigue, and achieve a leap from simple threshold alarms to state-based, precise predictive maintenance.

[0036] 3. Through continuous health status grading and trend analysis, the system can output clear predictive maintenance information, guiding users to perform planned and purposeful maintenance and avoiding unplanned downtime. The case library supports retrieval and analysis by event and type, which helps to quickly diagnose recurring faults. In addition, the cloud platform interface design supports data uploading to the cloud, facilitating more complex big data analysis, remote expert diagnosis, and full lifecycle management, significantly improving the intelligent operation and maintenance level of the entire power system. Attached Figure Description

[0037] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0038] Figure 1 This is a diagram of the overall system architecture of the present invention.

[0039] Figure 2 This is a schematic diagram of the intelligent measurement node composition of the present invention.

[0040] Figure 3This is a flowchart illustrating the protection logic execution of the present invention. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] This invention provides, for example Figure 1 The integrated intelligent monitoring and protection system for power electronic components in a medium-frequency power supply, as shown, includes:

[0043] The distributed intelligent measurement layer includes multiple intelligent measurement nodes that are set up one-to-one with each power electronic component. These nodes are used to synchronously and frequently sample the multi-dimensional electrical operating parameters of the corresponding components, calculate local characteristic parameters based on the sampled data, and generate a primary anomaly flag when the local characteristic parameters exceed a local threshold.

[0044] The power electronic components include IGBTs, MOSFETs, or thyristors.

[0045] Furthermore, in the above technical solution, the intelligent measurement node is composed as follows: Figure 2 As shown, it includes:

[0046] The current measurement unit is used to measure the load current flowing through power electronic components;

[0047] A voltage measurement unit is used to simultaneously measure the on-state voltage drop and off-state transient voltage of the components;

[0048] The junction temperature estimation unit is used to back-calculate the transient junction temperature of the component in real time based on the on-state voltage drop and through a pre-calibrated voltage drop-temperature relationship curve.

[0049] Furthermore, in the above technical solution, the junction temperature estimation unit injects a constant measuring current into the power electronic component that does not affect the operation of the main circuit, detects its on-state voltage drop under this current, and back-calculates the transient junction temperature based on the pre-calibrated voltage drop-temperature relationship.

[0050] It should be further noted that, in the preferred embodiment of the present invention, the current measurement unit adopts a closed-loop Hall current sensor with a measurement range of ±10A, a bandwidth of 500kHz, and an accuracy of ±0.5%FS; the sensor is connected in series in the main current loop of the component, and the output terminal is connected to the ADC interface of the edge processing unit through a shielded twisted pair cable, with the shielding layer grounded at one end.

[0051] The voltage measurement unit adopts a voltage divider sampling circuit. The on-state voltage drop measurement branch uses a high-precision metal film resistor to form a 1:100 voltage divider ratio, and the off-state transient voltage measurement branch is connected in series with a TVS tube to achieve 60V clamping protection. The sampling signal is isolated by an isolation amplifier and then input to the edge processing unit. The isolation voltage is ≥2.5kVrms, and the measurement accuracy is ±0.2%FS.

[0052] The junction temperature estimation unit consists of a constant current source module and a high-precision operational amplifier. The injected constant measurement current is 10mA and the frequency is 1kHz (asynchronous with the main circuit switching frequency to avoid interference). The output of the constant current source is connected to the control electrode and emitter (IGBT / MOSFET) or anode and cathode (thyristor) of the component through a dedicated test pin. The on-state voltage drop detection terminal shares the isolation amplifier channel with the voltage measurement unit.

[0053] An edge processing unit, connected to the current measurement unit, voltage measurement unit, and junction temperature estimation unit respectively, is used to perform the synchronous sampling and calculation to obtain the local characteristic parameters including peak current, current change rate, on-state voltage drop, voltage change rate, instantaneous power consumption, and transient junction temperature.

[0054] Furthermore, in the above technical solution, the edge processing unit also includes a self-calibration module and a self-diagnosis module. The self-calibration module is used to periodically perform zero-point and gain calibration on the current measurement unit and the voltage measurement unit; the self-diagnosis module is used to determine the working status of the intelligent measurement node itself by analyzing the rationality of the measurement data.

[0055] It should be further noted that, in the preferred embodiment of the present invention, the edge processing unit is a 32-bit industrial-grade MCU with a main frequency of 480MHz and a built-in 24-bit ADC (sampling rate of 5MSps); the MCU expands Flash for local data caching through the SPI interface and connects to status indicator lights (normal green light, abnormal red light) through the GPIO interface.

[0056] Furthermore, synchronous sampling is implemented as follows: the edge processing unit triggers ADC synchronous sampling through an internal timer, and the sampling time deviation of current, voltage, junction temperature and related signals is ≤50ns; the sampling rate is configured to 200kHz (which can be adjusted by software, ranging from 50kHz to 1MHz), and every 100 sampling points form a data frame, with the frame header containing a timestamp (calibrated by the system synchronization clock, with an accuracy of ±1μs).

[0057] The logic for calculating local feature parameters is as follows:

[0058] Peak current: The maximum value in each frame of data, updated using a sliding window method (window size 10ms);

[0059] Current change rate (di / dt): The average value of the absolute value of the current difference between adjacent sampling points divided by the sampling period (5μs);

[0060] On-state voltage drop: The effective value after removing transient spikes when the junction temperature estimation unit injects a 10mA current.

[0061] Voltage change rate (dv / dt): The difference between adjacent sampling points of the turn-off transient voltage divided by the sampling period, taking the maximum value;

[0062] Instantaneous power consumption: the product of real-time current and on-state voltage drop, calculated at the sampling point level and then averaged for each frame;

[0063] Transient junction temperature: Based on the pre-calibrated "voltage drop-temperature" relationship curve, the curve fitting formula is Tj=a×V. ce +b, where a = 8.5℃ / V, b = -20℃, V ce The on-state pressure drop is used, with a fitting error of ±1.5℃.

[0064] Furthermore, taking IGBT as an example, the preset thresholds for local characteristic parameters are as follows: peak current threshold = 11A, current change rate threshold = 50A / μs, on-state voltage drop threshold = 3.5V, voltage change rate threshold = 100V / μs, instantaneous power consumption threshold = 200W, and transient junction temperature threshold = 125℃.

[0065] A high-speed data aggregation layer is used to interconnect the multiple intelligent measurement nodes in a deterministic communication network topology, enabling time alignment and aggregation of local feature parameters of each node for uploading.

[0066] Furthermore, in the above technical solution, the high-speed data aggregation layer adopts a topology based on time-sensitive networking or daisy-chain fiber optic Ethernet, and the intelligent measurement node is embedded in the topology as a data acquisition terminal.

[0067] It should be further noted that, in the preferred embodiment of this invention, the high-speed data aggregation layer adopts a Time-Sensitive Network (TSN) topology, with the following specific configuration:

[0068] Core equipment: TSN switch, supporting 8 Ethernet ports, 1 optical port, and a transmission rate of 1Gbps;

[0069] Transmission medium: single-mode optical fiber, core diameter 9μm, outer sheath diameter 125μm, maximum transmission distance 2km; intelligent measurement nodes are connected to the optical fiber via SFP optical modules, with optical module transmit power -5~0dBm and receive sensitivity ≤-15dBm;

[0070] Topology: A hybrid star and daisy-chain topology is adopted. Each TSN switch can connect up to 32 smart measurement nodes. The fiber optic length between nodes is ≤50m. Switch cascading supports up to 4 levels of expansion (total number of nodes ≤128).

[0071] Furthermore, time alignment is achieved using the IEEE 802.1AS-2020 protocol, with the switch acting as the time master and the intelligent measurement node acting as the slave; the synchronization period is 10ms, and the time synchronization accuracy is ±1μs.

[0072] The data aggregation and uploading process is based on IEEE 802.1Qbv (time-aware scheduling), allocating dedicated time slices for measurement data (each frame transmission time ≤ 50μs) to avoid data conflicts; the data frame format conforms to IEEE 802.3, with the frame header containing node ID, timestamp, and data type identifier, and the payload containing local characteristic parameters; the data aggregation process is as follows:

[0073] 1. The intelligent measurement node generates data frames at a sampling rate of 200kHz, and each 100 frames (500μs) is packaged into a transmission packet;

[0074] 2. The transmission packets are uploaded to the switch via the TSN network. The switch sorts the data of each node by timestamp to achieve time alignment.

[0075] 3. The aligned data is grouped by "node ID + time slice" and uploaded to the collaborative analysis and evaluation layer via optical port, with a transmission delay of ≤1ms.

[0076] It should be further noted that, in the preferred technical solution of the present invention, key nodes (such as measurement nodes close to the core power supply components) adopt dual fiber optic link backup, and the link switching time is ≤100μs;

[0077] TSN switches are equipped with dual power supplies, with a power switching time of ≤5ms, to avoid data loss due to power outages.

[0078] The collaborative analysis and evaluation layer, connected to the high-speed data aggregation layer, is used to receive and fuse local feature parameters from all intelligent measurement nodes, perform component-level electro-thermal-aging collaborative analysis, and output health status assessment results and predictive maintenance information.

[0079] The collaborative analysis and evaluation layer includes:

[0080] The condition assessment and diagnosis unit, implemented based on software algorithms, is used to run the electrothermal coupling model and the lifetime prediction model. Based on the local characteristic parameters, it calculates the real-time junction temperature, junction temperature fluctuation, power loss, cumulative aging damage and remaining useful life of each component, and performs health status classification and fault warning based on preset rules.

[0081] It should be further noted that, in the preferred embodiment of this invention, the collaborative analysis and evaluation layer hardware consists of "main controller + FPGA + storage module + cloud interface module":

[0082] Main controller: Industrial-grade CPU, 2.5GHz, 16GB DDR4 memory, 512GB SSD (for system software and model storage).

[0083] FPGA: Field Programmable Gate Array, with 325k logic cells and 202 built-in DSP slices for real-time execution of protection logic;

[0084] Storage module: Industrial-grade SATA hard drive, 1TB capacity, used for diagnostic case library storage, supports RAID 1 redundancy backup;

[0085] Cloud interface module: 4G / 5G industrial module, supporting TD-LTE / FDD-LTE / 5G NR networks, with an upload speed of ≥10Mbps; or Ethernet interface (RJ45, 1Gbps), supporting wired network connection.

[0086] Furthermore, in the above technical solution, the condition assessment and diagnosis unit is used for:

[0087] Based on the instantaneous power consumption and transient junction temperature in the local characteristic parameters, the junction temperature fluctuation spectrum and thermal stress are calculated through the electrothermal coupling model;

[0088] Based on the thermal stress, the cumulative lifetime damage is calculated using the lifetime prediction model, and the remaining useful lifetime is estimated.

[0089] Based on long-term historical data of the local characteristic parameters, the performance degradation trend of the components is analyzed.

[0090] It should be further noted that the electrothermal coupling model uses an RC thermal network model to describe the heat transfer process of the components, and the model formula is as follows:

[0091] Thermal resistance network equation: T j =P loss ×(R th(j−c) +R th(c−s) +R th(s−a) )+T a

[0092] Wherein: T j For junction temperature (°C), P loss R is the instantaneous power consumption (W). th(j−c) R is the junction-to-shell thermal resistance (default 0.5℃ / W). th(c−s) The thermal resistance between the casing and the heatsink (default 0.2℃ / W), Rth(s−a) For heatsink-ambient thermal resistance (default 0.3℃ / W), T a Ambient temperature (°C, measured by an external temperature sensor, accuracy ±0.5°C);

[0093] The junction temperature fluctuation spectrum was analyzed using Fast Fourier Transform (FFT) on the junction temperature time series (sampling interval 5ms) with a frequency resolution of 0.1Hz. The spectral peaks in the 0~10Hz frequency band were extracted as the basis for thermal stress assessment.

[0094] Thermal stress calculation: σ = k × ΔT j ×α, where k is the material coefficient (2.3 for IGBT), ΔT j The value represents the junction temperature fluctuation amplitude (°C), and α is the coefficient of thermal expansion (2.6 × 10⁻⁶ for silicon). -6 / ℃).

[0095] The lifetime prediction model uses rainflow counting combined with the Arrhenius lifetime model. The specific steps are as follows:

[0096] Junction temperature cycle extraction: Rainflow counting is performed on the junction temperature time series, and the amplitude ΔT and mean T of each temperature cycle are calculated. avg ;

[0097] Single-cycle lifetime calculation:

[0098] Where: L0 is the baseline lifetime (default 1×10). 6 (second cycle), E a To determine the activation energy (0.7 eV for IGBT), k B Boltzmann constant (1.38 × 10⁻⁶) -23 J / K), T0 is the reference temperature (125℃), and n is the temperature cycle index (default 5);

[0099] Cumulative aging damage: , where N is the number of temperature cycles counted;

[0100] Remaining useful life (RUL): RUL = (1 − D) × L total / P avg L total Total component lifespan (default 2×10) 6 (nth cycle), P avg This represents the average number of daily temperature cycles (obtained from historical data).

[0101] Furthermore, in the above technical solution, when the performance degradation trend indicates that the linear rate of increase of the on-state saturation voltage drop of the component continuously exceeds a specified proportion of its initial value, or when the junction-shell thermal resistance estimated based on the electrothermal model shows a continuous increasing trend, an early warning message indicating aging of the bonding line or substrate solder layer is generated.

[0102] It should be further explained that the performance degradation trend analysis and early warning rules are as follows:

[0103] Data statistics period: Long-term historical data is stored in "daily / weekly / monthly" hierarchical format, with a statistical window of 30 days;

[0104] Early warning of degradation of on-state saturation pressure drop: Calculate the slope k of the linear fit of the monthly on-state saturation pressure drop. V If k V >5%×V ce0 / 1000h (V ce0 The initial on-state voltage drop is set to 2.0V, and this continues for 3 months, triggering a bond wire aging warning.

[0105] Early warning of thermal resistance degradation in the crust: Inverse calculation of R using an electrothermal model th(j−c) (R) th(j−c) =(T j−Tc ) / P loss T c (The shell temperature is measured by the shell temperature sensor). If R th(j−c) A monthly growth rate greater than 3% for two consecutive months triggers an aging warning for the substrate solder layer.

[0106] Warning levels: Mild warning (initial signs of degradation), Moderate warning (accelerated degradation rate), Severe warning (remaining lifespan <20%), corresponding to different maintenance prompts.

[0107] Furthermore, the health status grading standards are as follows:

[0108] When RUL>80% and D<0.2, it is grade A (excellent): the key parameters show no degradation trend, and the maintenance recommendation is normal monitoring;

[0109] When RUL∈[50%, 80%) or D∈[0.2, 0.5%), it is grade B (good): the key parameter status is slightly degraded, and the maintenance recommendation is to check regularly;

[0110] When RUL∈[20%, 50%) or D∈[0.5, 0.8%), it is Level C (Warning): the key parameter status is significantly degraded, and the maintenance recommendation is planned replacement;

[0111] When RUL < 20% or D > 0.8, it is classified as Level D (urgent): the critical parameter status is close to failure, and the maintenance recommendation is immediate replacement.

[0112] The data recording and case library management unit is used to continuously record system operation data, and when the primary anomaly flag is triggered or the health status level reaches the warning level, it automatically saves the original sampling data and local feature parameters of all relevant intelligent measurement nodes within 30 seconds before and after the event, forming a traceable diagnostic case library.

[0113] Furthermore, in the above technical solution, the data recording and case library management unit is configured to: store the original sampling data, local feature parameters and system operating condition snapshots in a standardized format; and store the diagnostic case library according to event type and component type, supporting case retrieval and optimized calling of life prediction models.

[0114] It should be further explained that the optimization process for the lifetime prediction model is as follows:

[0115] For every 100 valid cases (cases with confirmed causes of failure), model optimization is triggered;

[0116] Extract feature parameters and lifetime data from the case study, and use gradient descent to update the lifetime prediction model's E. a Parameters such as n;

[0117] The optimized model replaces the original model while retaining historical versions (up to 5 versions), and supports rollback.

[0118] Furthermore, in the above technical solution, the collaborative analysis and evaluation layer also includes a protection logic execution unit, and the protection logic execution process is as follows: Figure 3 As shown, it is implemented by a field-programmable gate array (FPGA) and is used to receive the health status assessment result or the primary anomaly identifier, and generate system-level protection control commands according to preset logic.

[0119] It should be further noted that, in the preferred embodiment of this invention, the field-programmable gate array is programmed using Verilog HDL, and the core logic includes:

[0120] Signal receiving module: Receives the health status assessment results of the main controller and the primary anomaly flags of the intelligent measurement nodes. The interface is AXI4-Stream with a data width of 32 bits.

[0121] Priority judgment module: Set trigger priority (primary anomaly flag > D-level warning > C-level warning > B-level warning > A-level normal), high-priority signals interrupt low-priority execution;

[0122] Protection command generation module: Generates corresponding commands based on the trigger signal. The specific commands are as follows:

[0123] Primary anomaly indicator: Generates an emergency stop command, cuts off the main circuit power supply, disconnects the component drive signal, and triggers an audible and visual alarm (buzzer + red warning light).

[0124] Level D warning: Generates bypass + shutdown command to bypass the faulty component (by switching via relay), cuts off power after 30 seconds, and prompts for immediate replacement;

[0125] Level C warning: Generates a derating command to reduce the power supply output power to 70% of the rated value, limit the maximum current to 8A, and continuously monitor the degradation trend;

[0126] Level B warning: Generates an alarm prompt command, illuminates a yellow warning light, and displays a maintenance prompt on the monitoring interface, without affecting normal operation.

[0127] Furthermore, in a preferred embodiment of the present invention, the field-programmable gate array has built-in dual-core redundancy, with the two cores executing protection logic synchronously, and the protection command is triggered only when the outputs of the two cores are consistent.

[0128] The protection command output terminal is equipped with a hardware latching circuit to prevent accidental triggering caused by momentary interference. The latching state needs to be manually reset (or remotely reset via the cloud platform).

[0129] Furthermore, in the above technical solution, the collaborative analysis and evaluation layer also includes a cloud platform interface, which is used to upload the health status evaluation and prediction results and the condensed feature data to the cloud server.

[0130] It should be further noted that, in the preferred embodiment of this invention, the cloud platform interface is implemented as follows:

[0131] The preferred communication protocol is MQTT v3.1.1 (lightweight and low bandwidth consumption), with HTTP / HTTPS as an alternative.

[0132] MQTT configuration: The Broker address is the public IP address of the cloud server, the Client ID is a unique system identifier, and username / password authentication is used (encrypted transmission).

[0133] Upload cycle:

[0134] Normal status (Level A / Level B): Upload condensed feature data (hourly statistics) every 5 minutes;

[0135] Warning status (Level C): Feature data is uploaded once every 1 minute, and raw data fragments (10 seconds in length) are uploaded once every 10 minutes.

[0136] Emergency Status (Level D / Basic Anomaly): Upload all data in real time (delay ≤ 1s) until the fault is resolved.

[0137] The cloud data interaction content configuration is as follows:

[0138] Uploaded data: health status classification results, remaining life prediction values, cumulative damage values, key characteristic parameter statistics (maximum / minimum / average values), operating condition snapshots, and fault warning indicators;

[0139] Data received: remote calibration parameters (such as threshold adjustment), model update instructions, maintenance plan notifications, and remote reset instructions;

[0140] Interface Specifications: Designed using RESTful API, data format is JSON, encoding is UTF-8, and supports breakpoint resume (for large volumes of raw data).

[0141] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An integrated intelligent monitoring and protection system for power electronic components in a medium-frequency power supply, characterized in that, include: The distributed intelligent measurement layer includes multiple intelligent measurement nodes that are set up one-to-one with each power electronic component. These nodes are used to synchronously and frequently sample the multi-dimensional electrical operating parameters of the corresponding components, calculate local characteristic parameters based on the sampled data, and generate a primary anomaly flag when the local characteristic parameters exceed a local threshold. A high-speed data aggregation layer is used to interconnect the multiple intelligent measurement nodes in a deterministic communication network topology, enabling time alignment and aggregation of local feature parameters of each node for uploading. The collaborative analysis and evaluation layer, connected to the high-speed data aggregation layer, is used to receive and fuse local feature parameters from all intelligent measurement nodes, perform component-level electro-thermal-aging collaborative analysis, and output health status assessment results and predictive maintenance information. The collaborative analysis and evaluation layer includes: The condition assessment and diagnosis unit, implemented based on software algorithms, is used to run the electrothermal coupling model and the lifetime prediction model. Based on the local characteristic parameters, it calculates the real-time junction temperature, junction temperature fluctuation, power loss, cumulative aging damage and remaining useful life of each component, and performs health status classification and fault warning based on preset rules. The data recording and case library management unit is used to continuously record system operation data, and when the primary anomaly flag is triggered or the health status level reaches the warning level, it automatically saves the original sampling data and local feature parameters of all relevant intelligent measurement nodes within 30 seconds before and after the event, forming a traceable diagnostic case library. The power electronic components include IGBTs, MOSFETs, or thyristors.

2. The integrated medium-frequency power supply power electronic component intelligent monitoring and protection system according to claim 1, characterized in that: The intelligent measurement node includes: The current measurement unit is used to measure the load current flowing through power electronic components; A voltage measurement unit is used to simultaneously measure the on-state voltage drop and off-state transient voltage of the components; The junction temperature estimation unit is used to back-calculate the transient junction temperature of the component in real time based on the on-state voltage drop and through a pre-calibrated voltage drop-temperature relationship curve. An edge processing unit, connected to the current measurement unit, voltage measurement unit, and junction temperature estimation unit respectively, is used to perform the synchronous sampling and calculation to obtain the local characteristic parameters including peak current, current change rate, on-state voltage drop, voltage change rate, instantaneous power consumption, and transient junction temperature.

3. The integrated medium-frequency power supply power electronic component intelligent monitoring and protection system according to claim 2, characterized in that: The junction temperature estimation unit injects a constant measuring current into the power electronic component that does not affect the operation of the main circuit, detects the on-state voltage drop under this current, and back-calculates the transient junction temperature based on the pre-calibrated voltage drop-temperature relationship.

4. The integrated medium-frequency power supply power electronic component intelligent monitoring and protection system according to claim 1, characterized in that: The high-speed data aggregation layer adopts a topology based on time-sensitive networking or daisy-chain fiber optic Ethernet, and the intelligent measurement node is embedded in the topology as a data acquisition terminal.

5. The integrated medium-frequency power supply power electronic component intelligent monitoring and protection system according to claim 1, characterized in that: The collaborative analysis and evaluation layer also includes a protection logic execution unit, implemented by a field-programmable gate array, used to receive the health status evaluation result or the primary anomaly identifier, and generate system-level protection control instructions according to preset logic.

6. The integrated medium-frequency power supply power electronic component intelligent monitoring and protection system according to claim 1, characterized in that: The condition assessment and diagnostic unit is used for: Based on the instantaneous power consumption and transient junction temperature in the local characteristic parameters, the junction temperature fluctuation spectrum and thermal stress are calculated through the electrothermal coupling model; Based on the thermal stress, the cumulative lifetime damage is calculated using the lifetime prediction model, and the remaining useful lifetime is estimated. Based on long-term historical data of the local characteristic parameters, the performance degradation trend of the components is analyzed.

7. The integrated medium-frequency power supply power electronic component intelligent monitoring and protection system according to claim 6, characterized in that: When the performance degradation trend indicates that the linear rate of increase of the on-state saturation voltage drop of the component continues to exceed a specified proportion of its initial value, or the junction-shell thermal resistance estimated based on the electrothermal model shows a continuous increasing trend, an early warning message indicating aging of the bonding wire or substrate solder layer is generated.

8. The integrated medium-frequency power supply power electronic component intelligent monitoring and protection system according to claim 1, characterized in that: The data recording and case library management unit is configured to store the original sampling data, local feature parameters, and system operating condition snapshots in a standardized format. The diagnostic case library is classified and stored according to event type and component type, and supports case retrieval and optimized invocation of life prediction models.

9. The integrated medium-frequency power supply power electronic component intelligent monitoring and protection system according to claim 1, characterized in that: The collaborative analysis and evaluation layer also includes a cloud platform interface for uploading the health status evaluation and prediction results and condensed feature data to a cloud server.

10. The integrated medium-frequency power supply power electronic component intelligent monitoring and protection system according to claim 2, characterized in that: The edge processing unit also includes a self-calibration module and a self-diagnosis module. The self-calibration module is used to periodically perform zero-point and gain calibration on the current measurement unit and the voltage measurement unit. The self-diagnosis module is used to determine the working status of the intelligent measurement node by analyzing the rationality of the measurement data.

Images