A super-calculating high-power silicon carbide power supply load optimization system and method

Abstract

The application discloses a kind of super-calculating high-power silicon carbide power load optimization system and method, to improve breakdown voltage and power density by using 4H-SiC material to make high-performance Schottky diode (SBD) and junction barrier Schottky diode (JBS), and apply Resurf theory and JTE technology optimization device structure.The system builds the power load optimization system containing multiple silicon carbide converters, flexibly adjusts output power, adapts to complex requirements, and improves efficiency.Introduce machine learning algorithm to predict load demand, adjust power parameters in advance, and maintain efficient operation.Install real-time monitoring unit to continuously monitor power state and load condition, ensure optimal operation and respond to abnormalities in time.Establish dynamic adjustment mechanism to quickly adjust power output characteristics to adapt to actual load changes.Embed automatic fault diagnosis and self-healing function to quickly restore normal operation and enhance reliability.Implement load optimization strategy to stabilize DC bus voltage and use MPPT technology to improve overall efficiency.

Description

Technical Field

[0001] This invention relates to the field of power management technology, and in particular to a supercomputing high-power silicon carbide power supply load optimization system and method. Background Technology

[0002] As the scale and performance requirements of supercomputing systems continue to increase, their power management systems face new challenges. Modern supercomputing systems not only need to process massive amounts of data and complex computational tasks, but also need to maintain efficient and stable operation in high-temperature and high-power environments. Traditional silicon-based power management systems are proving inadequate in meeting these demands.

[0003] Traditional power management systems perform poorly in high-temperature environments, leading to performance degradation and shortened lifespan. The high power consumption of supercomputing systems exacerbates heat dissipation problems, making it difficult for traditional power supplies to maintain efficient operation. Furthermore, traditional systems lack intelligent predictive and response mechanisms, failing to quickly adjust output parameters when the load changes drastically, easily causing voltage fluctuations or even system crashes. Simultaneously, traditional power management systems fail to effectively utilize excess energy when the load decreases, resulting in energy waste and increased maintenance costs. Due to the lack of automatic fault diagnosis and self-healing capabilities, traditional systems rely on manual intervention when faults occur, increasing downtime and maintenance complexity.

[0004] Traditional power management systems are no longer adequate for the requirements of modern supercomputing systems. A new solution is urgently needed to overcome the limitations of existing technologies and ensure that supercomputing systems operate stably and efficiently in complex and ever-changing working environments. Summary of the Invention

[0005] A supercomputing high-power silicon carbide power supply load optimization system and method includes the following steps;

[0006] S1. Silicon Carbide Power Device Design: Schottky diodes (SBD) and junction barrier Schottky diodes (JBS) are fabricated using 4H-SiC material; the device structure is optimized by applying the Resurf theory and junction termination extension (JTE) technology to improve the breakdown voltage and power density of the devices;

[0007] S2. Power Load Optimization System Construction: Design and construct a power load optimization system comprising multiple silicon carbide power converters. Each converter can obtain electrical energy from its corresponding DC power source and convert it into output power. The multi-stage converters are combined in series or parallel to form a power supply path that can flexibly adjust the output power.

[0008] S3. Intelligent Prediction: Introducing machine learning algorithms to predict load demand over a future period, and adjusting power output parameters in advance based on the prediction results so that the entire system operates at its optimal operating point;

[0009] S4. Real-time monitoring unit configuration: Install a real-time monitoring unit to continuously monitor power status and load conditions, ensuring that the system always operates at the optimal operating point and can respond immediately when any abnormal conditions are detected;

[0010] S5. Fault diagnosis and self-healing function integration: Embedded automatic fault diagnosis and self-healing function, automatically detects and repairs hardware or software faults that may affect system stability;

[0011] S6. Load optimization method: Real-time monitoring of the output voltage and current of each stage of silicon carbide power converter, and dynamic adjustment of the output power of the converter according to the load demand; adjustment of the output voltage of other converters to stabilize the DC bus voltage and prevent system collapse; use maximum power point tracking (MPPT) technology to make each stage of converter operate at its maximum power point as much as possible.

[0012] Furthermore, a high-power silicon carbide power supply load optimization system and method for supercomputing are provided.

[0013] In step S1, 4H-SiC material is used to fabricate Schottky diodes (SBD) and junction barrier Schottky diodes (JBS); the device structure is optimized by applying the Resurf theory and the junction termination extension (JTE) technique to improve the breakdown voltage and power density of the device. The specific steps are as follows.

[0014] S11. Material selection and preparation: 4H polytype silicon carbide (4H-SiC) was selected as the base material. Before the start of manufacturing, high-quality crystal growth and substrate treatment were performed on 4H-SiC to ensure that it has excellent electrical properties.

[0015] S12. Device Structure Design: Based on the Resurf theory, design the structures of Schottky diodes and junction barrier Schottky diodes, optimizing their breakdown voltage and power density. Determine the optimal doping concentration, layer thickness, and other key parameters to ensure stable device performance under high temperature and high power environments.

[0016] S13. Junction Termination Extension Technology Application (JTE): By adopting junction termination extension technology, the length of the drift region is increased or special-shaped termination structures such as ring or multi-finger terminals are introduced to further improve the breakdown voltage and reduce the electric field concentration in the edge region.

[0017] S14. Process Flow Optimization: Develop and implement a series of sophisticated semiconductor manufacturing processes, including photolithography, etching, deposition, and diffusion;

[0018] S15. Performance Testing and Verification: After the device manufacturing is completed, comprehensive electrical and thermal tests are conducted to verify whether the power device meets the design goals, especially whether it can provide the required stability and performance under high temperature and high power conditions.

[0019] Furthermore, a high-power silicon carbide power supply load optimization system and method for supercomputing are provided.

[0020] Step S2 involves designing and constructing a power load optimization system comprising multiple silicon carbide power converters. Each converter can obtain electrical energy from its corresponding DC power source and convert it into output power. The multi-stage converters are then connected in series or parallel to form a power supply path that can flexibly adjust the output power. The specific steps are as follows:

[0021] S21. System Architecture Planning: Define the overall architecture of the power load optimization system, including determining the number, type (e.g., DC-DC converter) and topology (e.g., buck, boost, or buck-boost type) of the required silicon carbide power converters.

[0022] S22. Converter Design and Selection: Select appropriate 4H-SiC material Schottky diodes (SBD) and junction barrier Schottky diodes (JBS) for each power converter, and design specific converter parameters, such as switching frequency, efficiency target and thermal management strategy;

[0023] S23. Topology configuration: Based on the load demand analysis of the supercomputing system, considering factors such as redundancy, reliability and flexibility, determine whether to use series, parallel or hybrid connection to combine multiple silicon carbide power converters.

[0024] S24. Control Algorithm Development: Develop advanced control algorithms to manage and optimize energy flow in the power supply chain. Monitor the operating status of each converter in real time and dynamically adjust its operating point according to the actual load to achieve efficient power transmission and minimize energy loss;

[0025] S25. System Integration and Testing: After completing the hardware assembly, integrate all components into a complete power load optimization system and perform performance tests including functional verification, stability assessment, and efficiency measurement under different operating conditions. Conduct a comprehensive check on the electromagnetic compatibility (EMC) and thermal characteristics of the entire system.

[0026] S26. Implementation, Deployment, and Maintenance: Install the power load optimization system into the supercomputing environment and put it into use. Establish a comprehensive monitoring and maintenance plan, regularly collect operational data for performance analysis, and promptly address any potential problems.

[0027] Furthermore, a high-power silicon carbide power supply load optimization system and method for supercomputing are provided.

[0028] In step S3, a machine learning algorithm is introduced to predict the load demand in the future. Based on the prediction results, the power supply output parameters are adjusted in advance so that the entire system operates at its optimal operating point. The specific steps are as follows.

[0029] S31. Data Collection and Preprocessing: Collect historical load data, operational status information, and external influencing factors (such as job scheduling patterns and time cycles) of the supercomputing system. Clean and normalize the raw data, and convert it into a format suitable for machine learning algorithms.

[0030] S32. Feature Engineering: Extract key feature variables that reflect load trends from existing datasets, such as average load rate, peak load interval, and task submission frequency. Construct new features, such as time lag terms or periodic components based on time series analysis, to enhance the model's predictive ability;

[0031] S33. Model Selection and Training: Choose an appropriate machine learning algorithm to build the predictive model, such as Support Vector Machine (SVM), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), or Deep Neural Network (DNN). Train the selected model using historical data and optimize the model parameters through cross-validation to ensure good generalization performance.

[0032] S34. Evaluation of Prediction Results: The model's prediction accuracy is evaluated using the test set or reserved data. Commonly used metrics include mean squared error (MSE), mean absolute error (MAE), and coefficient of determination (R²). Features are adjusted multiple times or other model types are tried until the most suitable prediction scheme is found.

[0033] Furthermore, a high-power silicon carbide power supply load optimization system and method for supercomputing are provided.

[0034] In step S4, a real-time monitoring unit is installed to continuously monitor the power supply status and load conditions, ensuring that the system always operates at the optimal operating point and can respond immediately when any abnormal conditions are detected. The specific steps are as follows;

[0035] S41. Determine monitoring needs and objectives: Identify the parameters that need to be monitored, such as power supply voltage, current, temperature, humidity, power factor, load rate, etc. Define the performance indicators of the monitoring system, such as sampling frequency, response time, accuracy requirements, and alarm thresholds;

[0036] S42. Select sensors and measurement equipment: Select high-precision sensors and advanced measurement equipment suitable for use in the supercomputing environment. For different physical quantities (such as electrical parameters and thermal parameters), select the corresponding dedicated sensors.

[0037] S43. Design the monitoring network architecture: Construct an efficient data transmission network to connect various sensors and the central monitoring platform. Select wired or wireless communication technologies such as Industrial Ethernet, RS485 bus, ZigBee, or Wi-Fi;

[0038] S44. Develop a monitoring software platform: Develop a user-friendly monitoring software that provides an intuitive interface for operators to view real-time data and historical records; the software includes a data analysis module that can automatically identify anomalies and trigger alarms; and integrates intelligent diagnostic functions to help technicians quickly locate the root cause of problems;

[0039] S45. Configure fault detection and response mechanism: Set reasonable alarm rules and emergency handling procedures, and immediately activate corresponding response measures when any parameter deviating from the normal range is detected, including automatically adjusting power output settings, notifying the maintenance team, and performing protective power cut-off.

[0040] Furthermore, a high-power silicon carbide power supply load optimization system and method for supercomputing are provided.

[0041] Step S5 incorporates automatic fault diagnosis and self-healing functions, which automatically detect and repair hardware or software faults that may affect system stability. The specific steps are as follows:

[0042] S51. Construct a fault monitoring network: Deploy comprehensive sensors and monitoring nodes in the supercomputing high-power silicon carbide power supply load optimization system to collect hardware status (such as temperature, voltage, and current) and software operation data (such as logs and performance indicators) in real time. Ensure coverage of all critical components and transmit data to the central monitoring platform via a high-speed communication network. Establish an anomaly detection mechanism to quickly identify potential problems;

[0043] S52. Develop intelligent diagnostic algorithms: Utilize machine learning and data analysis techniques to develop an intelligent fault diagnosis system that automatically analyzes data streams, identifies abnormal behavior patterns, and accurately determines the fault type and its location. Supervised and unsupervised learning methods are used to train the model, capturing known and unknown fault patterns.

[0044] S53. Implement self-healing strategies and automated repair: Develop self-healing strategies for different fault types, such as automatic restart, parameter reconfiguration, and switching to backup paths. For software faults, achieve self-repair through firmware updates and vulnerability fixes; for hardware faults, design redundant architectures to isolate and replace faulty components. Introduce automated recovery processes to reduce downtime and manual intervention.

[0045] S54. Testing, Verification, and Continuous Improvement: Conduct comprehensive testing, ranging from unit testing to system cascading testing, simulating various failure scenarios to evaluate response speed, repair success rate, and the ability to restore normal operation. Based on test results, continuously optimize diagnostic algorithms and self-healing strategies to ensure stable and reliable system operation.

[0046] Furthermore, a high-power silicon carbide power supply load optimization system and method for supercomputing are provided.

[0047] In step S6, the output voltage and current of each stage of the silicon carbide power converter are monitored in real time, and the output power of the converter is dynamically adjusted according to the load demand; the output voltage of other converters is adjusted to stabilize the DC bus voltage and prevent system collapse; maximum power point tracking (MPPT) technology is used to make each stage of the converter operate at its maximum power point as much as possible. The specific steps are as follows.

[0048] S61. Real-time Monitoring and Data Analysis: Establish a real-time monitoring system to continuously collect output voltage, current, and other key parameters (such as temperature and power factor) of each stage of the silicon carbide power converter. Utilize advanced data processing technologies and algorithms to quickly analyze the collected data to identify current load patterns and trends;

[0049] S62. Dynamic Power Adjustment Strategy: Based on real-time monitoring information, an intelligent control algorithm is developed to dynamically adjust the output power of each converter, ensuring its consistently efficient operation. When the DC power supply corresponding to a certain converter decreases, the system quickly responds and adjusts the output voltage of other converters to stabilize the DC bus voltage and prevent system performance degradation or collapse.

[0050] S63. DC Bus Voltage Stabilization Mechanism: The control system is designed to automatically calculate the optimal compensation scheme, determine which converters should increase or decrease their output, and maintain DC bus voltage stability. A buffer energy storage device (such as a supercapacitor or battery) is introduced to absorb or supplement instantaneous energy fluctuations, enhancing system stability.

[0051] S64. Maximum Power Point Tracking (MPPT) Technology Application: Integrating the MPPT algorithm ensures that each stage of the converter operates at its maximum power point as much as possible. The operating point is dynamically adjusted based on the input source characteristics, improving efficiency, reducing energy loss, and optimizing the overall energy efficiency of the supercomputing system.

[0052] A supercomputing high-power silicon carbide power supply load optimization system and method are disclosed. The supercomputing high-power silicon carbide power supply load optimization system and method are used to implement any of the aforementioned supercomputing high-power silicon carbide power supply load optimization systems and methods. The supercomputing high-power silicon carbide power supply load optimization system and method include: a silicon carbide power device design module, a power supply load optimization system construction module, an intelligent prediction module, a real-time monitoring unit configuration module, a fault diagnosis and self-healing function integration module, and a load optimization method module.

[0053] The silicon carbide power device design module uses 4H-SiC material to fabricate high-performance Schottky diodes (SBDs) and junction barrier Schottky diodes (JBSs), and optimizes the structure through Resurf theory and JTE technology to improve breakdown voltage and power density, ensuring stability under high temperature and high power environments.

[0054] Power Load Optimization System Building Module: This module designs and builds a system containing multiple silicon carbide power converters, each drawing power from a DC power source and converting it into output power. By combining them in series or parallel, a power supply path is formed that flexibly adjusts the output power to adapt to complex needs and improve efficiency.

[0055] Intelligent prediction module: Introduces machine learning algorithms to predict future load demand and adjust power output parameters in advance, so that the system always operates at its optimal operating point and maintains high efficiency.

[0056] Real-time monitoring unit configuration module: Install a real-time monitoring unit to continuously monitor the power status and load conditions, ensuring that the system always operates at the optimal operating point and can respond immediately when abnormal conditions are detected, ensuring safety and reliability.

[0057] Fault diagnosis and self-healing function integrated module: Embedded automatic fault diagnosis and self-healing function, automatically detects and repairs hardware or software faults that may affect system stability, quickly restores normal operation, and enhances continuous reliability.

[0058] Load optimization module: Real-time monitoring of the output voltage and current of each stage of silicon carbide power converter, dynamic adjustment of the converter's output power according to load demand, stabilization of DC bus voltage, and adoption of maximum power point tracking (MPPT) technology to improve the overall efficiency of the system.

[0059] The beneficial effects of this invention are:

[0060] Improve system stability: Ensure the stability of DC bus voltage and avoid system crashes by real-time monitoring and adjustment of the output power of each stage of the converter;

[0061] Improve system efficiency: Employ maximum power point tracking (MPPT) technology to ensure that each stage of the converter operates at its maximum power point, thereby improving the overall system efficiency;

[0062] Reduce system costs: By optimizing the structure and process of silicon carbide power devices, the manufacturing cost of the devices can be reduced, thereby reducing the overall system cost;

[0063] By using silicon carbide materials to fabricate power devices and combining multi-stage converter design and load optimization methods, efficient management and optimization of high-power power loads for supercomputing have been achieved. Attached Figure Description

[0064] Figure 1 This is a flowchart illustrating a method for optimizing the power supply load of a high-power silicon carbide power supply in a supercomputing system. Detailed Implementation

[0065] A supercomputing high-power silicon carbide power supply load optimization system and method includes the following steps;

[0066] S1. Silicon Carbide Power Device Design: Schottky diodes (SBD) and junction barrier Schottky diodes (JBS) are fabricated using 4H-SiC material; the device structure is optimized by applying the Resurf theory and junction termination extension (JTE) technology to improve the breakdown voltage and power density of the devices;

[0067] S2. Power Load Optimization System Construction: Design and construct a power load optimization system comprising multiple silicon carbide power converters. Each converter can obtain electrical energy from its corresponding DC power source and convert it into output power. The multi-stage converters are combined in series or parallel to form a power supply path that can flexibly adjust the output power.

[0068] S3. Intelligent Prediction: Introducing machine learning algorithms to predict load demand over a future period, and adjusting power output parameters in advance based on the prediction results so that the entire system operates at its optimal operating point;

[0069] S4. Real-time monitoring unit configuration: Install a real-time monitoring unit to continuously monitor power status and load conditions, ensuring that the system always operates at the optimal operating point and can respond immediately when any abnormal conditions are detected;

[0070] S5. Fault diagnosis and self-healing function integration: Embedded automatic fault diagnosis and self-healing function, automatically detects and repairs hardware or software faults that may affect system stability;

[0071] S6. Load optimization method: Real-time monitoring of the output voltage and current of each stage of silicon carbide power converter, and dynamic adjustment of the output power of the converter according to the load demand; adjustment of the output voltage of other converters to stabilize the DC bus voltage and prevent system collapse; use maximum power point tracking (MPPT) technology to make each stage of converter operate at its maximum power point as much as possible.

[0072] Furthermore, a high-power silicon carbide power supply load optimization system and method for supercomputing are provided.

[0073] In step S1, 4H-SiC material is used to fabricate Schottky diodes (SBD) and junction barrier Schottky diodes (JBS); the device structure is optimized by applying the Resurf theory and the junction termination extension (JTE) technique to improve the breakdown voltage and power density of the device. The specific steps are as follows.

[0074] S11. Material selection and preparation: 4H polytype silicon carbide (4H-SiC) was selected as the base material. Before the start of manufacturing, high-quality crystal growth and substrate treatment were performed on 4H-SiC to ensure that it has excellent electrical properties.

[0075] S12. Device Structure Design: Based on the Resurf theory, design the structures of Schottky diodes and junction barrier Schottky diodes, optimizing their breakdown voltage and power density. Determine the optimal doping concentration, layer thickness, and other key parameters to ensure stable device performance under high temperature and high power environments.

[0076] S13. Junction Termination Extension Technology Application (JTE): By adopting junction termination extension technology, the length of the drift region is increased or special-shaped termination structures such as ring or multi-finger terminals are introduced to further improve the breakdown voltage and reduce the electric field concentration in the edge region.

[0077] S14. Process Flow Optimization: Develop and implement a series of sophisticated semiconductor manufacturing processes, including photolithography, etching, deposition, and diffusion;

[0078] S15. Performance Testing and Verification: After the device manufacturing is completed, comprehensive electrical and thermal tests are conducted to verify whether the power device meets the design goals, especially whether it can provide the required stability and performance under high temperature and high power conditions.

[0079] Furthermore, a high-power silicon carbide power supply load optimization system and method for supercomputing are provided.

[0080] Step S2 involves designing and constructing a power load optimization system comprising multiple silicon carbide power converters. Each converter can obtain electrical energy from its corresponding DC power source and convert it into output power. The multi-stage converters are then connected in series or parallel to form a power supply path that can flexibly adjust the output power. The specific steps are as follows:

[0081] S21. System Architecture Planning: Define the overall architecture of the power load optimization system, including determining the number, type (e.g., DC-DC converter) and topology (e.g., buck, boost, or buck-boost type) of the required silicon carbide power converters.

[0082] S22. Converter Design and Selection: Select appropriate 4H-SiC material Schottky diodes (SBD) and junction barrier Schottky diodes (JBS) for each power converter, and design specific converter parameters, such as switching frequency, efficiency target and thermal management strategy;

[0083] S23. Topology configuration: Based on the load demand analysis of the supercomputing system, considering factors such as redundancy, reliability and flexibility, determine whether to use series, parallel or hybrid connection to combine multiple silicon carbide power converters.

[0084] S24. Control Algorithm Development: Develop advanced control algorithms to manage and optimize energy flow in the power supply chain. Monitor the operating status of each converter in real time and dynamically adjust its operating point according to the actual load to achieve efficient power transmission and minimize energy loss;

[0085] S25. System Integration and Testing: After completing the hardware assembly, integrate all components into a complete power load optimization system and perform performance tests including functional verification, stability assessment, and efficiency measurement under different operating conditions. Conduct a comprehensive check on the electromagnetic compatibility (EMC) and thermal characteristics of the entire system.

[0086] S26. Implementation, Deployment, and Maintenance: Install the power load optimization system into the supercomputing environment and put it into use. Establish a comprehensive monitoring and maintenance plan, regularly collect operational data for performance analysis, and promptly address any potential problems.

[0087] Furthermore, a high-power silicon carbide power supply load optimization system and method for supercomputing are provided.

[0088] In step S3, a machine learning algorithm is introduced to predict the load demand in the future. Based on the prediction results, the power supply output parameters are adjusted in advance so that the entire system operates at its optimal operating point. The specific steps are as follows.

[0089] S31. Data Collection and Preprocessing: Collect historical load data, operational status information, and external influencing factors (such as job scheduling patterns and time cycles) of the supercomputing system. Clean and normalize the raw data, and convert it into a format suitable for machine learning algorithms.

[0090] S32. Feature Engineering: Extract key feature variables that reflect load trends from existing datasets, such as average load rate, peak load interval, and task submission frequency. Construct new features, such as time lag terms or periodic components based on time series analysis, to enhance the model's predictive ability;

[0091] S33. Model Selection and Training: Choose an appropriate machine learning algorithm to build the predictive model, such as Support Vector Machine (SVM), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), or Deep Neural Network (DNN). Train the selected model using historical data and optimize the model parameters through cross-validation to ensure good generalization performance.

[0092] S34. Evaluation of Prediction Results: The model's prediction accuracy is evaluated using the test set or reserved data. Commonly used metrics include mean squared error (MSE), mean absolute error (MAE), and coefficient of determination (R²). Features are adjusted multiple times or other model types are tried until the most suitable prediction scheme is found.

[0093] Furthermore, a high-power silicon carbide power supply load optimization system and method for supercomputing are provided.

[0094] In step S4, a real-time monitoring unit is installed to continuously monitor the power supply status and load conditions, ensuring that the system always operates at the optimal operating point and can respond immediately when any abnormal conditions are detected. The specific steps are as follows;

[0095] S41. Determine monitoring needs and objectives: Identify the parameters that need to be monitored, such as power supply voltage, current, temperature, humidity, power factor, load rate, etc. Define the performance indicators of the monitoring system, such as sampling frequency, response time, accuracy requirements, and alarm thresholds;

[0096] S42. Select sensors and measurement equipment: Select high-precision sensors and advanced measurement equipment suitable for use in the supercomputing environment. For different physical quantities (such as electrical parameters and thermal parameters), select the corresponding dedicated sensors.

[0097] S43. Design the monitoring network architecture: Construct an efficient data transmission network to connect various sensors and the central monitoring platform. Select wired or wireless communication technologies such as Industrial Ethernet, RS485 bus, ZigBee, or Wi-Fi;

[0098] S44. Develop a monitoring software platform: Develop a user-friendly monitoring software that provides an intuitive interface for operators to view real-time data and historical records; the software includes a data analysis module that can automatically identify anomalies and trigger alarms; and integrates intelligent diagnostic functions to help technicians quickly locate the root cause of problems;

[0099] S45. Configure fault detection and response mechanism: Set reasonable alarm rules and emergency handling procedures, and immediately activate corresponding response measures when any parameter deviating from the normal range is detected, including automatically adjusting power output settings, notifying the maintenance team, and performing protective power cut-off.

[0100] Furthermore, a high-power silicon carbide power supply load optimization system and method for supercomputing are provided.

[0101] Step S5 incorporates automatic fault diagnosis and self-healing functions, which automatically detect and repair hardware or software faults that may affect system stability. The specific steps are as follows:

[0102] S51. Construct a fault monitoring network: Deploy comprehensive sensors and monitoring nodes in the supercomputing high-power silicon carbide power supply load optimization system to collect hardware status (such as temperature, voltage, and current) and software operation data (such as logs and performance indicators) in real time. Ensure coverage of all critical components and transmit data to the central monitoring platform via a high-speed communication network. Establish an anomaly detection mechanism to quickly identify potential problems;

[0103] S52. Develop intelligent diagnostic algorithms: Utilize machine learning and data analysis techniques to develop an intelligent fault diagnosis system that automatically analyzes data streams, identifies abnormal behavior patterns, and accurately determines the fault type and its location. Supervised and unsupervised learning methods are used to train the model, capturing known and unknown fault patterns.

[0104] S53. Implement self-healing strategies and automated repair: Develop self-healing strategies for different fault types, such as automatic restart, parameter reconfiguration, and switching to backup paths. For software faults, achieve self-repair through firmware updates and vulnerability fixes; for hardware faults, design redundant architectures to isolate and replace faulty components. Introduce automated recovery processes to reduce downtime and manual intervention.

[0105] S54. Testing, Verification, and Continuous Improvement: Conduct comprehensive testing, ranging from unit testing to system cascading testing, simulating various failure scenarios to evaluate response speed, repair success rate, and the ability to restore normal operation. Based on test results, continuously optimize diagnostic algorithms and self-healing strategies to ensure stable and reliable system operation.

[0106] Furthermore, a high-power silicon carbide power supply load optimization system and method for supercomputing are provided.

[0107] In step S6, the output voltage and current of each stage of the silicon carbide power converter are monitored in real time, and the output power of the converter is dynamically adjusted according to the load demand; the output voltage of other converters is adjusted to stabilize the DC bus voltage and prevent system collapse; maximum power point tracking (MPPT) technology is used to make each stage of the converter operate at its maximum power point as much as possible. The specific steps are as follows.

[0108] S61. Real-time Monitoring and Data Analysis: Establish a real-time monitoring system to continuously collect output voltage, current, and other key parameters (such as temperature and power factor) of each stage of the silicon carbide power converter. Utilize advanced data processing technologies and algorithms to quickly analyze the collected data to identify current load patterns and trends;

[0109] S62. Dynamic Power Adjustment Strategy: Based on real-time monitoring information, an intelligent control algorithm is developed to dynamically adjust the output power of each converter, ensuring its consistently efficient operation. When the DC power supply corresponding to a certain converter decreases, the system quickly responds and adjusts the output voltage of other converters to stabilize the DC bus voltage and prevent system performance degradation or collapse.

[0110] S63. DC Bus Voltage Stabilization Mechanism: The control system is designed to automatically calculate the optimal compensation scheme, determine which converters should increase or decrease their output, and maintain DC bus voltage stability. A buffer energy storage device (such as a supercapacitor or battery) is introduced to absorb or supplement instantaneous energy fluctuations, enhancing system stability.

[0111] S64. Maximum Power Point Tracking (MPPT) Technology Application: Integrating the MPPT algorithm ensures that each stage of the converter operates at its maximum power point as much as possible. The operating point is dynamically adjusted based on the input source characteristics, improving efficiency, reducing energy loss, and optimizing the overall energy efficiency of the supercomputing system.

[0112] A supercomputing high-power silicon carbide power supply load optimization system and method are disclosed. The supercomputing high-power silicon carbide power supply load optimization system and method are used to implement any of the aforementioned supercomputing high-power silicon carbide power supply load optimization systems and methods. The supercomputing high-power silicon carbide power supply load optimization system and method include: a silicon carbide power device design module, a power supply load optimization system construction module, an intelligent prediction module, a real-time monitoring unit configuration module, a fault diagnosis and self-healing function integration module, and a load optimization method module.

[0113] The silicon carbide power device design module uses 4H-SiC material to fabricate high-performance Schottky diodes (SBDs) and junction barrier Schottky diodes (JBSs), and optimizes the structure through Resurf theory and JTE technology to improve breakdown voltage and power density, ensuring stability under high temperature and high power environments.

[0114] Power Load Optimization System Building Module: This module designs and builds a system containing multiple silicon carbide power converters, each drawing power from a DC power source and converting it into output power. By combining them in series or parallel, a power supply path is formed that flexibly adjusts the output power to adapt to complex needs and improve efficiency.

[0115] Intelligent prediction module: Introduces machine learning algorithms to predict future load demand and adjust power output parameters in advance, so that the system always operates at its optimal operating point and maintains high efficiency.

[0116] Real-time monitoring unit configuration module: Install a real-time monitoring unit to continuously monitor the power status and load conditions, ensuring that the system always operates at the optimal operating point and can respond immediately when abnormal conditions are detected, ensuring safety and reliability.

[0117] Fault diagnosis and self-healing function integrated module: Embedded automatic fault diagnosis and self-healing function, automatically detects and repairs hardware or software faults that may affect system stability, quickly restores normal operation, and enhances continuous reliability.

[0118] Load optimization module: Real-time monitoring of the output voltage and current of each stage of silicon carbide power converter, dynamic adjustment of the converter's output power according to load demand, stabilization of DC bus voltage, and adoption of maximum power point tracking (MPPT) technology to improve the overall efficiency of the system.

Claims

1. A method for optimizing the power supply load of a high-power silicon carbide power supply for a supercomputer, characterized in that, Includes the following steps: S1. Silicon Carbide Power Device Design: Schottky diodes and junction barrier Schottky diodes are fabricated using 4H-SiC material. Device structure is optimized by applying the theory of reducing surface electric field effects and junction termination extension techniques to improve breakdown voltage and power density. Specifically, this includes: selecting 4H-SiC as the base material for crystal growth and substrate treatment; designing the device structure based on the theory of reducing surface electric field effects and determining doping concentration and layer thickness parameters; increasing the drift region length or introducing ring or multi-finger termination structures using junction termination extension techniques; developing and implementing semiconductor manufacturing processes; and conducting electrical and thermal tests to verify device performance. S2. Power Load Optimization System Construction: Design and construct a power load optimization system comprising multiple silicon carbide power converters. Each converter obtains electrical energy from a corresponding DC power source and converts it into output power. The multi-stage converters are combined in series or parallel to form a power supply path that can flexibly adjust the output power. Specifically, this includes: defining the overall system architecture; determining the number, type, and topology of the silicon carbide power converters; selecting Schottky diodes and junction barrier Schottky diodes based on 4H-SiC material for each power converter and designing converter parameters; determining the series, parallel, or hybrid connection method for the multiple silicon carbide power converters based on the load requirements of the supercomputing system; developing control algorithms to monitor the operating status of each converter in real time and dynamically adjust its operating point; completing system integration and performing functional verification, stability assessment, and efficiency testing; and deploying the system to the supercomputing environment and establishing a monitoring and maintenance plan. S3. Intelligent Prediction: Introducing machine learning algorithms to predict future load demand, and adjusting power output parameters in advance based on the prediction results to ensure the system operates at its optimal point; specifically including: collecting and preprocessing historical load data and operating status information of the supercomputing system; extracting key feature variables reflecting load trends from the dataset, constructing time lag terms or periodic component features; selecting machine learning algorithms to build a prediction model, training it with historical data and optimizing model parameters through cross-validation; evaluating the model's prediction accuracy using a test set, and adjusting it to the optimal prediction scheme; S4. Real-time monitoring unit configuration: Install a real-time monitoring unit to continuously monitor the power status and load conditions, ensuring that the system always operates at the optimal operating point and responding immediately when an anomaly is detected; S5. Fault diagnosis and self-healing function integration: Embedded automatic fault diagnosis and self-healing function, automatically detects and repairs hardware or software faults that affect system stability; S6. Load Optimization: Real-time monitoring of the output voltage and current of each stage of silicon carbide power converter, dynamically adjusting the output power of the converter according to load demand; adjusting the output voltage of other converters to stabilize the DC bus voltage; employing maximum power point tracking technology to ensure that each stage of converter operates at its maximum power point.

2. The method for optimizing the high-power silicon carbide power supply load for supercomputing as described in claim 1, characterized in that, In step S4, a real-time monitoring unit is installed to continuously monitor the power status and load conditions, ensuring that the system always operates at the optimal operating point and responding immediately when an anomaly is detected. The specific steps are as follows: S41. Determine monitoring needs and objectives: Identify the parameters that need to be monitored, including power supply voltage, current, temperature, humidity, power factor, and load rate; define the performance indicators of the monitoring system, including sampling frequency, response time, accuracy requirements, and alarm thresholds. S42. Selecting Sensors and Measurement Equipment: Select high-precision sensors and technologically advanced measurement equipment suitable for use in the supercomputing environment. For different physical quantities, select the corresponding dedicated sensors. S43. Design the monitoring network architecture: Build an efficient data transmission network to connect various sensors and the central monitoring platform, and select at least one communication technology from industrial Ethernet, RS485 bus, ZigBee, and Wi-Fi; S44. Develop a monitoring software platform: Develop a user-friendly monitoring software that provides an intuitive interface for operators to view real-time data and historical records; the software includes a data analysis module that can automatically identify anomalies and trigger alarms; and integrates intelligent diagnostic functions to help technicians quickly locate the root cause of problems; S45. Configure fault detection and response mechanism: Set reasonable alarm rules and emergency handling procedures, and immediately activate corresponding response measures when any parameter deviating from the normal range is detected, including automatically adjusting power output settings, notifying the maintenance team, and performing protective power cut-off.

3. The method for optimizing the high-power silicon carbide power supply load of a supercomputing system as described in claim 1, characterized in that, Step S5 incorporates automatic fault diagnosis and self-healing functions, automatically detecting and repairing hardware or software faults that affect system stability. The specific steps are as follows: S51. Construct a fault monitoring network: Deploy comprehensive sensors and monitoring nodes in the supercomputing high-power silicon carbide power load optimization system to collect hardware status and software operation data in real time, ensuring coverage of all critical components, and transmit the data to the central monitoring platform through a high-speed communication network to establish an anomaly detection mechanism and quickly identify potential problems. S52. Develop intelligent diagnostic algorithms: Utilize machine learning and data analysis techniques to develop an intelligent fault diagnosis system that automatically analyzes data streams, identifies abnormal behavior patterns, accurately determines fault types and their locations, and trains models using supervised and unsupervised learning methods to capture known and unknown fault patterns. S53. Implement self-healing strategies and automated repair: Develop self-healing strategies for different fault types, including automatic restart, reconfiguration of parameters, and switching to backup paths. For software faults, self-repair is achieved by updating firmware and fixing vulnerabilities. For hardware faults, design a redundant architecture, isolate and replace faulty components, and introduce an automatic recovery process to reduce downtime and manual intervention. S54. Testing, Verification and Continuous Improvement: Conduct comprehensive testing, covering unit testing to system cascade testing, simulate various failure scenarios, evaluate response speed, repair success rate and ability to restore normal operation, and continuously optimize diagnostic algorithms and self-healing strategies based on test results to ensure the system works stably and reliably.

4. The method for optimizing the high-power silicon carbide power supply load of a supercomputing system as described in claim 1, characterized in that, In step S6, the output voltage and current of each stage of the silicon carbide power converter are monitored in real time, and the output power of the converter is dynamically adjusted according to the load demand; the output voltage of other converters is adjusted to stabilize the DC bus voltage; the maximum power point tracking (MPPT) technology is used to make each stage of the converter work at the maximum power point. The specific steps are as follows: S61. Real-time monitoring and data analysis: Establish a real-time monitoring system to continuously collect the output voltage, current, and other key parameters including temperature and power factor of each stage of silicon carbide power converters. Utilize data processing technology and algorithms to quickly analyze the collected data in order to identify the current load patterns and trends. S62. Dynamic power adjustment strategy: Based on real-time monitoring information, develop intelligent control algorithms to dynamically adjust the output power of each converter to ensure that it always operates efficiently. When the DC power of a certain converter decreases, it quickly responds and adjusts the output voltage of other converters to stabilize the DC bus voltage and prevent the system performance from deteriorating or crashing. S63. DC bus voltage stabilization mechanism: The control system is designed to automatically calculate the optimal compensation scheme, determine which converters should increase or decrease their output, maintain DC bus voltage stability, introduce a buffer energy storage device to absorb or supplement instantaneous energy fluctuations, and enhance system stability. S64. Maximum Power Point Tracking (MPPT) Technology Application: Integrating the MPPT algorithm, each stage of the converter operates at its maximum power point as much as possible, dynamically adjusting the operating point according to the characteristics of the input source, improving efficiency, reducing energy loss, and optimizing the overall energy efficiency of the supercomputing system.

5. A supercomputing high-power silicon carbide power supply load optimization system, characterized in that, A method for optimizing the high-power silicon carbide power supply load for supercomputing as described in any one of claims 1-4 includes: Silicon carbide power device design module: Schottky diodes and junction barrier Schottky diodes are fabricated using 4H-SiC material, and the structure is optimized by reducing surface electric field effect theory and junction termination extension technology to improve breakdown voltage and power density; Power load optimization system construction module: Construct a system containing multiple silicon carbide power converters. Each converter obtains electrical energy from a DC power source and converts it into output power. Through series or parallel combination, a power supply path that can flexibly adjust the output power is formed. Intelligent prediction module: Introduces machine learning algorithms to predict future load demand and adjust power output parameters in advance so that the system operates at the optimal operating point; Real-time monitoring unit configuration module: Install a real-time monitoring unit to continuously monitor power status and load conditions, ensure that the system operates at the optimal operating point and respond immediately when an anomaly is detected; Fault diagnosis and self-healing function integration module: Embedded automatic fault diagnosis and self-healing function, automatically detects and repairs hardware or software faults that affect system stability; Load optimization module: Real-time monitoring of the output voltage and current of each stage of silicon carbide power converter, dynamic adjustment of output power according to load demand, stabilization of DC bus voltage, and use of maximum power point tracking technology to improve system efficiency.

Images