Power supply modular combination method and system based on standardized interface

Abstract

The application discloses a power supply modular combination method and system based on a standardized interface, and the method comprises the following steps: step one, collecting data through a sensor; step two, fusing data by a central control hub and establishing a four-dimensional collaborative model; step three, adjusting according to a control instruction of the central control hub by a MOSFET driving circuit; and step four, predicting a load change trend and a performance attenuation of each module of the current power supply system by an AI prediction current sharing module through a built-in machine learning model, dynamically adjusting current distribution of each module based on a prediction result and a target current value in the instruction, and the system comprises the central control hub, a sensor unit, the AI prediction current sharing module and the MOSFET driving circuit; the application has the beneficial effects that dynamic impedance matching and current sharing control are closed loops, the AI prediction current sharing is realized by using the machine learning model, the output impedance of the power supply module and the current distribution are accurately controlled, the load change is dynamically adapted, and the current sharing precision and the system stability are improved.

Description

Technical Field

[0001] This invention relates to the field of power modules, specifically to a method and system for modular power module assembly based on standardized interfaces. Background Technology

[0002] With the rapid development of new energy, industrial automation, data centers and other fields, higher requirements are being placed on the flexibility, reliability and efficiency of power systems. Traditional power systems suffer from low modularity and complex fault maintenance, making it difficult to meet diverse load demands and intelligent management needs. Power systems still face technical bottlenecks in multi-module collaborative control, dynamic performance optimization and predictive maintenance. For example, insufficient current sharing control accuracy between modules leads to reduced system efficiency and shortened module lifespan. Limited rapid reconfiguration and adaptive adjustment capabilities in the event of a fault affect system reliability. In addition, with the deep integration of industrial internet and artificial intelligence technologies, power systems urgently need to have data-driven intelligent decision-making capabilities to achieve multi-objective dynamic balance and full lifecycle management functions. Summary of the Invention

[0003] The purpose of this invention is to solve the problems mentioned above, such as low modularity, complex fault maintenance, and difficulty in meeting diverse load requirements and intelligent management requirements of traditional power supply systems. Therefore, this invention proposes a modular power supply combination method and system based on standardized interfaces.

[0004] The objective of this invention can be achieved through the following technical solution: a modular power supply assembly method and system based on standardized interfaces, comprising:

[0005] Step 1: Collect output impedance data, current data, junction temperature data, and fault signals through sensors, and transmit these data to the central control hub;

[0006] Step 2: The central control hub, as the data fusion center, collects the data fed back by the sensors, builds a decision scheduling engine, establishes a four-dimensional collaborative model, iteratively optimizes the four-dimensional collaborative model through the decision scheduling engine, and converts the model optimization results into control commands to feed back to the AI ​​prediction current sharing module and MOSFET drive circuit.

[0007] Step 3: The MOSFET driver circuit receives control commands from the central control hub, adjusts the MOSFET's conduction level and switching frequency to change the power module's output impedance and current output, and feeds back the executed status data to the central control hub.

[0008] Step 4: The AI ​​predictive current sharing module uses a built-in machine learning model to predict the current load change trend of the power system and the performance degradation of each module. Based on the prediction results and the target current value in the command, it dynamically adjusts the current distribution of each module. At the same time, the adjusted current distribution results are fed back to the central control hub for subsequent optimization and adjustment of control strategies.

[0009] The present invention also provides a power supply modular combination system based on a standardized interface, wherein the above method is applied to the power supply modular combination system, which includes a central control hub, a sensor unit, an AI predictive current sharing module, and a MOSFET drive circuit.

[0010] Furthermore, the sensor unit is used to collect data and transmit the data to the central control hub. The sensor unit includes a vector impedance sensor, a current sensor, a temperature sensor, and a fault detection sensor. The vector impedance sensor is used to collect the output impedance data of the power module, the current sensor is used to collect real-time current data, the temperature sensor is used to collect junction temperature data, and the fault detection sensor is used to collect fault signals.

[0011] Furthermore, the central control hub, acting as a data fusion center, uses clock synchronization technology to collect data transmitted by sensor units, construct a decision scheduling engine, and establish a four-dimensional collaborative model.

[0012] Furthermore, the four-dimensional collaborative model achieves multi-objective dynamic balance by coupling and modeling four key indicators: power output stability, module lifespan, energy efficiency, and fault response. It also utilizes the CANFD bus to build an instruction distribution system, forming a closed-loop control.

[0013] The established four-dimensional collaborative model enables precise control of power output stability, effectively reducing voltage fluctuations, providing a more stable power supply to the load, meeting the power requirements of high-precision equipment, and performing thermal management based on junction temperature data to monitor and adjust module temperature in real time, avoiding performance degradation and component damage caused by overheating, significantly extending the service life of the power module, reducing maintenance costs, optimizing real-time current and impedance to reduce losses during power transmission, improving the overall energy efficiency of the power system, conforming to the development trend of green energy saving, and quickly processing fault signals to effectively improve the system's fault tolerance and reliability, ensuring power supply continuity.

[0014] Furthermore, the central control hub has a built-in current sharing control algorithm to calculate the control parameters required to adjust the power module output, thereby generating control commands to be sent to the MOSFET drive circuit. By adjusting the MOSFET drive circuit, the output impedance and current output of the power module are changed.

[0015] By introducing a current sharing error compensation term, the current deviation between each power supply module is monitored and corrected in real time, so that the current distribution error between modules can be reduced to within ±1%, ensuring system load balance and avoiding performance degradation or failure due to overload of some modules.

[0016] Furthermore, the AI ​​predictive current sharing module uses a built-in machine learning model to predict the load change trend of the power system and the performance degradation of each module. Based on the prediction results and the target current value in the central control hub command, it dynamically adjusts the current distribution of each module. At the same time, it feeds back the adjusted current distribution results to the central control hub for subsequent optimization and adjustment of the control strategy, thereby forming inner and outer loop control.

[0017] By anticipating load change trends in advance, the current output is proactively increased when the load is about to increase, effectively preventing voltage drops. When the load decreases, the current is reduced in a timely manner to prevent energy waste, significantly improving the dynamic response speed and stability of the power supply system. The current is dynamically allocated according to the performance status of each module, and intelligent compensation is performed for modules with performance degradation to ensure the overall current sharing effect of the system. This reduces the current imbalance caused by differences in module performance and extends the service life of the power supply modules. Through the dynamic current allocation strategy, the system can maintain stable operation when some modules experience performance degradation, reducing the impact of single-point failures on the overall system.

[0018] Furthermore, the flow sharing control algorithm is as follows:

[0019]

[0020] Where ΔD(k) is the conduction ratio adjustment in the k-th iteration, and μ(k) is the adaptive step size. Let I be the output impedance gradient vector, λ be the current sharing weighting coefficient, and I be the current sharing weighting coefficient. avg (k) represents the average output current of the module group, I i (k) represents the current output current of the module.

[0021] Furthermore, the specific method for dynamically adjusting the current distribution of each module is as follows:

[0022] When an increase in load is anticipated, the current output of each module is increased in advance; when the load decreases, the current of each module is reduced proportionally.

[0023] For modules whose performance has degraded, based on their historical data and current status, the current allocation ratio of the module is reduced, and the reduced current is allocated to other modules to ensure the overall current sharing effect and stability of the system. At the same time, the adjusted current allocation result is fed back to the central control hub.

[0024] Furthermore, the built-in machine learning model employs a machine learning model that integrates long short-term memory networks and support vector regression.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] 1. Construct a four-dimensional collaborative model based on the improved NSGA-II algorithm, which integrates four dimensions—power output stability, module lifespan, energy efficiency, and fault response—into coupled modeling to achieve multi-objective dynamic balance. Through closed-loop control, improve the overall performance of the power system, extend module lifespan, reduce energy consumption, and at the same time, respond quickly to faults to ensure stable system operation.

[0027] 2. Dynamic impedance matching and current sharing control closed loop: Introducing a current sharing error compensation term, using a fusion machine learning model to achieve AI prediction of current sharing, accurately controlling the output impedance and current distribution of the power module, dynamically adapting to load changes, compensating for module performance degradation, and improving current sharing accuracy and system stability.

[0028] 3. Build an AI-driven predictive maintenance system, including a multimodal deep neural network model that includes subnetworks for life prediction, thermal runaway early warning, and fault location. This system can predict module life, thermal runaway risk, and fault location in advance, and use coordinated control strategies to reduce the failure rate and improve system reliability. Attached Figure Description

[0029] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0030] Figure 1 This is a flowchart of the power supply modular combination method based on standardized interfaces according to the present invention. Detailed Implementation

[0031] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0032] Please see Figure 1 As shown, the modular power supply assembly method based on standardized interfaces comprises the following steps:

[0033] Step 1: Collect output impedance data, current data, junction temperature data, and fault signals through sensors, and transmit these data to the central control hub;

[0034] Step 2: The central control hub, as the data fusion center, collects the data fed back by the sensors, builds a decision scheduling engine, establishes a four-dimensional collaborative model, iteratively optimizes the four-dimensional collaborative model through the decision scheduling engine, and converts the model optimization results into control commands to feed back to the AI ​​prediction current sharing module and MOSFET drive circuit.

[0035] Step 3: The MOSFET driver circuit receives control commands from the central control hub, adjusts the MOSFET's conduction level and switching frequency to change the power module's output impedance and current output, and feeds back the executed status data to the central control hub.

[0036] Step 4: The AI ​​prediction current sharing module uses a built-in machine learning model to predict the current load change trend of the power system and the performance degradation of each module. Based on the prediction results and the target current value in the command, it dynamically adjusts the current distribution of each module. At the same time, it feeds back the adjusted current distribution results to the central control hub for subsequent optimization and adjustment of control strategies.

[0037] This modular power supply combination method establishes a data acquisition framework in step one, which provides data support for step two. Step two relies on step one, integrates the information provided in step one, and establishes a four-dimensional collaborative model for multi-objective dynamic balancing. Step two feeds back the generated control commands based on the four-dimensional collaborative model to step three. The MOSFET drive circuit in step three makes corresponding adjustments according to the control commands. Step four predicts the current load change trend of the power supply system and the performance degradation of each module. Based on the prediction results and the target current value in the command in step two, the current distribution of each module is dynamically adjusted, thereby completing the process from data acquisition to command generation, command implementation, and predictive adjustment.

[0038] The present invention also provides a power modular combination system based on a standardized interface, wherein the above method is applied to the power modular combination system, and the system includes: a central control hub, a sensor unit, an AI prediction current sharing module, and a MOSFET drive circuit;

[0039] The sensor unit includes a vector impedance sensor, a current sensor, a temperature sensor, and a fault detection sensor. The sensor unit is used to collect data and transmit the data to the central control hub.

[0040] The central control hub, as the data fusion center, uses clock synchronization technology to collect sensor unit signals, build a decision scheduling engine based on the improved NSGA-II algorithm, and establish a four-dimensional collaborative model.

[0041] The four-dimensional collaborative model couples and models four key indicators: power output stability, module lifespan, energy efficiency, and fault response, to achieve dynamic balance of multiple objectives. It also uses the CANFD bus to build a command distribution system and form a closed-loop control.

[0042] The process of building the four-dimensional collaborative model is as follows: the power module output impedance, real-time current, junction temperature data and fault signal raw data and power output voltage collected by clock synchronization technology are filtered and normalized to eliminate dimensional differences and outlier interference and ensure data quality.

[0043] Power output stability model: The voltage fluctuation range and ripple coefficient are converted into quantitative parameters. The voltage fluctuation range is the difference between the maximum and minimum measured voltage within a given time range. The ripple coefficient is the ratio of the peak value of AC ripple voltage to the DC output voltage. A voltage stability function is constructed, and the output voltage curve is fitted by the least squares method. The standard deviation is calculated as a stability evaluation index.

[0044] Module life loss model: A thermal stress model is established based on junction temperature data. Combined with the Arrhenius equation, the junction temperature is correlated with the module aging rate to generate a life loss function.

[0045] Energy efficiency optimization model: Based on real-time current and impedance data, a power loss model is constructed, and the system energy loss is reduced by optimizing impedance matching, thus establishing an objective function for improving energy efficiency;

[0046] Fault suppression model: Features are extracted from fault signals, fault response time thresholds and processing priorities are set, and a fault suppression function is constructed;

[0047] Using multi-objective optimization theory, the quantization functions of the above four dimensions are taken as sub-objectives. Through the weighted coefficient method, a joint optimization model containing the four sub-objectives is constructed. At the same time, constraints are introduced, including the power output power range and the module temperature safety range, to ensure that the model meets the actual engineering requirements.

[0048] The improved NSGA-II algorithm is used to iteratively optimize the four-dimensional collaborative model. Through selection, crossover, mutation and genetic operations, the optimal solution set is searched in the feasible solution space to achieve a multi-objective dynamic balance of power output stability, module life, energy efficiency and fault response.

[0049] A command distribution system is built using the CANFD bus to convert model optimization results into control commands and feed them back to the power module. Real-time data after execution is collected for verification. By comparing the target value with the actual output value, the model parameters and algorithm strategy are dynamically adjusted to form a closed-loop control system of data acquisition, modeling optimization, command execution and effect feedback.

[0050] During use, the vector impedance sensor collects the output impedance data of the power module, the current sensor collects real-time current data, the temperature sensor collects junction temperature data, and the fault detection sensor collects fault signals. These data are then transmitted to the central control hub.

[0051] After receiving the data, the central control hub calculates the control parameters required to adjust the power module output based on the built-in four-dimensional collaborative model and current sharing control algorithm, and generates control commands including current distribution instructions. The current sharing control algorithm is as follows: ΔD(k) is the conduction ratio adjustment in the k-th iteration, and μ(k) is the adaptive step size, dynamically adjusted based on the fuzzy PID algorithm. Let I be the output impedance gradient vector, λ be the current sharing weighting coefficient, and I be the current sharing weighting coefficient. avg (k) represents the average output current of the module group, I i (k) represents the current output current of the current module. This instruction is based on the real-time current I of each module. i (k) System average current I avg (k) and the flow equalization error compensation term λ·(I) avg (k)-I i (k) , which clarifies the target current value that each module needs to output;

[0052] After receiving the current distribution command from the central control hub, the AI-predictive current sharing module first uses its built-in machine learning model to predict the load change trend of the power system and the performance degradation of each module. Based on the prediction results and the target current value in the command, it dynamically adjusts the current distribution of each module, including:

[0053] Load change response: When an increase in load is predicted, the current output of each module is increased in advance according to the preset value to avoid voltage fluctuations due to insufficient current supply. When the load decreases, the current of each module is reduced according to the preset ratio to prevent energy waste.

[0054] Module performance compensation: For modules with degraded performance, the AI ​​predictive current sharing module reduces the current allocation ratio of the module based on its historical data and current status, allocating more current to other modules to ensure the overall current sharing effect and stability of the system. At the same time, the adjusted current allocation result is fed back to the central control hub for subsequent optimization and adjustment of control strategies, thus forming inner and outer loop control. The MOSFET drive circuit receives control commands from the central control hub and changes the output impedance and current output of the power module by adjusting the conduction level and switching frequency of the MOSFET drive circuit, executing the adjustment strategy of the impedance-current dual closed loop controller, and feeding back the status data after execution to the central control hub.

[0055] The module performance compensation process is as follows: The AI ​​predictive current sharing module constructs a module health assessment model based on historical data and real-time monitoring data. It quantifies module performance from three dimensions: output voltage stability, efficiency degradation rate, and temperature rise slope. A weighted scoring method is used, assigning weights of 0.4 to the number of times voltage fluctuations exceed the threshold, the percentage of efficiency decrease, and the junction temperature growth rate, respectively, to calculate the module health score H. i A lower score indicates more severe performance degradation. Based on the average health score H of all modules, modules with a score below H are defined as performance degradation modules, and their current distribution is reduced by a percentage ΔP. i It is inversely proportional to the health score, and the formula is: α is an adjustment coefficient that redistributes the current reduction caused by the performance degradation module according to the proportion of the health scores of other modules, i.e., the current increment ΔI of the normal performance module. j for While ensuring the overall current sharing effect of the system, the system efficiency and stability are maximized. At the same time, the adjusted current distribution results are fed back to the central control hub for subsequent optimization and adjustment of control strategies, thus forming inner and outer loop control.

[0056] Where, ΔI j H represents the current increment of the j-th power module with better performance. j This represents the health score of the j-th power module. This score reflects the current performance status of the module; a higher score indicates better module performance. ∑ k≠i H k It is the sum of the health scores of all modules except the i-th module with performance degradation, used to determine the weight ratio of each module during current redistribution, ∑ i ΔP i I represents the sum of the power reductions of all performance-degraded modules, reflecting the total power lost due to module performance degradation. avg The average current of the system is used to convert power changes into current regulation.

[0057] The AI-predicted current sharing module is also used for predictive maintenance, the specific process of which is as follows:

[0058] Building a lifetime prediction model: Based on historical junction temperature data and current fluctuations, a module lifetime prediction model is constructed by combining the Arrhenius equation and the long short-term memory network. The Arrhenius equation quantifies the impact of temperature on the aging rate of the device, while the long short-term memory network processes time series data, learns the degradation trend of module performance over time, and predicts the remaining lifetime.

[0059] Build a thermal runaway early warning model: Utilize temperature gradient data collected in real time by temperature sensors, and analyze data features through convolutional neural networks. Convolutional neural networks are good at capturing spatial features, and here they are used to identify abnormal temperature change patterns, such as temperature abrupt changes and overheating zone diffusion, to predict the risk of thermal runaway in advance.

[0060] A fault location model is built. Based on the characteristics of fault signals, such as sudden current changes and abnormal voltage fluctuations, a graph neural network is used to analyze the connection topology of the power module. The graph neural network can quickly locate the fault location through the information transmission between nodes and edges, and achieve accurate identification of the fault module.

[0061] Based on the prediction results, the corresponding control mechanism is triggered. When the life prediction model determines that the remaining life of a certain module is lower than the threshold, the central control hub adjusts the current distribution of the module and reduces its workload to the preset value. When the thermal runaway early warning model issues an alarm, the heat dissipation system automatically increases the heat dissipation power to the preset value. At the same time, the central control hub limits the output power of the power module. When the fault location model identifies the faulty module, the central control hub immediately starts the solid-state switch group, cuts off the faulty circuit, and performs redundancy reconstruction.

[0062] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of protection of the claims.

[0063] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. 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. A modular power supply assembly method based on standardized interfaces, characterized in that, include: Step 1: Collect output impedance data, current data, junction temperature data, and fault signals through sensors, and transmit these data to the central control hub; Step 2: The central control hub, as the data fusion center, collects the data fed back by the sensors, builds a decision scheduling engine, establishes a four-dimensional collaborative model, iteratively optimizes the four-dimensional collaborative model through the decision scheduling engine, and converts the model optimization results into control commands to feed back to the AI ​​prediction current sharing module and MOSFET drive circuit. Step 3: The MOSFET driver circuit receives control commands from the central control hub, adjusts the MOSFET's conduction level and switching frequency to change the power module's output impedance and current output, and feeds back the executed status data to the central control hub. Step 4: The AI ​​prediction current sharing module uses a built-in machine learning model to predict the current load change trend of the power system and the performance degradation of each module. Based on the prediction results and the target current value in the command, it dynamically adjusts the current distribution of each module. At the same time, it feeds back the adjusted current distribution results to the central control hub for subsequent optimization and adjustment of control strategies. The power system includes a central control hub, sensor units, an AI predictive current sharing module, and MOSFET drive circuitry. The specific method for dynamically adjusting the current distribution of each module is as follows: When an increase in load is anticipated, the current output of each module is increased in advance; when the load decreases, the current of each module is reduced proportionally. For modules whose performance has degraded, based on their historical data and current status, the current allocation ratio of the module is reduced, and the reduced current is allocated to other modules to ensure the overall current sharing effect and stability of the system. At the same time, the adjusted current allocation result is fed back to the central control hub. The machine learning model employs a fusion of long short-term memory network and support vector regression.

2. A power supply modular assembly system based on a standardized interface, used to execute the power supply modular assembly method based on a standardized interface as described in claim 1, characterized in that, The sensor unit is used to collect data and transmit the data to the central control hub. The sensor unit includes a vector impedance sensor, a current sensor, a temperature sensor, and a fault detection sensor. The vector impedance sensor is used to collect the output impedance data of the power module, the current sensor is used to collect real-time current data, the temperature sensor is used to collect junction temperature data, and the fault detection sensor is used to collect fault signals.

3. The power supply modular combination system based on a standardized interface according to claim 2, characterized in that, The central control hub, serving as a data fusion center, uses clock synchronization technology to collect data transmitted by sensor units, construct a decision scheduling engine, and establish a four-dimensional collaborative model.

4. The power supply modular combination system based on a standardized interface according to claim 2, characterized in that, The four-dimensional collaborative model achieves multi-objective dynamic balance by coupling and modeling four key indicators: power output stability, module lifespan, energy efficiency, and fault response. It also utilizes the CANFD bus to build an instruction distribution system, forming a closed-loop control.

5. The power supply modular combination system based on a standardized interface according to claim 2, characterized in that, The central control hub has a built-in current sharing control algorithm to calculate the control parameters required to adjust the power module output, thereby generating control commands to be sent to the MOSFET drive circuit. By adjusting the MOSFET drive circuit, the output impedance and current output of the power module are changed.

6. The power supply modular combination system based on a standardized interface according to claim 5, characterized in that, The flow sharing control algorithm is as follows: ； in, This is the conduction ratio adjustment amount for the k-th iteration. For adaptive step size, The output impedance gradient vector. The flow equalization weighting coefficient is... The average output current of the module group. Output current for the current module.

Images