Method, device, equipment, medium and program product for regulating a load supply circuit

Abstract

The application discloses a load power supply loop regulation method, device, equipment, medium and program product, relates to the technical field of servers, and comprises the following steps: when a hot plug event occurs, the current change rate, contact temperature rise rate and voltage change rate of a load interface are acquired to comprehensively reflect the rapid change of the circuit state at the hot plug moment; then, the current change rate, contact temperature rise rate and voltage change rate are input into a surge probability prediction model to obtain a surge occurrence probability, so that the surge risk is predicted in advance; when the surge occurrence probability is greater than or equal to a first preset probability, the resistance-capacitance parameters of the load power supply loop are calculated according to the monitored load interface current, load interface voltage and load interface temperature, so that when the hot plug event occurs, the circuit can be quickly adjusted according to the resistance-capacitance parameters, the abnormal change of the current and voltage is timely suppressed, the protection response speed is improved, the damage of equipment caused by untimely protection is effectively avoided, and the stable operation of the server under complex conditions such as load hot plug is ensured.

Description

Technical Field

[0001] This application relates to the field of server technology, and in particular to a method, apparatus, equipment, medium, and program product for regulating a load power supply circuit. Background Technology

[0002] The stable operation of servers heavily relies on continuous and effective heat dissipation, and the reliable operation of fans, as the core heat dissipation component of most current server systems, is crucial. To improve the convenience of server maintenance and the continuous operation capability of the system, most mainstream servers are equipped with hot-swappable cooling fans, allowing direct plugging and unplugging of fans during system operation. However, the protection measures for the hot-swapping process in related technologies are not yet perfect. Typically, only electronic fuses (eFuse) are used to protect the fans to deal with abnormal conditions such as overcurrent, overvoltage, or short circuits, preventing damage to the fans and related circuits due to such faults.

[0003] While eFuses offer some short-circuit and overcurrent protection, in scenarios like fan hot-swapping, the rapid changes in system state can easily generate instantaneous inrush currents. These inrush currents can cause the eFuse to malfunction or even damage related circuits. Furthermore, the eFuse's thermal shutdown function relies on the temperature rise of its internal integrated circuit for triggering, which often results in a protection response that lags behind actual needs, especially during rapid hot-swapping events. This lag can lead to untimely protection and consequently, equipment damage. Summary of the Invention

[0004] This application provides a method, apparatus, equipment, medium, and program product for regulating a load power supply circuit, so as to at least solve the problem of system failure caused by surge current generated during hot-plugging of the load in the related art.

[0005] This application provides a method for regulating a load power supply circuit, comprising: when a baseboard management controller detects a hot-plug event of a load, acquiring the current change rate, contact heating rate, and voltage change rate of the load interface; wherein the hot-plug event is identified by the level signal of the presence detection pin of the load power supply circuit; inputting the current change rate, contact heating rate, and voltage change rate into a surge probability prediction model to obtain the surge occurrence probability output by the surge probability prediction model; wherein the surge probability prediction model is constructed based on a long short-term memory network and trained on a historical dataset; when the surge occurrence probability is greater than or equal to a first preset probability, calculating the resistance and capacitance parameters of the load power supply circuit based on the monitored load interface current, load interface voltage, and load interface temperature; and regulating the load power supply circuit according to the resistance and capacitance parameters.

[0006] This application also provides a control device for a load power supply circuit, comprising:

[0007] The acquisition module is used to acquire the current change rate, contact heating rate, and voltage change rate of the load interface when the baseboard management controller detects a hot-plug event of the load.

[0008] The surge prediction module is used to input the current change rate, contact heating rate and voltage change rate into the surge probability prediction model to obtain the surge occurrence probability output by the surge probability prediction model; wherein, the surge probability prediction model is built based on a long short-term memory network and trained on historical datasets;

[0009] The calculation module is used to calculate the resistance and capacitance parameters of the load power supply circuit based on the monitored load interface current, load interface voltage and load interface temperature when the surge occurrence probability is greater than or equal to the first preset probability.

[0010] The control module is used to regulate the load power supply circuit according to the resistance and capacitance parameters.

[0011] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for executing the computer program to implement the above-described load power supply circuit control method.

[0012] This application also provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the above-described load power supply circuit control method.

[0013] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described load power supply circuit control method.

[0014] This application acquires the current change rate, contact heating rate, and voltage change rate of the load interface when a hot-plug event occurs. These key indicators comprehensively reflect the rapid changes in the circuit state at the moment of hot-plugging. Subsequently, these key indicators are input into a surge probability prediction model to obtain the surge probability, thereby predicting surge risks in advance. When the surge probability is greater than or equal to a first preset probability, it means that there may be a significant surge risk. At this time, based on the monitored load interface current, load interface voltage, and load interface temperature, the resistance and capacitance parameters of the load power supply circuit are calculated. Thus, when a hot-plug event occurs, the circuit can be quickly adjusted according to the resistance and capacitance parameters to suppress abnormal changes in current and voltage in a timely manner, improve the protection response speed, effectively avoid equipment damage caused by untimely protection, and ensure the stable operation of the server under complex conditions such as load hot-plugging. Attached Figure Description

[0015] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A schematic diagram of the specific hardware architecture on which the execution of a load power supply circuit regulation method provided in this application embodiment depends;

[0017] Figure 2 A schematic flowchart illustrating a method for regulating a load power supply circuit according to an embodiment of this application;

[0018] Figure 3 This is a schematic diagram of the structure of an energy feedback circuit provided in an embodiment of this application;

[0019] Figure 4 A flowchart illustrating the resistance and capacitance parameters of a dynamically switching load power supply circuit provided in an embodiment of this application;

[0020] Figure 5 A flowchart illustrating the load power supply circuit control method provided in this application embodiment. Figure 2 ;

[0021] Figure 6 A schematic diagram of the structure of a load power supply circuit control device provided in an embodiment of this application;

[0022] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0024] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0025] To more clearly illustrate the embodiments of this application, the technical terms used in the embodiments will be briefly introduced below:

[0026] An electronic fuse (eFuse) is a programmable protection device based on semiconductor technology, replacing traditional mechanical fuses. By monitoring parameters such as current, voltage, or temperature in a circuit, it can quickly cut off the circuit when abnormalities such as overcurrent or overvoltage occur (the response time is typically in the microsecond range). Some models also support a reset function, making them reusable. They are widely used in power management, consumer electronics, and other fields.

[0027] Inrush current refers to the short-term peak current that far exceeds the steady-state operating current generated when a circuit is switched on (or when a device is started).

[0028] The negative temperature coefficient (NTC) refers to the physical property that the resistance of a component decreases as the temperature increases, and vice versa. Components with this property (such as NTC thermistors) are commonly used in temperature measurement, circuit current limiting (surge protection at startup), and temperature compensation.

[0029] An analog-to-digital converter (ADC) is an electronic device that converts continuously changing analog signals (such as voltage and current) into discrete digital signals (binary code). The conversion process typically includes three core steps: sampling, quantization, and encoding. Its performance is determined by resolution (number of bits, such as 8 bits or 12 bits), conversion rate, and accuracy. It is a key interface connecting analog circuits and digital systems (such as microcontrollers and processors) and is widely used in fields such as sensor signal acquisition and audio processing.

[0030] Long Short-Term Memory (LSTM) networks are a special variant of Recurrent Neural Networks (RNNs) designed to address the long-sequence dependency problem of traditional RNNs, namely their inability to effectively remember early information in long sequences. By designing forget gates, input gates, and output gates, they can selectively retain or discard historical information, enabling them to process time-series data such as text, speech, and time-series signals. They are commonly used in tasks such as natural language processing, speech recognition, and time-series prediction.

[0031] Overcurrent protection (OCP) is a circuit safety protection mechanism. Its core is to trigger protective actions such as cutting off the circuit or limiting the current when the current in the circuit exceeds a preset safety threshold, so as to prevent excessive current from damaging the load, wires or power supply devices.

[0032] A MOSFET array is a device or circuit unit that integrates multiple metal-oxide-semiconductor field-effect transistors (MOSFETs) in a specific topology (such as parallel, series, or hybrid connection). Its purpose is to improve overall current carrying capacity, increase withstand voltage, or implement complex switching logic. It is widely used in high-power applications such as high-power power conversion, motor control, and load switching, and can reduce the power loss and heat generation of a single MOSFET.

[0033] Back electromotive force (EMF) is an electromotive force generated in electromagnetic induction phenomena by a conductor or coil due to changes in its own current or movement in a magnetic field, which is opposite in direction to the external driving voltage. It is commonly found in inductive components such as inductors, motors, and transformers. For example, when a motor is running, the back EMF generated by the coil cutting the magnetic field will suppress the input current. If the power is suddenly cut off, the back EMF may cause high voltage, which requires protection circuits through devices such as freewheeling diodes.

[0034] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] The specific application environment architecture or specific hardware architecture on which the control method of the load power supply circuit depends is described here.

[0036] like Figure 1 As shown, Figure 1 A schematic diagram of the specific hardware architecture upon which the regulation method for the power supply circuit to the load depends.

[0037] The intelligent prediction and pre-regulation system collects current, temperature, and voltage signals from the load power supply circuit through a high-frequency current sensor, temperature detection module, and voltage detection module. The field-programmable gate array (FPGA) predictive controller monitors the collected multi-dimensional data, analyzes whether the circuit may face risks such as surges, generates impedance adjustment commands, provides a basis for subsequent dynamic adjustment of circuit parameters, and performs graded pre-charging.

[0038] The dynamic resistance-capacitance regulation system uses a proportional-integral-derivative (PID) control chip to receive impedance regulation commands from the intelligent prediction and pre-adjustment system and generate resistance-capacitance regulation commands based on its own calculations. The numerically controlled resistor array and resistance-capacitance network dynamically adjust the resistance and capacitance parameters of the load power supply circuit according to the resistance-capacitance regulation commands, optimize impedance characteristics, thereby suppressing surges and ensuring stable power supply.

[0039] The energy buffering and reuse module, when there is residual surge energy in the circuit, directs the residual surge energy into the energy storage capacitor bank for storage through the energy feedback circuit. If the system bus voltage is insufficient or the load operating parameters do not meet the standards, the energy storage capacitor bank can release energy through the energy feedback circuit to replenish the load power supply circuit, realizing energy recycling and improving energy efficiency.

[0040] The specific hardware architecture upon which the load power supply circuit regulation method relies to implement end-to-end protection of prediction, suppression and recovery, enhances the stability of load power supply, improves energy utilization efficiency, and balances safety and energy efficiency.

[0041] The embodiments of this application provide a method for regulating a load power supply circuit. The method is described in detail below in conjunction with the execution flow of the method for regulating a load power supply circuit.

[0042] like Figure 2 As shown, the control method for the load power supply circuit includes the following steps S201~S204:

[0043] S201. When the baseboard management controller detects a hot-plug event of the load, it acquires the current change rate, contact heating rate, and voltage change rate of the load interface.

[0044] The load can be a powered component such as a fan (e.g., P12V_FAN), hard drive, or module. A load hot-plug event refers to the operation or event in which the load is connected to or disconnected from the circuit while the circuit or device is still powered on.

[0045] The Baseboard Management Controller (BMC) determines whether a hot-plug event has occurred by identifying the level signal of the presence detection pin on the load power supply loop. The BMC communicates directly with the load power supply loop interface via the server's internal hardware management bus to monitor the physical connection status of the load in real time. When a load is inserted, the presence detection pin on the load power supply loop interface generates a fixed level signal via hardware circuitry, such as a high level indicating "present" and a low level indicating "absent". The BMC reads this level signal via the hardware management bus to determine the physical connection status.

[0046] In some embodiments, the board management controller can monitor load hot-plug events through at least one of the following methods: A small-value sampling resistor is connected in series in the load power supply circuit; the voltage across the resistor is amplified by an operational amplifier; when the voltage surge exceeds a threshold, an interrupt signal is triggered to confirm a load hot-plug event; or, a voltage sampling circuit is connected in parallel at the load power supply terminal, a normal voltage range is set, and when the voltage exceeds the range and the rate of change exceeds a threshold, a hot-plug event is determined; or, a hot-plug event is determined by comparing the difference between the real-time load current and its rated value, for example, if the fan current suddenly drops from its rated value to 0 and the speed signal disappears, it can be determined that the fan is unplugged; if the fan current suddenly rises from 0 to its rated value and the speed signal reappears, it can be determined that the fan is plugged in. Alternatively, a query command can be periodically sent to the load; if there is no response after multiple consecutive attempts or a sudden response is received, a hot-plug event can be determined.

[0047] The above embodiments monitor load hot-plug events by capturing subtle signals of current surges through sampling resistors and operational amplifier amplification. Combined with voltage change rate, load rating comparison, and an active query mechanism, a dual judgment logic of passive sensing and active verification is formed. This effectively distinguishes parameter surges caused by hot-plugging from fluctuations during normal operation, reducing the false alarm rate. Different monitoring methods can be matched to different load types, allowing users to flexibly select or combine them according to load characteristics, without relying on a single monitoring mode. By capturing sudden current changes and voltage anomalies caused by hot-plugging in real time, interrupt signals or judgment mechanisms can be quickly triggered, providing sufficient response time for subsequent surge current suppression. This prevents damage to power supply circuit components from surge currents during hot-plugging and also prevents voltage spikes caused by sudden load removal from affecting other devices. The active query mechanism complements passive parameter monitoring. Even if passive monitoring has sensing blind spots due to differences in load characteristics, periodic communication queries can accurately determine hot-plug events through response loss / recovery, avoiding the limitations of a single monitoring method and ensuring stable monitoring of hot-plug events under different load types and power supply environments.

[0048] In some embodiments, when the board management controller detects a hot-plug event of the load, it monitors the load interface current, load interface temperature, and load interface voltage. Then, it calculates the current change rate based on the load interface current, the contact temperature rise rate based on the load interface temperature, and the voltage change rate based on the load interface voltage.

[0049] Optionally, the load interface current is acquired using a high-frequency current sensor, and the rate of change of current (di / dt) is calculated. The load interface temperature is acquired using a temperature detection module, and the contact heating rate (ΔT / Δt) is calculated. This temperature detection module may include a surface-mount negative temperature coefficient NTC thermistor. The load interface voltage is acquired using a voltage detection module, and the rate of change of voltage (ΔV / Δt) is calculated. This voltage detection module may include a high-voltage differential probe. Optionally, the load interface current, load interface temperature, and load interface voltage are pre-processed and converted from analog to digital signals to obtain digital signals.

[0050] Specifically, when connecting a high-frequency current sensor in series or through non-contact coupling to the load power supply circuit, it is necessary to ensure that the sensor bandwidth matches the load current change rate and to acquire the instantaneous current signal of the load interface in real time. The current signal output by the high-frequency current sensor is conditioned by an operational amplifier and then sent to an analog-to-digital converter (ADC) for high-frequency sampling. The digital signal processing unit then performs differential calculations on the continuously sampled current values ​​and outputs the di / dt values ​​in real time, where di is the current difference and dt is the sampling interval.

[0051] A surface-mount NTC thermistor is mounted near the load interface contacts. A voltage divider circuit converts the resistance change of the NTC thermistor into a measurable voltage signal, which is then fed into a signal conditioning circuit. Based on the temperature-resistance characteristic curve of the NTC thermistor, the acquired resistance signal is first converted into a corresponding temperature value. Then, a sliding window algorithm is used to calculate the temperature change per unit time, obtaining ΔT / Δt, where ΔT is the temperature difference and Δt is the calculation period, typically set to 10~100ms.

[0052] A high-voltage differential probe is connected in parallel across the load interface. The probe must have a high common-mode rejection ratio to resist power supply circuit interference. It directly acquires the instantaneous voltage across the interface and outputs a differential signal to the subsequent processing unit. The voltage signal output by the high-voltage differential probe is attenuated and filtered before being synchronously sampled by an ADC. Similarly, the continuous voltage value is processed through differential operations to obtain ΔV / Δt, where ΔV is the voltage drop and Δt is the calculation period. During this process, a software algorithm is used to eliminate baseline drift caused by power supply voltage fluctuations.

[0053] The above embodiments, through the combination of a high-frequency current sensor and a high-speed ADC, can capture the sudden current change during hot-plugging, thereby improving the response speed; through cross-validation of three types of dynamic parameters, di / dt, ΔT / Δt, and ΔV / Δt, the interference of single parameter fluctuations can be effectively eliminated.

[0054] In some embodiments, it is first determined whether the current change rate is greater than a first current change rate threshold. When the current change rate is greater than the first current change rate threshold, a current-limiting resistor is connected to the load power supply circuit to suppress the maximum peak current. The first preset change rate I... limit A / μs, Ilimit = Overcurrent protection value of the power supply chip / number of loads being powered, e.g., 20 A / μs. limit It can adapt to power supply scenarios with different load numbers, eliminating the need to redesign the current limiting threshold based on changes in load number. The current limiting resistor has a resistance of 10Ω. Connecting the current limiting resistor to the load power supply circuit can be defined as pre-charge mode. In pre-charge mode, the load has already undergone initial charging, thus avoiding cold start impact.

[0055] The fan power supply circuit includes a digitally controlled resistor array and a resistor-capacitor (RC) network. The digitally controlled resistor array includes digital potentiometers and a MOSFET array; the RC network includes capacitor banks and an RC topology. The digitally controlled resistor array can flexibly adjust the equivalent resistance value, providing a stable 10Ω resistance during current limiting and dynamically adjusting according to current changes after load startup. For example, it can reduce the current-limiting resistance or remove current limiting after the load enters steady-state operation, avoiding power loss and heat generation caused by continuous current limiting. The RC network helps suppress voltage spikes during current surges and buffers circuit impedance changes when the current-limiting resistor is connected, reducing the impact on the starting performance of the load motor and preventing load startup failure or speed fluctuations due to instantaneous impedance surges.

[0056] When the rate of change of current is less than or equal to a first preset rate of change, determine whether the contact heating rate is greater than a preset heating threshold. The preset heating rate can be 10°C / μs. If so, continue to step S202.

[0057] The above embodiment first uses a first current change rate threshold as the trigger condition, and only intervenes to adjust when the current surge reaches the risk threshold, avoiding unnecessary current limiting during normal operation that could affect load operation. Once the trigger condition is met, a 10Ω current-limiting resistor is quickly connected to the load power supply circuit, directly suppressing the maximum peak current during hot-plugging. Combined with the microsecond-level response of the previously used high-frequency current sensor, current limiting can be completed before the surge current reaches the power chip's overcurrent protection value, significantly reducing the probability of the power chip triggering protection or being damaged due to overcurrent. The current-limiting resistor is adjusted via a digitally controlled resistor array, offering a faster response than pure software current limiting. Even if the digital signal processing unit experiences a short-term fault, the hardware current limiting mechanism can still operate independently, further improving the operational reliability of the load power supply circuit in hot-plugging scenarios.

[0058] When the rate of change of current is less than or equal to a first current rate of change threshold, or when the rate of change of current is less than or equal to the first current rate of change threshold and the contact heating rate is less than or equal to a preset heating threshold, or when the contact heating rate is less than or equal to a preset heating threshold, the load power supply circuit is switched to steady-state mode. Steady-state mode is based on the pre-charge mode with a preset RC network connected in parallel, for example, a RC network containing a 5Ω resistor and a 100nF capacitor. The 5Ω resistor stabilizes the circuit impedance and suppresses minor current fluctuations during steady-state load operation, while the 100nF capacitor quickly absorbs voltage spikes, such as voltage fluctuations caused by instantaneous grid disturbances or slight power changes in the load, keeping the supply voltage ripple within a lower range, typically below 10mV. Standardized RC parameters reduce hardware design complexity, eliminating the need to customize RC specifications for different loads. Furthermore, the parallel structure of the RC network facilitates parameter adjustments based on load characteristics (e.g., the capacitor can be increased to 220nF in high-interference environments), providing excellent scalability.

[0059] From a technical perspective, on-demand switching is achieved through multi-dimensional condition determination, balancing protection and energy consumption. Three types of switching conditions are set: current change rate meets the standard, both current change rate and contact heating rate meet the standard, and contact heating rate meets the standard. This avoids protection deficiencies caused by premature switching; it can also flexibly adapt to different load states. For example, low-power loads may only require the contact heating rate to meet the standard, while high-power loads require the current change rate to assist in the determination, preventing switching delays caused by over-reliance on a single condition.

[0060] S202. Input the current change rate, contact heating rate and voltage change rate into the surge probability prediction model to obtain the surge occurrence probability output by the surge probability prediction model.

[0061] The surge probability prediction model is built upon a Long Short-Term Memory (LSTM) network and trained using historical datasets. Optionally, the historical dataset includes historical current change rates, historical contact heating rates, historical load interface voltages, historical surge markers, historical ambient temperatures, and historical impedance parameters of the load power supply circuit. The training process for the surge probability prediction model includes: first, acquiring the historical dataset; then, using the historical dataset as training samples to adjust the initial model built on the LSTM network to obtain a converged surge probability prediction model.

[0062] Specifically, the historical dataset is first preprocessed, including but not limited to outlier removal, missing value imputation, standardization, and labeling. For example, extreme data such as outliers with current change rates far exceeding physical limits and negative impedance values ​​caused by sensor malfunctions are removed using the 3σ principle or box plot method to avoid interfering with model training; short-term missing data are imputed using linear interpolation, and long-term missing segments are marked as "invalid samples" and removed; current change rate, heating rate, voltage, ambient temperature, and impedance are standardized using Z-Score to a dataset with a mean of 0 and a variance of 1, eliminating the impact of dimensional differences on the LSTM model weight update; and "historical surge markers" are converted into binary labels, where 1 indicates a surge occurred and 0 indicates no surge occurred, ensuring a one-to-one correspondence between the labels and the timestamps of the feature data.

[0063] Next, a sequence window is constructed, and the preprocessed historical dataset is split. Based on the temporal characteristics of surge events (e.g., hot-plug surges typically occur within 10-100 μs), the sequence window length is set, and a sliding window is used to generate training samples, ensuring the model can predict future surges based on historical time-series features. The preprocessed historical dataset is divided into a training set for updating model parameters, a validation set for adjusting hyperparameters, and a test set for evaluating final performance, in a 7:2:1 ratio. Temporal continuity must be maintained during the division; for example, the first 70% is used as the training set, the middle 20% as the validation set, and the last 10% as the test set, to avoid data leakage.

[0064] The initial model is then built based on LSTM, with a core structure consisting of an input layer, an LSTM layer, a fully connected layer, and an output layer. The input layer's input dimensions include batch size, sequence window length, and the number of features. The LSTM layer can have two LSTM units; the first layer enables the return sequence and outputs a time-series sequence, while the second layer disables the return sequence and outputs a single time-series feature vector. A dropout layer is also added to prevent overfitting. The fully connected layer maps the feature vector output by the LSTM layer to low-dimensional features. The output layer uses a single output unit with a sigmoid activation function, outputting the probability of a surge occurring after the current time window ends.

[0065] Further batch training is performed, stopping when the validation set loss converges. In each training round, the model calculates the predicted probability through forward propagation, compares it with the true label to calculate the loss value, and then updates the weights and biases of the LSTM and fully connected layers through back propagation through time (BPTT) to minimize the loss function. Key hyperparameters, such as the number of LSTM units, sequence window length, and dropout ratio, are adjusted based on the validation set performance, selecting the hyperparameter combination with the highest F1 score on the validation set. The loss, precision, and recall of the training and validation sets are monitored in real time. If the loss on the training set continues to decrease but the loss on the validation set increases, it indicates that the model is overfitting, and the dropout ratio needs to be increased or the number of LSTM units reduced. The learning rate change in each training round is recorded to ensure that the learning rate stabilizes at a low level in the later stages to avoid model oscillations in the later training phase. The trained model is then used for inference on the test set to calculate precision, recall, and the area under the receiver operating characteristic curve (AUC), which evaluates the discrimination ability of the binary classification model. The overall performance of the model is quantified by plotting the ROC curve with the false positive rate (FPR, i.e., the proportion of non-surge events that are mistakenly identified as surges) on the horizontal axis and the true positive rate (TPR, i.e., the proportion of surge events that are correctly identified as surges) on the vertical axis, and calculating the AUC between the curve and the horizontal axis. An AUC ≥ 0.95 means that the model can effectively distinguish between surge and non-surge events.

[0066] The training process of the aforementioned surge probability prediction model, through the optimization of the entire process of data, model, training and verification, can accurately capture surge events with a low false positive rate, providing reliable early warning for surge protection in scenarios such as hot-plugging of circuits; the monitoring, optimization and verification mechanism of the entire process ensures stable model training and performance compliance, reduces the uncertainty risk in practical applications, and provides a solid predictive foundation for subsequent protection measures.

[0067] In some embodiments, when the rate of change of current is greater than a first rate of change of current threshold, the rate of change of current (di / dt), the contact heating rate (ΔT / Δt), and the rate of change of voltage (ΔV / Δt) are input into the surge probability prediction model to obtain the surge occurrence probability output by the surge probability prediction model.

[0068] In other embodiments, when the rate of change of current is less than or equal to a first threshold for the rate of change of current, and the contact heating rate is greater than a preset threshold for heating, the rate of change of current (di / dt), the contact heating rate (ΔT / Δt), and the rate of change of voltage (ΔV / Δt) are input into the surge probability prediction model to obtain the surge probability output by the surge probability prediction model. The contact heating rate can be 10℃ / μs. Using this as a trigger condition allows for timely intervention when a dangerous heating trend occurs at the contact, while avoiding frequent model triggering due to a threshold that is too low (e.g., 5℃ / μs) or a threshold that is too high (e.g., 15℃ / μs) leading to intervention lag.

[0069] In the above embodiments, if the current change rate is less than or equal to a first current change rate threshold and the contact heating rate is greater than a preset heating threshold, it is initially predicted to be a surge precursor. At this time, model prediction is triggered, reducing the model call frequency in non-surge scenarios and reducing the consumption of computing resources. di / dt, ΔT / Δt, and ΔV / Δt are instantaneous dynamic characteristics before a surge occurs. The combination of the three allows the model to simultaneously capture long-term environmental influences and instantaneous risk mutations. By linking the three, the model can accurately determine the surge probability and avoid missed detections.

[0070] After applying the surge probability prediction model to infer the surge occurrence probability, it is determined whether the surge occurrence probability is greater than or equal to a first preset probability. If so, the load power supply circuit is upgraded from pre-charging mode to steady-state mode. For example, assuming the surge occurrence probability is greater than or equal to 90%, a 10Ω resistor is connected to the load power supply circuit as a basic protection component, and a 5Ω resistor is connected in parallel with a 100nF capacitor. This combination of the 10Ω resistor, 5Ω resistor, and 100nF capacitor under a probability greater than 90% provides the following protection: the 10Ω resistor handles residual surges before the upgrade, the 5Ω resistor meets the low-loss requirements of steady-state power supply, and the 100nF capacitor ensures power supply stability, further enhancing current limiting capability and filtering effect, preventing surge peak current from damaging circuit components.

[0071] If the surge probability is less than the first preset probability, then it is determined whether the surge probability is greater than or equal to the second preset probability, where the second preset probability is less than the first preset probability. For example, the first preset probability is 90%, and the second preset probability is 80%. When the surge probability is greater than or equal to the second preset probability but less than the first preset probability, a current-limiting resistor is connected to the drive load power supply circuit. When the surge probability is greater than 80% but less than 90%, a 10Ω current-limiting resistor is connected to avoid excessive circuit impedance and a surge in power loss due to over-upgrading.

[0072] Here, a second preset probability is set as a medium-risk threshold. When the surge probability is in the medium-risk range of 80%-90%, a current-limiting resistor is connected to the active drive circuit. This proactive preventative measure effectively suppresses potential surge currents, preventing them from reaching the eFuse trigger threshold and reducing the possibility of eFuse malfunction from the source. Probability-based tiered control allows for targeted measures at the initial stage of risk, avoiding overprotection in low-risk scenarios and reducing surge intensity in high-risk scenarios through the early connection of the current-limiting resistor. This makes the protection response more accurate and timely, compensating for the lag in the eFuse thermal shutdown function.

[0073] S203. When the surge probability is greater than or equal to the first preset probability, calculate the resistance and capacitance parameters of the load power supply circuit based on the monitored load interface current, load interface voltage and load interface temperature.

[0074] This application monitors the rate of change of current, the rate of increase of contact temperature, and the rate of change of voltage in real time. Utilizing a surge probability prediction model based on a long short-term memory network, it pre-determines the probability of a surge and intervenes when the probability of a surge is greater than or equal to a first preset probability, thus avoiding false triggering caused by the passive response of eFuse. When a high surge risk is predicted, the application actively adjusts the circuit state before the surge occurs by calculating and controlling the resistance and capacitance parameters of the load power supply circuit in real time. Compared to the passive mechanism of eFuse, which relies on its own temperature rise to trigger protection, this significantly improves the response speed and can promptly address instantaneous changes in rapid hot-plugging scenarios, preventing equipment damage caused by untimely protection.

[0075] In some embodiments, when the probability of a surge occurrence is greater than or equal to a first preset probability, the load interface current is first subjected to spectrum analysis to extract the surge frequency f; then the target impedance is calculated based on the load interface voltage and the limiting current; then the capacitance value of the load power supply circuit is calculated based on the surge frequency f and the target impedance; and finally, the resistance value of the load power supply circuit is calculated based on the temperature compensation coefficient corresponding to the load interface voltage, the target current and the load interface temperature.

[0076] Among them, the surge dominant frequency is the frequency component with the strongest energy in the surge current. Target impedance Z target =V / I limit V is the load interface voltage; I limit = Overcurrent protection OCP value of the power chip ÷ Number of loads supplied, representing the maximum allowable current in the load power supply circuit.

[0077] Specifically, the current spectrum is analyzed using a PID control chip via Fast Fourier Transform to find the surge frequency f. Then, the target impedance is calculated based on the load interface voltage and the limiting current. Finally, the target impedance Z is used to... targetGiven the surge frequency f, calculate the capacitance C of the load power supply circuit according to the following formula (1). opt :

[0078] (1)

[0079] According to the following formula (2), based on the load interface voltage V and the target current I... target Temperature compensation coefficient K corresponding to the load interface temperature temp Calculate the resistance value R of the load power supply circuit. opt :

[0080] (2)

[0081] In formula (2), the target current I target This is the rated current of the load power supply circuit during normal operation. Temperature compensation coefficient K. temp Used to correct for the effect of temperature on resistance, ensuring that the calculated resistance value meets the requirements at different temperatures.

[0082] In the above embodiment, when the surge probability is greater than or equal to a first preset probability, the current spectrum is first analyzed using the Fast Fourier Transform (FFT) of the PID control chip to extract the surge frequency f. This avoids capacitor and resistor mismatch caused by using fixed frequency parameters. The PID control chip ensures that the calculation of the RC parameters is completed before the surge energy erupts, allowing sufficient time for subsequent RC network adjustments and preventing protection delays due to excessive analysis time. Through spectrum analysis, frequency extraction, and RC calculation, the RC parameters are dynamically adjusted according to surge characteristics, load status, and temperature, improving the surge suppression effect. This parameter-customized control based on surge characteristics can specifically suppress the energy concentration frequency band of the surge current, reducing surge intensity from the source and preventing it from reaching the eFuse trigger threshold, thus reducing the possibility of malfunction. The entire control process is achieved through a proactive mechanism of real-time monitoring of current, voltage, and temperature, Fast Fourier Transform, and precise calculation of RC parameters, ultimately resulting in dynamic adjustment. Compared to the passive triggering mode of eFuse, which relies on its own temperature rise, the response speed is significantly improved. In particular, by introducing target impedance and temperature compensation coefficient, it ensures that the power supply circuit parameters are always in the optimal protection state during the instantaneous changes of rapid hot-swapping, thus solving the problem of untimely protection caused by eFuse response lag.

[0083] In some embodiments, when the surge probability is greater than or equal to a first preset probability, firstly, a spectrum analysis is performed on the load interface current to extract the surge dominant frequency f; then, the target impedance is calculated based on the load interface voltage and the limiting current, and finally, based on the target impedance Z... target And the surge frequency f, the solution satisfies The resistance and capacitance parameters. Because resistors consume power (P=I). 2R), provided that the impedance requirements are met, a lower resistance and capacitance parameter can be selected to regulate the load power supply circuit.

[0084] The above embodiments extract the surge frequency f through spectrum analysis and calculate the RC parameters that meet the target impedance based on the impedance formula. This accurately matches the frequency characteristics of the surge, ensuring that the total impedance Z after the RC series connection equals the target impedance. This avoids insufficient or excessive surge suppression due to total impedance deviation. This customized impedance design effectively attenuates the energy of the surge current, preventing it from reaching the eFuse trigger threshold and reducing malfunctions at the source. Simultaneously, selecting a smaller resistor while meeting impedance requirements ensures surge suppression while avoiding normal operation interference caused by excessive current limiting, resolving the contradiction between eFuse protection and normal equipment operation. Compared to the passive triggering mechanism of eFuse, which relies on its own temperature rise, this forward-looking control based on surge characteristics has a faster response speed, completing circuit adjustment before the surge energy spreads. This compensates for the protection hysteresis defect caused by the thermal conduction delay of eFuse, making it particularly suitable for rapid transient scenarios during fan hot-plugging.

[0085] S204. Adjust the load power supply circuit according to the resistance and capacitance parameters.

[0086] The RC parameters include resistance and capacitance values. The load power supply circuit includes a digitally controlled variable resistor array and an RC network. The digitally controlled variable resistor array consists of digital potentiometers and a power MOSFET array, capable of rapidly switching resistance values ​​to limit current amplitude. The RC network consists of capacitor banks and an RC topology. The capacitor banks support switching capacitance values ​​via control circuitry, and the RC topology refers to the connection method of the resistors and capacitors. The RC network can dynamically match RC parameters, suppress inrush current, and reduce unstable oscillations in the circuit caused by high-frequency signals.

[0087] The numerically controlled resistor array of the load power supply circuit is adjusted according to the resistance value; the RC network of the load power supply circuit is adjusted according to the capacitance value. Specifically, when the impedance formula calculates the target resistance value R... opt Subsequently, the main control unit sends an 8-bit control command to the numerically controlled resistor array via the SPI / I2C communication protocol. The first 4 bits control the MOSFET switching state, and the last 4 bits control the digital potentiometer resistance value. Upon receiving the command, the MOSFET switch completes its on / off state within 100ns, and the digital potentiometer synchronously adjusts its sliding contact position. This combines coarse adjustment of the fixed resistor combination with fine adjustment of the digital potentiometer to achieve the target resistance value, ultimately controlling the resistance error within ±0.01Ω. Based on the calculated C... optWith the real-time main frequency f, control signals are sent through the GPIO port to drive the analog switch to complete the turn-on / turn-off within 50ns, and the target capacitance value is obtained. At the same time, according to the voltage-capacitance characteristics of the varactor diode, its reverse bias voltage is adjusted to achieve pF-level fine adjustment, and the final capacitance error is controlled within ±0.1nF.

[0088] For example, in a scenario with f=2MHz and Ztarget=6Ω, a precise combination of R=5Ω and C=13nF can reduce the surge peak current from 30A to 2A (I=V / Z=12V / 6Ω=2A), which is completely controlled within the power chip's OCP value (e.g., 4A), thus avoiding overcurrent protection triggering.

[0089] In any of the above embodiments, the system bus voltage is monitored, and the voltage difference between the system bus voltage and the load interface voltage is calculated. If the voltage difference is less than or equal to a voltage difference threshold and the current change rate is less than a second current change rate threshold for a preset duration, the resistance value of the load power supply circuit in steady-state mode is reduced.

[0090] Among them, the system bus voltage V bus It is the energy source for powering the load, and the load interface voltage V fan This is the actual supply voltage. The voltage difference threshold can be set to 0.8V, the second current change rate threshold can be set to 1Aμs, and the preset duration can be set to 10μs.

[0091] Specifically, monitor the system bus voltage V bus Calculate the system bus voltage V bus With load interface voltage V fan The voltage difference ΔV between them. When ΔV ≤ 0.8V and the current change rate di / dt < 1A / μs over 10μs, it indicates that the voltage drop from the bus to the load is small, reducing the resistance of the load power supply circuit from 5Ω to 1Ω. The bus has sufficient power supply capability, and reducing the resistance will not cause V to drop due to the increased voltage drop across the resistor. fan Too low.

[0092] The voltage difference is ≤ voltage difference threshold and the current change rate within a preset time is < second current change rate threshold as the trigger condition for reducing the resistance value. This ensures that the resistance adjustment is only performed in scenarios where the power supply is stable and there is no surge risk, avoids circuit safety hazards caused by misoperation, reduces power loss, improves power supply efficiency, and ensures the stability of load operation.

[0093] If the voltage difference is greater than the voltage difference threshold, or the current change rate is less than the second current change rate threshold for a duration shorter than the preset duration, or the current change rate is greater than the second current change rate threshold, then the steady-state mode of the load power supply circuit is maintained for a delay of 50μs, and the voltage difference and current change rate are re-evaluated after 50μs.

[0094] If ΔV > 0.8V, it indicates insufficient bus power supply. Blindly reducing the resistance will further increase the current (I = Vfan / R), causing ΔV to continue to expand and even triggering bus overcurrent protection. Therefore, a 50μs delay is used to maintain the steady-state mode of the load power supply circuit. Assuming the current change rate is stable at 0.5A / μs, a 10μs delay indicates that the circuit has no instantaneous surges and can maintain the steady-state mode of the load power supply circuit. If the current change rate briefly drops below 1A / μs but does not reach 10μs, it may be a current adjustment before a surge. In this case, not adjusting the resistance can prevent current runaway due to insufficient resistance during a surge.

[0095] In some embodiments, the remaining surge energy of the load power supply circuit is monitored. If the remaining surge energy exceeds a preset energy threshold, the remaining surge energy is stored in an energy storage capacitor bank through an energy feedback circuit. Optionally, the remaining current i is collected by a high-frequency current sensor, and the remaining surge energy is calculated in real time in conjunction with the circuit resistance R.

[0096] Among them, the residual surge energy is the instantaneous energy remaining after surge suppression; the preset energy threshold can be 30%. The 30% setting ensures that the recovered energy has practical utilization value, while avoiding the fact that the startup cost of the feedback circuit is higher than the recovery benefit due to the low residual energy.

[0097] Energy feedback circuits can be H-bridge circuits, such as... Figure 3 As shown, when the remaining surge energy exceeds a preset energy threshold, control paths S1 and S4 are activated, transferring the remaining surge energy from the load power supply circuit to the energy storage capacitor bank. The energy storage capacitor bank can be a supercapacitor. A supercapacitor is a new type of energy storage device that falls between traditional capacitors and batteries. Its core characteristics are a much higher specific capacity than ordinary capacitors and a much faster charging and discharging speed than batteries. Its energy storage principle is based on the double-layer capacitance at the electrode-electrolyte interface or the Faraday pseudocapacitance on the electrode surface, enabling millisecond-level rapid charging and discharging, a cycle life of tens of thousands to hundreds of thousands of cycles, and a wide operating temperature range.

[0098] Residual surge energy monitoring can directly reflect the energy status. This closed loop can avoid secondary impacts of residual surge energy on the circuit and improve the utilization rate of surge energy.

[0099] When the remaining surge energy is less than or equal to a preset energy threshold, the load power supply circuit is adjusted to steady-state mode. A remaining surge energy of ≤30% means that the residual surge energy has been reduced to a safe range, and at this time, there is no possibility of secondary energy impact on the circuit.

[0100] For example, if the total surge energy is 100mJ, and the remaining 25mJ (≤30mJ) can be released through a 5Ω resistor, it will only cause a brief temperature rise of 0.3℃ and will not damage the load. During surge protection, a high-resistance configuration is used to suppress peak energy, which effectively limits current but results in higher power loss. A low-resistance configuration, adjusted to steady-state mode, can reduce losses. Combined with capacitor filtering to reduce voltage fluctuation losses, this lowers the overall power consumption of the circuit after a surge.

[0101] In the above embodiments, when the remaining energy is >30%, it is preferentially recovered to the supercapacitor bank through the H-bridge circuit to avoid energy waste; when the energy is ≤30% and the recovery value is low, it directly switches to steady-state mode to avoid the startup cost of the feedback circuit. Both cover the entire scenario of high energy recovery and low energy steady state, so that the energy utilization rate after the surge can be improved.

[0102] After storing a certain amount of surge energy in the energy storage capacitor bank, when the system bus voltage is detected to be lower than the voltage threshold, or the load operating parameters are lower than the required parameters, the energy in the energy storage capacitor bank is transferred to the load power supply circuit through the energy feedback circuit. For example, when the system bus voltage is less than 11V, or the fan speed is less than the preset speed, the energy feedback circuit controls the conduction paths S2 and S3, and the energy storage capacitor bank releases energy. When the system bus voltage briefly drops to 11V, the load power supply may be insufficient; the energy storage capacitor bank releases energy to maintain a stable load voltage, avoiding triggering the resistor adjustment protection action. When the fan speed is less than the preset speed, the energy feedback can directly supplement the load current to push the speed back to the rated value, avoiding a decrease in heat dissipation efficiency due to insufficient speed and ensuring normal load function.

[0103] When a back electromotive force (EMF) is detected at the load interface, the back EMF is suppressed by controlling the current direction of the energy feedback circuit. For example, when the fan is unplugged, a -15V back EMF is generated in the power supply circuit. The energy feedback circuit's conduction paths S2 and S4 are controlled to direct the energy corresponding to the back EMF into the energy storage capacitor bank, suppressing the voltage spike. This achieves active suppression and reuse of the back EMF.

[0104] When inductive loads such as fans are unplugged, the inductor generates a -15V back electromotive force due to the sudden drop in current. If not handled properly, this could damage the MOSFETs and power supply chips in the power supply circuit. However, the conduction of paths S2 and S4 can establish a reverse energy path from the load power supply circuit to the energy feedback circuit and then to the energy storage capacitor bank within 1-2μs. This quickly introduces the energy corresponding to the -15V back electromotive force into the energy storage capacitor bank, suppressing voltage spikes and improving energy recovery rate.

[0105] Based on any of the above embodiments, after the hot-plug event is completed, the load power supply circuit is switched to a low-power mode. For example, in low-power mode, the resistance R of the load power supply circuit is 0.1Ω, and the capacitance C is 470μF. After hot-plugging, the circuit enters a long-term steady state and is switched to a low-power mode, reducing long-term operating costs and avoiding circuit temperature rise caused by high resistance heating, thus extending component lifespan. Although there is no sudden surge after hot-plugging, there may be minor fluctuations in the power grid and slight changes in load power. The large capacitance of the 470μF capacitor provides strong energy buffering capacity, preventing performance deviations in the load due to power supply fluctuations.

[0106] In some embodiments, parameters, including but not limited to impedance parameters and residual surge energy, are uploaded to a database during the regulation of the load power supply circuit to update the surge probability prediction model.

[0107] In summary, the load power supply circuit control method provided in this application comprehensively reflects the rapid changes in the circuit state at the moment of hot-plugging by acquiring three key indicators: the rate of change of current at the load interface, the rate of increase of contact temperature, and the rate of change of voltage. These rates of change are then input into a surge probability prediction model to obtain the surge probability. Unlike traditional methods that rely solely on the eFuse to passively handle overcurrent and overvoltage, this method can predict surge risks in advance. Since the instantaneous surge current generated during hot-plugging is a key factor causing eFuse malfunction, by predicting the surge probability in advance, measures can be taken to prevent the eFuse from malfunctioning due to surge impact when a surge is highly probable, thereby protecting the circuit from damage.

[0108] When the surge probability is greater than or equal to the first preset probability, it indicates a potentially significant surge risk. In this case, the method calculates the resistance and capacitance parameters of the load power supply circuit based on the monitored load interface current, load interface voltage, and load interface temperature. When power supply fluctuations are large or the design is unstable, the eFuse's overvoltage protection capability is insufficient. However, by calculating the resistance and capacitance parameters in real time, the circuit's impedance characteristics can be flexibly adjusted. For example, when the power supply voltage rises, by changing the capacitor or resistor values, the circuit current can be adjusted to share some of the overvoltage energy, compensating for the eFuse's shortcomings in overvoltage protection under complex power supply environments and providing more comprehensive overvoltage protection for the load power supply circuit.

[0109] The key execution step of this method is to regulate the load power supply circuit based on the calculated resistance and capacitance parameters. Unlike the eFuse thermal shutdown function, which relies on the internal temperature of the IC and has a slow response speed, this regulation is real-time and based on dynamic calculation of multiple parameters. For example, in the event of rapid hot-plugging or short-circuit events, the circuit can be quickly adjusted according to the calculated resistance and capacitance parameters. This can be achieved by changing the connection method of capacitors or resistors through switching elements, rapidly changing the circuit characteristics, and timely suppressing abnormal changes in current and voltage. This greatly improves the protection response speed, effectively avoids equipment damage caused by untimely protection, and ensures stable operation of the server under complex conditions such as load hot-plugging.

[0110] like Figure 4 As shown, Figure 4 This is a flowchart illustrating the RC parameters of the dynamically switching load power supply circuit provided in this application embodiment. First, the current change rate, contact heating rate, and voltage change rate of the load interface are obtained. Then, it is determined whether the current change rate of the load interface is greater than or equal to a first current change rate threshold I. limit If so, it means the load power supply circuit meets the pre-charge trigger condition, and a 10Ω current-limiting resistor is first connected; this is the initial current limiting process. Next, the current change rate, contact heating rate, and voltage change rate of the load interface are input into the surge probability prediction model to obtain the surge probability. It is then determined whether the surge probability is greater than or equal to the second preset probability of 80%. If so, it means the load power supply circuit meets the pre-charge trigger condition, and a 10Ω current-limiting resistor is connected. Then, it is determined whether the surge probability is greater than or equal to the first preset probability of 90%. If so, a 5Ω resistor is connected in parallel with the load power supply circuit, and a 100nF capacitor is also connected. Then, it is determined whether the steady-state switching condition is met: whether the voltage difference between the system bus voltage and the load interface voltage is less than or equal to the voltage difference threshold of 0.8V, and whether the current change rate within a preset time of 10μs is less than the second current change rate threshold of 1A / μs. If the voltage difference is ≤0.8V and less than 1A / μs within 10μs, the steady-state switching condition is met, and the load power supply circuit is switched to connect a 1Ω resistor. If the steady-state switching conditions are not met, the configuration of the 5Ω resistor and 100nF capacitor is maintained, and the steady-state switching conditions are reassessed after a 50μs delay. The above process, by adjusting the resistor and capacitor parameters in stages and combining indicators such as current change rate, power prediction, and voltage change rate, ensures that the circuit transitions smoothly from pre-charging to steady-state operation, avoids the impact of sudden current and voltage changes on circuit components, and optimizes energy efficiency and stability.

[0111] like Figure 5 As shown, Figure 5 A flowchart illustrating the load power supply circuit control method provided in this application embodiment. Figure 2When a hot-plug event occurs, real-time signals such as current, temperature, and voltage of the load power supply circuit are collected using high-frequency current sensors, temperature sensors, and voltage sensors. The collected real-time signals are preprocessed, and then the current change rate di / dt, contact heating rate, and voltage change rate are calculated and their spectra analyzed. Subsequently, it is determined whether the current change rate exceeds a first current change rate threshold of 20 A / μs. If so, the current change rate, contact heating rate, and voltage change rate are input into a surge probability prediction model to obtain the surge probability. If not, it is further determined whether the contact heating rate exceeds a preset heating rate threshold of 10 °C / μs. If the contact heating rate exceeds the preset heating rate threshold, the current change rate, contact heating rate, and voltage change rate are input into the surge probability prediction model to obtain the surge probability. If the contact heating rate is less than or equal to the preset heating rate threshold, a steady-state mode (parallel connection of a 5Ω resistor and a 100nF capacitor) is maintained.

[0112] When the surge probability is greater than or equal to 90%, the PID control chip calculates the resistance and capacitance parameters R and C of the load power supply circuit based on the monitored load interface current, load interface voltage, and load interface temperature. The numerically controlled resistor array and RC network of the load power supply circuit are then adjusted based on these resistance and capacitance parameters.

[0113] The remaining surge energy is then monitored to determine if it exceeds the energy threshold. If so, the remaining surge energy is recovered and stored in the supercapacitor; otherwise, the load power supply circuit is adjusted to a steady-state mode. Through the above description of the implementation methods, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented using software plus necessary general-purpose hardware platforms. Of course, they can also be implemented using hardware, but in many cases, the former is a better implementation method.

[0114] like Figure 6 As shown, embodiments of this application also provide a control device for a load power supply circuit, the device comprising:

[0115] The acquisition module 601 is used to acquire the current change rate, contact heating rate and voltage change rate of the load interface when the baseboard management controller detects a hot-plug event of the load; wherein, the hot-plug event is identified by the level signal of the presence detection pin of the load power supply circuit.

[0116] The surge prediction module 602 is used to input the current change rate, contact heating rate and voltage change rate into the surge probability prediction model to obtain the surge occurrence probability output by the surge probability prediction model; wherein, the surge probability prediction model is constructed based on a long short-term memory network and trained according to historical datasets;

[0117] The calculation module 603 is used to calculate the resistance and capacitance parameters of the load power supply circuit based on the monitored load interface current, load interface voltage and load interface temperature when the surge occurrence probability is greater than or equal to the first preset probability.

[0118] The control module 604 is used to control the load power supply circuit according to the resistance and capacitance parameters.

[0119] As an optional implementation provided in this application, the acquisition module 601 is specifically used to: monitor the load interface current, load interface temperature and load interface voltage when the baseboard management controller detects a hot-plug event of the load; calculate the current change rate based on the load interface current, calculate the contact heating rate based on the load interface temperature and calculate the voltage change rate based on the load interface voltage.

[0120] As an optional implementation provided in this application, the acquisition module 601 is further configured to: determine whether the current change rate is greater than the first current change rate threshold; the control module 604 is further configured to: drive the load power supply circuit to connect a current limiting resistor when the current change rate is greater than the first current change rate threshold.

[0121] As an optional implementation provided in this application, the surge prediction module 602 is also used to acquire historical datasets, which include historical current change rate, historical contact heating rate, historical load interface voltage, historical surge markers, historical ambient temperature, and historical impedance parameters of the load power supply circuit; and adjust the initial model constructed based on the long short-term memory network according to the historical datasets to obtain a surge probability prediction model.

[0122] As an optional implementation provided in this application, the surge prediction module 602 is further used to determine whether the current change rate is greater than a first current change rate threshold; if the current change rate is less than or equal to the first current change rate threshold, it determines whether the contact heating rate is greater than a preset heating threshold, so that if the contact heating rate is greater than the preset heating threshold, the current change rate, the contact heating rate and the voltage change rate are input into the surge probability prediction model.

[0123] As an optional implementation provided in this application, the surge prediction module 602 is further configured to drive the load power supply circuit to upgrade from the pre-charging mode to the steady-state mode when the surge occurrence probability is greater than or equal to the first preset probability; wherein, the steady-state mode is based on the pre-charging mode with a preset resistor-capacitor network connected in parallel.

[0124] As an optional implementation provided in this application, the control module 604 is further configured to: switch the load power supply circuit to steady-state mode when the current change rate is less than or equal to a first current change rate threshold and / or the contact heating rate is less than or equal to a preset heating threshold.

[0125] As an optional implementation provided in this application, the control module 604 is further configured to: drive the load power supply circuit to connect a current-limiting resistor when the surge occurrence probability is greater than or equal to a second preset probability and less than a first preset probability; wherein the second preset probability is less than the first preset probability.

[0126] As an optional implementation provided in this application, the calculation module 603 is specifically used for: performing spectrum analysis on the load interface current and extracting the surge frequency when the surge occurrence probability is greater than or equal to a first preset probability; calculating the target impedance based on the load interface voltage and the limiting current; calculating the capacitance value of the load power supply circuit based on the surge frequency and the target impedance; and calculating the resistance value of the load power supply circuit based on the temperature compensation coefficient corresponding to the load interface voltage, the target current, and the load interface temperature.

[0127] As an optional implementation provided in this application, the control module 604 is specifically used for: controlling the numerically controlled resistor array of the load power supply circuit according to the resistance value; and controlling the resistor-capacitor network of the load power supply circuit according to the capacitance value.

[0128] As an optional implementation provided in this application, the control module 604 is also used to: monitor the system bus voltage; calculate the voltage difference between the system bus voltage and the load interface voltage; and if the voltage difference is less than or equal to the voltage difference threshold and the current change rate is less than the second current change rate threshold for a preset duration, then reduce the resistance value of the load power supply circuit in steady state mode.

[0129] As an optional implementation provided in this application, the device further includes an energy feedback module for: monitoring the remaining surge energy of the load power supply circuit; and storing the remaining surge energy into an energy storage capacitor bank through an energy feedback circuit when the remaining surge energy is greater than a preset energy threshold.

[0130] As an optional implementation provided in this application, the control module 604 is further configured to: control the load power supply circuit to a steady-state mode when the remaining surge energy is less than or equal to a preset energy threshold.

[0131] As an optional implementation provided in this application, the energy feedback module is also used to: transfer the energy in the energy storage capacitor bank to the load power supply circuit through the energy feedback circuit when the system bus voltage is detected to be less than the voltage threshold or the load operating parameters are less than the required parameters.

[0132] As an optional implementation provided in this application, the energy feedback module is further used to: suppress the reverse electromotive force by controlling the current direction of the energy feedback circuit when the load interface voltage is detected to be a reverse electromotive force.

[0133] As an optional implementation provided in this application, the control module 604 is further configured to: control the load power supply circuit to a low power mode after the hot-plug event is processed.

[0134] For a description of the features of the control device for the load power supply circuit in the corresponding embodiment, please refer to the relevant description of the control method for the load power supply circuit in the corresponding embodiment, which will not be repeated here.

[0135] like Figure 7 As shown, embodiments of this application also provide an electronic device, including a memory 701 and a processor 702. The memory 701 stores a computer program, and the processor 702 is configured to run the computer program to execute the steps in any of the above-described embodiments of the load power supply circuit control method.

[0136] Embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above-described embodiments of the load power supply circuit regulation method when running.

[0137] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0138] The embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above-described load power supply circuit control method embodiments.

[0139] Embodiments of this application also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in any of the above-described embodiments of the load power supply circuit control method.

[0140] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0141] The above provides a detailed description of the method, apparatus, equipment, medium, and program product for regulating a load power supply circuit provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only intended to help understand the method and core ideas of this application. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A method for regulating a load power supply circuit, characterized in that, include: When the baseboard management controller detects a hot-plug event of the load, it acquires the current change rate, contact heating rate, and voltage change rate of the load interface; wherein, the hot-plug event is identified by the level signal of the presence detection pin of the load power supply circuit. The current change rate, the contact heating rate, and the voltage change rate are input into the surge probability prediction model to obtain the surge occurrence probability output by the surge probability prediction model; wherein, the surge probability prediction model is constructed based on a long short-term memory network and trained on a historical dataset, the historical dataset including historical current change rate, historical contact heating rate, historical load interface voltage, historical surge marker, historical ambient temperature, and historical impedance parameters of the load power supply circuit; When the surge occurrence probability is greater than or equal to the first preset probability, the resistance and capacitance parameters of the load power supply circuit are calculated based on the monitored load interface current, load interface voltage and load interface temperature. The load power supply circuit is adjusted according to the aforementioned resistance and capacitance parameters; After inputting the current change rate, the contact heating rate, and the voltage change rate into the surge probability prediction model to obtain the surge occurrence probability output by the surge probability prediction model, the method further includes: when the surge occurrence probability is greater than or equal to the first preset probability, driving the load power supply circuit to upgrade from the pre-charging mode to the steady-state mode; wherein, the steady-state mode is based on the pre-charging mode with a preset resistor-capacitor network connected in parallel. The preset RC network consists of a capacitor bank and an RC topology. The capacitor bank supports switching the capacitance value through a control circuit. The RC topology refers to the connection method of the resistor and capacitor. The RC network can dynamically match the RC parameters according to the load characteristics to suppress surge current.

2. The method according to claim 1, characterized in that, When the baseboard management controller detects a hot-plug event of the load, it acquires the current change rate, contact heating rate, and voltage change rate of the load interface, including: When the baseboard management controller detects a hot-plug event of the load, it monitors the load interface current, load interface temperature, and load interface voltage. The current change rate is calculated based on the load interface current, the contact heating rate is calculated based on the load interface temperature, and the voltage change rate is calculated based on the load interface voltage.

3. The method according to claim 1, characterized in that, After obtaining the current change rate, contact heating rate, and voltage change rate of the load interface, the method further includes: Determine whether the rate of change of current is greater than a first rate of change of current threshold. When the rate of change of current is greater than the first rate of change of current threshold, the load power supply circuit is driven to connect a current-limiting resistor.

4. The method according to claim 1, characterized in that, Before acquiring the current change rate, contact heating rate, and voltage change rate of the load interface when the substrate management controller detects a hot-plug event of the load, the method further includes: Obtain the historical dataset; The initial model built on the long short-term memory network is adjusted based on the historical dataset to obtain the surge probability prediction model.

5. The method according to claim 1, characterized in that, When the baseboard management controller detects a hot-plug event of the load, after acquiring the current change rate, contact heating rate, and voltage change rate of the load interface, and before inputting the current change rate, contact heating rate, and voltage change rate into the surge probability prediction model to obtain the surge occurrence probability, the method further includes: Determine whether the rate of change of current is greater than a first rate of change of current threshold. If the current change rate is less than or equal to a first current change rate threshold, determine whether the contact heating rate is greater than a preset heating threshold, so that if the contact heating rate is greater than the preset heating threshold, the current change rate, the contact heating rate, and the voltage change rate are input into the surge probability prediction model.

6. The method according to claim 5, characterized in that, The method further includes: If the current change rate is less than or equal to the first current change rate threshold, and / or the contact heating rate is less than or equal to the preset heating threshold, the load power supply circuit is switched to steady-state mode.

7. The method according to claim 1, characterized in that, The method further includes: When the surge occurrence probability is greater than or equal to the second preset probability and less than the first preset probability, the load power supply circuit is driven to connect a current limiting resistor. Wherein, the second preset probability is less than the first preset probability.

8. The method according to claim 1, characterized in that, When the surge occurrence probability is greater than or equal to a first preset probability, the resistance and capacitance parameters of the load power supply circuit are calculated based on the monitored load interface current, load interface voltage, and load interface temperature, including: When the surge occurrence probability is greater than or equal to the first preset probability, the load interface current is subjected to spectrum analysis to extract the surge main frequency; Calculate the target impedance based on the load interface voltage and the limiting current; Calculate the capacitance value of the load power supply circuit based on the surge frequency and the target impedance; The resistance value of the load power supply circuit is calculated based on the load interface voltage, target current, and temperature compensation coefficient corresponding to the load interface temperature.

9. The method according to claim 8, characterized in that, The step of regulating the load power supply circuit according to the resistance and capacitance parameters includes: The numerically controlled resistor array of the load power supply circuit is adjusted according to the resistance value; The resistor-capacitor network of the load power supply circuit is adjusted according to the capacitance value.

10. The method according to claim 1, characterized in that, The method further includes: Monitor system bus voltage; Calculate the voltage difference between the system bus voltage and the load interface voltage; If the voltage difference is less than or equal to the voltage difference threshold and the current change rate is less than the second current change rate threshold for a preset duration, then the resistance value of the load power supply circuit in steady-state mode will be reduced.

11. The method according to claim 1, characterized in that, After adjusting the load power supply circuit according to the resistance and capacitance parameters, the method further includes: Monitor the remaining surge energy of the load power supply circuit; If the remaining surge energy is greater than a preset energy threshold, the remaining surge energy is stored in an energy storage capacitor bank through an energy feedback circuit.

12. The method according to claim 11, characterized in that, The method further includes: When the remaining surge energy is less than or equal to the preset energy threshold, the load power supply circuit is adjusted to a steady-state mode.

13. The method according to claim 11, characterized in that, The method further includes: If the system bus voltage is detected to be lower than the voltage threshold, or the load operating parameters are lower than the required parameters, the energy in the energy storage capacitor bank is transferred to the load power supply circuit through the energy feedback circuit.

14. The method according to claim 11, characterized in that, The method further includes: When the load interface voltage is detected to be a reverse electromotive force, the reverse electromotive force is suppressed by controlling the current direction of the energy feedback circuit.

15. The method according to claim 1, characterized in that, The method further includes: After the hot-plug event is processed, the load power supply circuit is switched to a low-power mode.

16. A control device for a load power supply circuit, characterized in that, include: The acquisition module is used to acquire the current change rate, contact heating rate, and voltage change rate of the load interface when the baseboard management controller detects a hot-plug event of the load; wherein, the hot-plug event is identified by the level signal of the in-situ detection pin of the load power supply circuit. The surge prediction module is used to input the current change rate, the contact heating rate, and the voltage change rate into the surge probability prediction model to obtain the surge occurrence probability output by the surge probability prediction model; wherein, the surge probability prediction model is constructed based on a long short-term memory network and trained on a historical dataset, the historical dataset including historical current change rate, historical contact heating rate, historical load interface voltage, historical surge markers, historical ambient temperature, and historical impedance parameters of the load power supply circuit; The calculation module is used to calculate the resistance and capacitance parameters of the load power supply circuit based on the monitored load interface current, load interface voltage and load interface temperature when the surge occurrence probability is greater than or equal to the first preset probability. The control module is used to control the load power supply circuit according to the resistance and capacitance parameters; The surge prediction module is further configured to drive the load power supply circuit to upgrade from a pre-charging mode to a steady-state mode when the surge occurrence probability is greater than or equal to the first preset probability; wherein, the steady-state mode is based on the pre-charging mode with a preset RC network connected in parallel; the preset RC network consists of a capacitor bank and an RC topology, the capacitor bank supports switching the capacitor size through a control circuit, the RC topology refers to the connection method of the resistor and capacitor, and the RC network can dynamically match the RC parameters according to the load characteristics to suppress surge current.

17. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the load power supply circuit control method as described in any one of claims 1 to 15 when executing the computer program.

18. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, it implements the steps of the load power supply circuit control method as described in any one of claims 1 to 15.

19. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the load power supply circuit control method as described in any one of claims 1 to 15.

Images