Intelligent power supply protection control method for weak current cabinet

Abstract

The present application belongs to the field of electric digital data processing, and discloses an intelligent power supply protection control method for a weak current cabinet, comprising: performing time length limited instantaneous power supply interruption and recovery operation on multiple ports in a determined order to carry out sequential probe scanning, the time length being less than the power-off holding time length of a load power supply; monitoring total current in parallel to identify abnormalities, and interrupting the sequential scanning and triggering focused probe scanning when abnormalities are identified; and before each interruption and recovery operation, dynamically determining the time length of the operation according to environmental parameters, the present application combines the controlled port switch action with the synchronous sampling of a single total current sensor, converts the past port level discrete information acquisition problem relying on multiple physical sensors into a timing data processing problem based on a single signal source, and constructs a real-time power consumption data set covering all ports without increasing hardware cost.

Description

Technical Field

[0001] This invention relates to an intelligent power supply protection and control method for low-voltage cabinets, belonging to the field of electrical digital data processing technology. Background Technology

[0002] In current data center and communication equipment room environments, rack-level power distribution units (PDUs) are widely used. Their core function is to provide a stable and reliable power supply for multiple servers, switches, and other low-voltage equipment within the rack. At the data processing level, monitoring the power supply status of the PDU relies on a single current sensor installed at the PDU's main input line. By collecting aggregated data such as total current and total power, the load status of the entire rack can be macroscopically assessed. This method has been widely deployed due to its simple hardware structure and low cost.

[0003] However, with the continuous increase in the deployment density of equipment inside the rack and the increasingly refined requirements for energy efficiency management, the limitations of this monitoring method, which can only provide macro-level aggregated data, have become increasingly prominent. It simplifies a complex system composed of dozens of independent dynamic loads into a single-variable static monitoring problem, forming a black box at the information acquisition level. Although maintenance personnel can know the total power consumption of the rack, they cannot actually perceive the precise power consumption, load status, and potential risks of each independent output port on the PDU. This lack of information granularity directly leads to the fact that in daily maintenance work, the capacity planning for new equipment can only rely on conservative reserve values ​​far exceeding actual needs. At the same time, there is a lack of effective technical means to identify and locate those devices that, although under low business load, maintain abnormal basic power consumption for some reason, i.e., so-called idle servers. In the long run, this results in a double waste of rack space and power resources.

[0004] To obtain power consumption data for each independent output port, the most direct technical approach in this field is to add an independent current or power metering module to each output port of the PDU. However, while this approach solves the problem of information granularity, it also means that the hardware cost, internal structural complexity, and potential failure points of the PDU will increase exponentially. This contradicts the overall development trend of modern data centers pursuing low cost, high reliability, and easy large-scale deployment. Specifically, existing technologies have the following shortcomings: 1. Obtaining independent power consumption data for each port requires a significant increase in hardware cost and system complexity; 2. Maintaining the existing low-cost single-sensor solution inevitably leads to inaccurate operation and maintenance decisions and energy efficiency management failures due to insufficient information granularity. Therefore, how to analyze the complete and accurate real-time power consumption map of each port through a data processing method without increasing the hardware complexity and cost of the PDU itself, and break the inherent contradiction between cost and information accuracy, becomes the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention provides an intelligent power supply protection and control method for low-voltage cabinets. Its main purpose is to solve the problem in the prior art that it is impossible to obtain accurate and real-time power consumption data of each independent port in the low-voltage cabinet without significantly increasing hardware costs.

[0006] To achieve the above objectives, this invention provides an intelligent power supply protection and control method for low-voltage cabinets, applicable to systems including a bus, a total current sensor, and multiple independently switchable power supply ports. The method performs the following operations within a unified control flow:

[0007] Step 1: Perform sequential probe scanning. In a determined order, perform time-limited instantaneous power interruption and recovery operations on multiple power supply ports in sequence. The duration is limited to less than the power outage retention time of the load power supply connected to the current port. During each time-limited instantaneous power interruption and recovery operation, the total current on the bus is synchronously collected to obtain the differential response current, and the instantaneous power of the current port is calculated based on the differential response current.

[0008] Step 2: Perform parallel anomaly monitoring, continuously monitor the total current on the bus to obtain the instantaneous total current value, compare the instantaneous total current value with the sum of the instantaneous power of multiple ports calculated based on sequential probe scanning, and when the instantaneous total current value exceeds the sum of the instantaneous power and meets the preset triggering conditions, interrupt the sequential probe scanning and trigger a focused probe scan for one or more candidate ports to obtain their transient peak power.

[0009] Step 3: Active adaptation during execution. Before each time-limited instantaneous power interruption and recovery operation, the ambient temperature and bus voltage are obtained, and the duration of the operation is dynamically determined based on the obtained ambient temperature and bus voltage.

[0010] Preferably, the method further includes: constructing and updating in real time a power map containing the power distribution status of all power supply ports based on the instantaneous power and transient peak power calculated in the sequential probe scan and focused probe scan; and when an overcurrent protection action of the system is detected, freezing the power map at the moment before the overcurrent protection action occurs, and locating the power supply port with abnormal power consumption based on the frozen power map.

[0011] Preferably, the preset triggering condition is that the magnitude and duration of the instantaneous total current value exceeding the sum of instantaneous power both meet their respective preset thresholds.

[0012] Preferably, before calculating the instantaneous power of the current port in the sequential probe scan, the method further includes: acquiring a historical total current time series before performing a time-limited instantaneous power interruption and recovery operation; establishing a prediction model for predicting the background current change trend based on the historical total current time series; and using the prediction output of the prediction model during the time-limited instantaneous power interruption and recovery operation as the baseline current for calculating the instantaneous power.

[0013] Preferably, during each time-limited transient power interruption and recovery operation, the method further includes: synchronously acquiring the transient change profile of the total current during the start and recovery phases of the time-limited transient power interruption and recovery operation at a time resolution higher than that when performing sequential probe scanning; extracting a health fingerprint based on the transient change profile to characterize the dynamic response characteristics of the load power input terminal connected to the current port; and generating a warning signal when the trend of the health fingerprint over time meets a preset drift condition.

[0014] Preferably, the method further includes: synchronously acquiring the transient change profile of the bus voltage during the recovery phase of a time-limited transient power interruption and recovery operation; and determining a connection health parameter to characterize the physical connection status of the current port, wherein the connection health parameter is contact resistance. Its calculation follows ,in This refers to the drop value that occurs synchronously with the transient change profile of the bus voltage during the recovery phase. This represents the peak value of the transient profile of the total current.

[0015] Preferably, the method further includes: identifying a power supply port whose power consumption does not match its corresponding service load status by analyzing the power map, and marking the identified power supply port as an abnormal state.

[0016] Preferably, focused probe scanning performs time-limited transient power interruption and recovery operations on one or more candidate ports at a higher execution frequency than sequential probe scanning.

[0017] Preferably, one or more candidate ports are ports whose instantaneous power values ​​calculated in the sequential probe scan are higher than a preset power threshold.

[0018] Preferably, the method further includes: monitoring the density of events where the instantaneous total current value exceeds the sum of instantaneous power; when the density exceeds a preset congestion threshold, pausing the focused probe scan and continuously recording the original current waveform on the bus until the density returns to below the preset congestion threshold; and after recording, performing a new round of active power probe operations on the active power supply ports and correlating the response results with the event characteristics in the recorded original current waveform to achieve attribution of multiple transient power anomaly events during congestion.

[0019] Compared with the prior art, the beneficial effects of the present invention are:

[0020] 1. By acquiring the baseline current and using it as a reference, the differential response current of the bus is acquired simultaneously with the instantaneous interruption and recovery operation of a single power supply port. By analyzing the correlation between the baseline and the differential current, a real-time power consumption data set covering all ports is constructed with only one physical acquisition point for the total current. This method transforms the port-level discrete information that previously required the superposition of multiple physical sensor hardware into a time-series data processing problem based on a single information source. It provides a data acquisition method that does not come at the cost of hardware costs for the refined management of the power supply status inside the weak current cabinet.

[0021] 2. While maintaining sequential probe scanning to establish a power map, the system monitors the instantaneous total current on the bus in parallel and compares it in real time with the predicted sum of power consumption of all ports in the power map. When the instantaneous total current exceeds the predicted sum and meets the preset conditions, the system interrupts the current sequential scan and triggers a focused probe scan. This data acquisition strategy, which combines background monitoring and event triggering, enables the system to capture and attribute short-term sudden power anomalies while maintaining low conventional resource overhead, thus solving the limitation of sequential sampling mechanism in terms of time resolution.

[0022] 3. During the active power probe operation, not only is the current amplitude of the differential response acquired, but also the transient change profile of the total current is simultaneously acquired with enhanced time resolution during the start and recovery phases of power outage. By extracting the characteristic parameters of this transient profile, a health fingerprint characterizing the dynamic response characteristics of the power input terminal of the low-voltage equipment is established. This allows a single power probe action to simultaneously carry out both power measurement and health assessment functions, providing an additional assessment of the dynamic response characteristics of the load power input terminal for diagnosis that delves into the equipment's internal physical characteristics from its operating status.

[0023] 4. The present invention further combines the acquisition of the transient current change profile with the synchronous acquisition of the transient bus voltage change profile. By analyzing the relationship between the peak current profile and the voltage profile drop value that occurs synchronously at the moment of power restoration, a health parameter characterizing the physical connection status of the power supply port is determined. This reuses the information acquisition process originally used for diagnosing load devices to assess the health of the power supply path itself, enabling the system to have the online monitoring capability for the potential risk of PDU internal connection degradation without adding any additional physical actions. Attached Figure Description

[0024] Figure 1 This is a dual-mode workflow diagram of the intelligent power supply protection and control method of the present invention;

[0025] Figure 2 This is a graph showing the dependence of the probe operation safety time of the present invention on environmental conditions;

[0026] Figure 3 This is a block diagram of the hardware composition and control firmware functional modules of the system of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in further detail below. 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.

[0028] This invention provides an intelligent power supply protection and control method for low-voltage cabinets, applicable to systems including buses, total current sensors, and multiple independently switchable power supply ports, such as data centers or communication equipment rooms with power distribution units (PDUs). A technical problem in such applications is how to acquire real-time power consumption data for each independent port without increasing hardware costs, in order to address inaccurate capacity planning and potential overload risks caused by dynamic load changes. The method of this invention allows a unified control flow to be executed by a microcontroller (MCU) deployed within the PDU. This flow mainly includes: performing sequential probe scanning to build a power data model, performing parallel anomaly monitoring to capture sudden power events, and performing proactive adaptive operations to adjust detection parameters.

[0029] In one specific implementation, the method begins with the execution of a sequential probe scan. This step aims to address the inability to acquire real-time power consumption data for each individual port. To achieve this, the system is configured to execute a data processing procedure based on active power probes and differential response analysis. This procedure utilizes the PDU's existing port on / off control capabilities for information detection. The specific implementation path is as follows: First, at the beginning of the scan cycle, the control flow acquires the baseline current. The controller, through its built-in analog-to-digital converter (ADC), acquires the signal output from the total current sensor at a sampling frequency of 1kHz and extracts a 100ms time window. The arithmetic mean of 100 sampling points within this window is calculated, and this average value is defined as the baseline current. The current is temporarily stored in the MCU's random access memory (RAM). Next, the control flow initiates a sequential scan cycle, performing a time-limited instantaneous power interruption and recovery operation on each port in a pre-defined order from port number 1 to N. This operation, as an active power probe operation, involves controlling the relay or MOSFET switch of the corresponding port to perform a disconnection and recovery action. The duration of the disconnection state is limited to a value shorter than the power-down retention time of the load power supply connected to the current port. In a server power supply application, the power-down retention time is typically greater than 100ms; therefore, this duration can be set to a value between 20ms and 50ms, such as 30ms. Third, during the 30ms period of the aforementioned instantaneous power interruption operation, the control flow synchronously acquires the total current on the bus and calculates the average current value within this 30ms window using the same 1kHz sampling frequency, denoted as the differential response current. ,in The current port number being probed; finally, the control flow executes the power resolution and attribution steps, determining the current port number. Instantaneous power consumption The calculation is based on its differential contribution to the total current, and the specific formula is as follows: Among them, the difference Port The differential response current, considering the actual effects of electrical noise and measurement drift, if the calculated differential response current is negative, then the current port... Instantaneous power consumption The value is zero in the formula. As an example of numerical deduction, consider the bus voltage measured by a voltage sampling circuit, for instance, 220V: If the measured... The voltage was 8.5A, measured when a probe operation was performed on port 3. It is 8.3A, bus voltage If the voltage is 220V, then the instantaneous power consumption of port 3 is... Calculated and attributed to The calculated instantaneous power consumption values ​​of each port are stored in a predefined data structure, which logically constructs a power map containing the power allocation status of all power supply ports.

[0030] In a low-voltage cabinet environment, electrical noise may affect measurement accuracy. To improve the fidelity of power map data, before performing power analysis and attribution steps, the system can execute a signal processing method based on sliding window trend prediction and synchronous residual verification. This is an alternative implementation method, and its procedure is as follows: Before performing power analysis on the port... Within 100ms prior to executing the instantaneous power interruption operation, the controller pre-collects 100 historical total current time series sample points. Subsequently, based on this historical total current time series, the system applies the least squares method to linearly fit these 100 data points, establishing a predictive model for predicting the baseline current change trend, in the form of: Where t is time, a is the slope of the linear model, used to characterize the trend of current change, and b is the intercept, used to characterize the initial baseline current; both are determined by calculation using the least squares method. Accordingly, during power outages, the baseline current used to calculate instantaneous power is no longer the previously measured baseline current. Instead, it uses the output value of the prediction model at the point in the interruption time window. The final power calculation formula was revised to By introducing a prediction model based on historical data trends to replace the static baseline, the interference of low-frequency drift and some noise on the measurement results can be suppressed. Furthermore, sequential probe scanning has limitations in temporal resolution and may miss transient power peaks commonly seen in computing devices. To compensate for this deficiency, the method of this invention executes an anomaly monitoring process in parallel while performing sequential scanning. This process continuously acquires the instantaneous total current value at a frequency of 5kHz and compares it with the predicted total current calculated from the sum of the instantaneous power of all ports in the power map. The triggering condition for a transient power anomaly event is defined as: when the instantaneous total current value exceeds the predicted total current in both magnitude and duration, and both conditions are met... If a preset threshold is reached, for example, if the amplitude exceeds 2A and the duration exceeds 3ms, an abnormal event is determined to have occurred. Once an event is triggered, the system immediately interrupts the ongoing sequential probe scan and triggers a focused probe scan. This focused probe scan performs a time-limited instantaneous power interruption and recovery operation on one or more candidate ports at a higher execution frequency than the regular sequential scan, for example, within 10ms. The candidate ports are determined based on ports whose instantaneous power values ​​calculated in the previous sequential scan are higher than a preset power threshold, for example, all ports with power exceeding 100W. This data acquisition method enables the system to capture and attribute short-term, sudden power anomaly events.

[0031] In situations where multiple power peaks occur concurrently, to ensure the integrity of information recording, the present invention also provides an alternative implementation method: monitoring the density of events where the instantaneous total current value exceeds the sum of instantaneous power. When this density exceeds a preset congestion threshold (e.g., 3 times within a unit of time, for example, 100ms), the system suspends all focused probe scans and instead uses the ADC buffer to continuously record the original high-frequency current waveform on the bus until the event density falls below the threshold. After recording, the system performs a new round of active power probe operations on all active power supply ports and correlates the response results with the event characteristics in the recorded original current waveform to attribute the causes of multiple transient power anomalies during congestion. Furthermore, instantaneous power interruption and... The safety of the recovery operation depends on the interrupt duration being less than the device's power-off retention time, which is affected by factors such as ambient temperature and bus voltage. To ensure the non-intrusiveness of the operation, the method of this invention performs an active adaptation step before each instantaneous power interruption and recovery operation. This step uses a temperature sensor and bus voltage sampling circuit integrated on the PDU controller to obtain the current ambient temperature and bus voltage in real time. A two-dimensional lookup table for probe safety duration is pre-defined in the controller's firmware. This table uses temperature and voltage as input axes and stores the maximum safe interrupt duration under different environmental combinations. Before performing the probe operation, the controller uses the real-time acquired ambient temperature and bus voltage as indexes to query this lookup table to dynamically determine the duration of this operation. For example, if the current (-5) bus voltage is found, the system will query the lookup table. Under the condition of 195V, the safety duration is 18ms, so the interrupt duration is set to 18ms. By changing this control parameter from a static constant to a real-time adaptive dynamic variable, the adaptability of the method in different working environments can be improved.

[0032] To further utilize the information acquired by probe operations, the method of this invention also includes a procedure for assessing the health status of the device. During each instantaneous power interruption and recovery operation, the system simultaneously acquires the transient profile of the total current during the start and recovery phases of the interruption with enhanced time resolution, such as instantaneously increasing the ADC sampling rate to 100kHz. Subsequently, based on this transient profile, the system executes a feature extraction algorithm that calculates morphological parameters such as the fall time from 90% to 10% and the overshoot during the recovery process. These parameters together constitute a health fingerprint characterizing the dynamic response characteristics of the load power input connected to the current port. When the trend of a port's health fingerprint over time meets a preset drift condition, such as its fall time accumulating a change exceeding 20% ​​of the initial baseline over three consecutive scan cycles, the system generates a warning signal. Without conflict, this information acquisition process can also simultaneously acquire the transient profile of the bus voltage during the power recovery phase and extract the peak value of the transient profile of the total current. Drop value occurring synchronously with voltage profile This is used to determine a connection health parameter that characterizes the physical connection status of the current port, namely contact resistance. Its calculation follows This enables online monitoring of the internal connection status of the PDU. The power map constructed and updated in real-time based on the above method can also be used for various power supply protection or control logics. For example, when an overcurrent protection action is detected, the power map at the moment before the overcurrent protection action occurs can be frozen, and the power supply port with abnormal power consumption can be located based on the frozen power map; or, by analyzing the power map, a power supply port whose power consumption does not match its corresponding service load status can be identified and marked as an abnormal state. As a numerical example, if the system measures the peak value of the transient change profile of the total current during the recovery phase of a probe operation... The current is 5.2A, and the instantaneous voltage drop of the bus is measured simultaneously. If the voltage is 0.11V, then the contact resistance of the current port is... identified as The contact resistance value is recorded as the initial baseline value of the health status when the device is first connected. If the value continues to rise in a unidirectional manner during subsequent operation, it indicates that the physical connection of the port or its internal relay contacts have deteriorated, and the system can issue an overheating risk warning accordingly.

[0033] Example 1: In a high-density data center rack providing backend computing power support for financial transactions, the intelligent power supply protection and control method for low-voltage cabinets of this invention is deployed. The rack houses 32 high-performance servers, each powered by a PDU with a rated current of 32A. Under normal operating conditions, the PDU's total current sensor displays a total current of 25A. At this time, the method continuously executes sequential probe scans in 2-second cycles. The resulting power map dataset also shows that the power consumption of all 32 ports is within the normal range. When the maintenance system distributes software update tasks to all servers in the rack in batches, one... The operating conditions presented a challenge: the update task would trigger a computational peak on a single server, lasting approximately 500ms and causing a sudden jump in power consumption from 200W to 750W. If the computational peaks of multiple servers overlapped in time, the combined peak current would exceed the PDU's capacity limit. Three seconds after the update task began, the server located on port 17 was the first to enter the computational peak, causing a sudden increase in the PDU's bus total current of approximately 2.5A. However, the sequential probe scan that was in progress at this time had not yet polled port 17, and the power map it maintained was still based on data from the previous cycle, failing to reflect this sudden change.

[0034] At this point, the parallel anomaly monitoring process in this method uses the power map established by the sequential probe scan as a dynamic benchmark. It continuously compares the collected instantaneous total bus current value with the sum of the instantaneous power of all ports in the power map at a frequency of 5kHz. When the process detects that the instantaneous total current value continuously exceeds the predicted sum of the power map by 2.5A for more than 5ms, this state meets the preset triggering conditions of an amplitude greater than 2A and a duration greater than 3ms. The system determines that a transient power anomaly event has occurred. Correspondingly, the system interrupts the currently executing sequential probe scan and triggers a focused probe scan. This focused probe scan, based on the list of ports with instantaneous power values ​​higher than 100W in the power map, identifies port 17 and other high-load servers as candidate ports, and sequentially executes the following steps on these candidate ports: The system performed a time-limited transient power interruption and recovery operation, which quickly confirmed that the differential response current of port 17 matched the abnormal increment of 2.5A. The system then updated the power map, recording the transient peak power of port 17 as 750W, and sent an early warning signal containing precise port location to the upper-level management system. Upon receiving the signal, the management system's scheduling strategy was triggered, adjusting the originally planned parallel update task to a batch-based, small-scale serial update, thereby avoiding the risk of multiple power peaks occurring concurrently. In this way, a local transient risk that could not be perceived due to relying on total average data was captured and located by a data processing method that combined background data modeling with high-speed anomaly detection, providing decision-making information to avoid a potential power failure without increasing physical measurement hardware.

[0035] After completing this early warning and handling, the system returned to normal operation. The power map maintained in its memory not only included the baseline power consumption of each port, but also additionally recorded the transient peak power event of port 17, providing a more accurate data model for subsequent load balancing and capacity planning. In another scenario, when the server on port 17 experiences a short circuit fault in its internal power module, causing a sharp increase in its instantaneous current and triggering the PDU's hardware overcurrent protection, the method of this invention automatically updates the power map data maintained in memory at the moment the circuit breaker trips. The system freezes the power supply to the PDU. After power is restored, maintenance personnel access the management interface to retrieve the frozen power map data. The data shows that during the last sampling period before the trip, the power values ​​of other ports were within the normal range, while the instantaneous power value of port 17 was an abnormal value far exceeding its rated power. Based on this, the maintenance system directly determines that port 17 is the source of the power outage and issues a repair order containing the precise port location to the maintenance personnel. This reduces the fault location time from several hours required for traditional manual troubleshooting to within seconds of data retrieval.

[0036] Example 2: To objectively verify the effectiveness of the method of the present invention in identifying different types of power anomaly events, this example constructs a reproducible hardware test platform. This platform is centered on a standard PDU with 16 independently controllable power supply ports. The PDU integrates a total current sensor with a range of 0-32A and an accuracy of ±1%, and is controlled by a microcontroller capable of executing the control flow of the present invention. The platform's load consists of 16 programmable DC electronic loads, each of which can have its operating current and duration set via host computer software to simulate the power consumption characteristics of a server under different operating conditions. Simultaneously, the test platform is equipped with a high-precision power analyzer with a sampling rate set to 10kHz and a measurement accuracy of ±0. 1% is used to record the total input of the PDU and the power of specific ports as a baseline for comparison. This experiment sets up two experimental groups: a control group and the sample group of this invention. The PDU used in the control group runs standard firmware that only has the function of total current monitoring. The PDU used in the sample group of this invention runs firmware that implements the complete technical solution of this invention. The complete cycle of sequential probe scanning is set to 2s, the duration of instantaneous power interruption is set to 30ms, and the trigger condition for parallel anomaly monitoring is set to the instantaneous total current exceeding the sum of the power map predictions by more than 1A and lasting for more than 5ms. The experiment includes two test items: Test item A is used to verify the ability to identify steady-state power anomalies, and Test item B is used to verify the ability to capture and attribute transient power peaks.

[0037] In test item A, 15 electronic loads were set to a normal standby power consumption of 15W, while the electronic load connected to port 8 was set to an abnormal standby power consumption of 60W to simulate an idle server scenario. After the test started, the control group could only observe a stable and slight increase in the total power of the PDU, but could not provide any information about the location of the abnormal source. In contrast, the sample of this invention constructed a complete power map in the first sequential probe scan cycle after startup. The map data showed that the instantaneous power of port 8 was 60.5W, while the power of the other ports was around 15W, thus locating the steady-state power anomaly within 2 seconds. Normal ports; In test item B, all 16 electronic loads were set to a normal standby power consumption of 15W. Then, the electronic load connected to port 5 was controlled by the host computer software to generate a current pulse with a power of 500W and a duration of 200ms at a random time point during the test to simulate ghost peaks. To further demonstrate the synergistic effect between technical features, a partially missing control group was added to this test item. The firmware of this group of PDUs only included the sequential probe scanning module and did not include the parallel anomaly monitoring module. After the test was repeated 100 times, the statistical results are shown in Table 1. Table 1: Comparison of the identification results of the two types of test events for different test groups.

[0038] Test sample group Test Events Event recognition results Response time control group Steady-state power anomaly Unidentified port N / A Sample of the present invention Steady-state power anomaly Identify port 8 (60.5W) <2s control group Transient power peak Unidentified port N / A Partially missing control group Transient power peak 9% success rate in recognition >2s Sample of the present invention Transient power peak 100% success rate in recognition <50ms

[0039] Referring to Table 1, the experimental data shows that the partial missing control group has a low probability of its probe operation sampling window coinciding with the event occurrence window due to its 2s scanning cycle being much longer than the 200ms event duration, resulting in an identification success rate of only 9%. In contrast, in the sample group of this invention, the parallel anomaly monitoring process uses the power map established by sequential probe scanning as a benchmark. At each peak event, it can detect the deviation between the instantaneous total current and the predicted sum and trigger a focused probe scan, thus achieving a 100% identification success rate and a response time of less than 50ms. The results of this experiment show that the method of this invention can effectively construct the power map of each port and locate steady-state power anomalies using a single total current sensor. At the same time, through the mechanism of parallel monitoring and event triggering, it can capture and attribute transient power peaks that are easily missed by conventional sampling methods, verifying the engineering feasibility and practical effect of the solution.

[0040] Example 3: This example combines Figures 1 to 3 This document describes an intelligent power supply protection and control method for a low-voltage electrical cabinet, such as... Figure 1As shown, the process begins with the total bus current as the single physical source, entering the system initialization and environment self-learning phase to calibrate the noise baseline and set monitoring thresholds. Subsequently, the system enters a core loop to execute sequential probe scans to build and update the power maps of each port. Simultaneously, a parallel anomaly monitoring module continuously compares the real-time total current with the predicted sum of the power map. If the total current does not exceed the predicted sum, it returns to continue executing sequential probe scans. If the total current exceeds the predicted sum, it triggers a focused probe scan to probe candidate ports at a higher scanning speed, thereby attributing transient anomalies. In this process, an active adaptive control module dynamically adjusts the probe operation duration based on real-time environmental parameters, namely temperature and bus voltage, to ensure operational safety. At the same time, a device health status assessment module extracts health fingerprints and connection status parameters during probe operations. Finally, the system integrates all information to generate a real-time power map and risk warning, outputting key information such as port power consumption, health status, and fault location.

[0041] like Figure 2 As shown, the horizontal axis of the graph represents ambient temperature, in units of... The vertical axis represents the safe interruption duration in milliseconds (ms). The graph uses three curves to illustrate the dependence of safe interruption duration on ambient temperature under three different operating conditions: bus voltages of 190V, 210V, and 230V. The safe interruption duration is shorter in the low-temperature region due to capacitor performance degradation, and shorter in the normal temperature region (approximately 25°C). The safe time is longest near the point where the capacitor has the best performance; it is shortened again in the high-temperature area due to the significant increase in leakage current; the three voltage values ​​of 190V, 210V and 230V in the figure are intended to simulate typical undervoltage, normal and overvoltage conditions under a nominal 220V power supply to verify the environmental adaptability of the invention. Under the same temperature, the lower the voltage, the shorter the safe time, because the capacitor needs to compensate for the low voltage with a larger current, thereby accelerating its own discharge.

[0042] like Figure 3 As shown, the system is centered around an intelligent power distribution unit (PDU), which integrates an embedded microcontroller (MCU). This MCU runs an intelligent power protection and control firmware, which includes a sequential probe scanning module, a parallel anomaly monitoring module, and a health fingerprint assessment module. The MCU obtains bus current information through a total current sensor and monitors temperature and voltage in real time through environmental sensors. It internally stores a probe safety duration lookup table as a decision-making basis, thereby controlling a multi-port independent switch matrix to supply power to backend load devices such as server A, server B, and network switches. At the same time, the system can report the generated warning signals and status data to a remote operation and maintenance management platform.

[0043] Example 4: This example is a specific engineering implementation of a health fingerprint extraction and assessment procedure. It aims to identify early signs of progressive aging in the internal energy storage capacitors of a server power module already in operation, caused by long-term operation. This type of aging is a potential risk, characterized by increased instantaneous power consumption of the device. While maintaining within the normal range, the dynamic response characteristics at its power input have deteriorated. In a specific application scenario, when a new server device is first connected to a power supply port of a PDU, the method of this invention is configured to perform a baseline calibration process for a health fingerprint. This process targets the power module of the new server, which is initially healthy and not aged. The enabling environment is a PDU with the method of this invention deployed, whose microcontroller has the capability to perform instantaneous analog-to-digital conversion sampling at a frequency of not less than 100kHz. After the process is started, the system performs a time-limited instantaneous power interruption and recovery operation on the port. During the power recovery phase, the controller continuously collects 512 data points at a sampling rate of 100kHz. The number of data points is set to an integer power of 2 to facilitate subsequent frequency domain analysis such as Fast Fourier Transform (FFT) or efficient memory processing, forming a high-resolution current transient change profile. The system then performs a feature extraction algorithm on the digital samples to generate a health fingerprint characterizing the device's dynamic response characteristics. The algorithm's path is as follows: First, it searches for the maximum value among the 512 sample points to determine the peak value of the total current transient change profile. The mean of the last 25% of the sample sequence was calculated as the steady-state current. The second step is to calculate the rise time. That is, the current flows from Rise to The time elapsed; the third step, calculating the oversurge. Its calculation formula is Fourth step, calculate the settlement time. That is, the peak value of the transient profile of the total current. Initially, current fluctuations enter and remain. The time required within the ±2% error band; these three parameters Together, they form a three-dimensional feature vector, which is defined as the initial health fingerprint of the device and associated with the port number. It is stored in the non-volatile memory of the PDU controller. This fingerprint vector is a quantitative representation of the health status of the equivalent capacitance at the device's power input terminal.

[0044] During long-term operation of the equipment, the system repeats the transient profile capture and fingerprint extraction process described above every time a sequential probe scan is performed, generating a current health fingerprint. The system compares the current fingerprint vector with the stored baseline fingerprint vector to determine the trend of changes in the equipment's health status. The drift condition threshold used for judgment is predetermined based on an offline statistical calibration procedure. This procedure establishes a characteristic parameter distribution model for healthy and faulty samples by performing fingerprint tests on a large number of new and artificially aged power modules of the same model, and sets a discrimination boundary that can distinguish between the two types of samples. When any characteristic parameter deviates from its baseline value by more than 3 standard deviations, it is judged as abnormal. As a numerical example: the baseline fingerprint of a server is 1.2ms, 5.1%, 3.5ms. After 6 months of operation, the current fingerprint obtained by a certain measurement is 0.8ms, 15.4%, 4.8ms. The system calculates its rise time. With overshoot The changes all exceeded the preset statistical thresholds; based on this, the system determined that the health fingerprint of the device had drifted significantly and generated a predictive maintenance alarm indicating that the device was in a sub-healthy state and had a potential electrical fault risk, which was then reported to the operation and maintenance management platform. Based on this, the operation and maintenance personnel could replace the power module of the server in advance, thereby avoiding unexpected equipment downtime or electrical short circuit faults that might be caused by the eventual failure of the capacitor. In this way, a process of data mining based on the additional information of existing detection actions extends the monitoring dimension from the power status of the external device to the physical health status of its internal components, providing a data basis for preventive maintenance.

[0045] Example 5: This example is a specific engineering implementation of the active adaptation step, aiming to provide a systematic and reproducible data filling and calibration procedure for the probe safety duration two-dimensional lookup table on which it relies. In a controlled engineering environment, this procedure experimentally obtains the maximum safe interruption duration that a specific load device can withstand under different combinations of environmental parameters, thereby transforming the construction of the lookup table from relying on empirical settings to a calibration process based on measured data. This calibration procedure is executed on a test platform containing an adjustable temperature chamber and a programmable AC power supply. The device under test, i.e., a specific model of server, is placed in the temperature chamber and powered by a PDU integrating the method of this invention. The input of the PDU is connected to the programmable AC power supply. At the beginning of the procedure, an environmental parameter combination is set, and the temperature of the temperature chamber is set to -10°C. The AC power output voltage is set to 190V. Subsequently, the PDU's control system executes an incremental interruption sequence on the port where the server is located. The sequence starts with an interruption duration of 1ms and increases by 1ms each time. After each time-limited instantaneous power interruption and recovery operation, the system continuously monitors the connectivity of the server's external network port to determine if a restart or operational interruption has occurred. When an abnormal server operating status is detected for the first time, the duration of the previous interruption that did not cause an abnormality is recorded as the critical safe duration under these environmental parameters. This critical safe duration is multiplied by a safety factor, such as 0.8, and the result is used as the final maximum safe interruption duration, which is then filled into the corresponding (-10) in a two-dimensional lookup table. Within the cell containing 190V; repeat this incremental probe sequence until all preset temperature and voltage combinations are traversed, thus generating a complete two-dimensional lookup table for this type of server.

[0046] During field deployment, when the device models connected to the PDU are unknown or diverse, a pre-emptive online safety boundary self-calibration process can be executed. In this process, before entering routine monitoring, the PDU performs the aforementioned interrupt duration incrementing detection sequence for each active power supply port under the current ambient temperature and bus voltage. This determines a critical safety duration for each specific device connected to the port under the current operating conditions. This duration is stored as a personalized safety parameter bound to that port. In subsequent routine operation, when the system performs proactive adaptation steps, it will prioritize using this more targeted safety duration parameter obtained from online calibration, thereby improving the safety of active power probe operation in heterogeneous device environments.

[0047] Example 6: This example is a standardized pre-deployment calibration and online operation parameter self-tuning procedure for a system. It aims to determine the key thresholds of the parallel anomaly monitoring module based on the unique electrical noise environment and load characteristics of the site, and to manage low signal-to-noise ratio ports. In a specific deployment scenario, when a PDU integrating the method of this invention completes hardware installation and is powered on for the first time in a new low-voltage cabinet, before entering the normal operating mode, the system is configured to first execute a 10-minute environmental self-learning process. During this process, all load devices in the cabinet are in a stable standby state. The PDU's control system continuously acquires the total current signal on the bus at a frequency of 10kHz, forming a time-series database containing 6 million sample points. The system then performs statistical analysis on this database, calculates the standard deviation of the sequence, and defines it as the background noise amplitude characteristic of the current deployment environment. Meanwhile, by analyzing the autocorrelation function of the noise signal, the average duration of the noise pulse is determined. .

[0048] Based on the above data collection and analysis, the system performs parameter self-tuning, and the amplitude threshold used for parallel anomaly monitoring is set as the background noise amplitude characteristic. A predetermined multiple, typically between 3 and 6; in this embodiment, 6 times is chosen, i.e. This is to control the probability of false triggering caused by random noise below a certain level; correspondingly, the duration threshold is set to... A multiple of, typically between 2 and 5, is chosen in this embodiment as 2 times, i.e. This distinguishes between genuine load transients and brief noise interference; simultaneously, during this self-learning process, the system performs a complete sequential probe scan for each port. If its calculated differential response current The amplitude is less than If the signal-to-noise ratio (SNR) is low, the port is marked as a low SNR port. In subsequent normal operation, for ports marked as low SNR, their power readings are marked in the power map as a specific state bit indicating that the power is below the detectable threshold, thus avoiding inaccurate power map data caused by the measurement signal being submerged by noise. After completing the above process, the system stores the self-tuned parameters in non-volatile memory and automatically switches to normal operating mode. In this way, a standardized deployment pre-procedure changes the key control parameters and judgment logic of the method of this invention from relying on general preset values ​​to being bound to the physical characteristics of specific application scenarios, thereby improving the adaptability and data reliability of the method in different electrical environments.

[0049] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent power supply protection and control of a low-voltage cabinet, applied to a system including a bus, a total current sensor, and multiple independently switchable power supply ports, characterized in that, The method operates within a unified control flow, performing the following actions: Step 1: Perform sequential probe scanning. In a determined order, perform time-limited instantaneous power interruption and recovery operations on multiple power supply ports in sequence. The duration is limited to less than the power outage retention time of the load power supply connected to the current port. During each time-limited instantaneous power interruption and recovery operation, the total current on the bus is synchronously collected to obtain the differential response current, and the instantaneous power of the current port is calculated based on the differential response current. Step 2: Perform parallel anomaly monitoring, continuously monitor the total current on the bus to obtain the instantaneous total current value, compare the instantaneous total current value with the predicted total current calculated by the sum of the instantaneous power of multiple ports obtained based on sequential probe scanning, and when the instantaneous total current value exceeds the predicted total current and meets the preset triggering conditions, interrupt the sequential probe scanning and trigger a focused probe scan for one or more candidate ports to obtain their transient peak power. Step 3: Actively adapt during execution. Before each time-limited instantaneous power interruption and recovery operation, acquire the ambient temperature and bus voltage, and dynamically determine the duration of the operation based on the acquired ambient temperature and bus voltage. Focused probe scanning, in particular, performs time-limited transient power interruption and recovery operations on one or more candidate ports at a higher execution frequency than sequential probe scanning.

2. The intelligent power supply protection and control method for a low-voltage cabinet according to claim 1, characterized in that, The method also includes: constructing and updating in real time a power map containing the power distribution status of all power supply ports based on the instantaneous power and transient peak power calculated in sequential probe scanning and focused probe scanning; and when an overcurrent protection action of the system is detected, freezing the power map at the moment before the overcurrent protection action occurs, and locating the power supply port with abnormal power consumption based on the frozen power map.

3. The intelligent power supply protection and control method for a low-voltage cabinet according to claim 1, characterized in that, The preset triggering conditions are: the magnitude and duration of the instantaneous total current value exceeding the sum of instantaneous power both meet their respective preset thresholds.

4. The intelligent power supply protection and control method for a low-voltage cabinet according to claim 1, characterized in that, In sequential probe scanning, before calculating the instantaneous power of the current port, the method further includes: acquiring a historical total current time series prior to performing a time-limited instantaneous power interruption and recovery operation; establishing a prediction model for predicting the background current change trend based on the historical total current time series; and using the prediction output of the prediction model during the time-limited instantaneous power interruption and recovery operation as the baseline current for calculating the instantaneous power.

5. The intelligent power supply protection and control method for a low-voltage cabinet according to claim 1, characterized in that, During each time-limited transient power interruption and recovery operation, the method further includes: synchronously acquiring the transient change profile of the total current during the start and recovery phases of the time-limited transient power interruption and recovery operation at a time resolution higher than that when performing sequential probe scanning; extracting a health fingerprint based on the transient change profile to characterize the dynamic response characteristics of the load power input terminal connected to the current port; and generating a warning signal when the trend of the health fingerprint over time meets a preset drift condition.

6. The intelligent power supply protection and control method for a low-voltage cabinet according to claim 5, characterized in that, The method also includes: synchronously acquiring the transient change profile of the bus voltage during the recovery phase of a time-limited transient power interruption and recovery operation; and determining a connection health parameter to characterize the physical connection status of the current port, wherein the connection health parameter is contact resistance. Its calculation follows ,in, This refers to the drop value that occurs synchronously with the transient change profile of the bus voltage during the recovery phase. This represents the peak value of the transient profile of the total current.

7. The intelligent power supply protection and control method for a low-voltage cabinet according to claim 2, characterized in that, The method also includes: identifying a power supply port whose power consumption does not match its corresponding service load status by analyzing the power map, and marking the identified power supply port as abnormal.

8. The intelligent power supply protection and control method for a low-voltage cabinet according to claim 1, characterized in that, One or more candidate ports are ports whose instantaneous power values ​​calculated during sequential probe scanning are higher than a preset power threshold.

9. The intelligent power supply protection and control method for a low-voltage cabinet according to claim 2, characterized in that, The method also includes: monitoring the density of events where the instantaneous total current value exceeds the sum of instantaneous power; when the density exceeds a preset congestion threshold, pausing the focused probe scan and continuously recording the original current waveform on the bus until the density returns to below the preset congestion threshold; and after recording, performing a new round of active power probe operations on the active power supply ports and correlating the response results with the event characteristics in the recorded original current waveform.

Images