A low-voltage intelligent power utilization control method and system based on the Internet of Things

Abstract

The embodiment of the application provides a low-voltage intelligent power utilization control method and system based on Internet of Things, relates to the field of low-voltage intelligent power utilization control, and comprises the following steps: identifying the power supply loop of a server cluster, monitoring the electric parameters of the power supply loop, triggering a potential overload event flag when the current exceeds a preset current and the duration exceeds a preset time, obtaining the predicted occurrence period of a computing task of the server cluster and the computing task execution state in response to the potential overload event flag, wherein the predicted occurrence period is a time period estimated based on historical tasks, and controlling the power supply or power-off of the power supply loop according to the electric parameters, the predicted occurrence period and the computing task execution state. The application can avoid unnecessary power-off caused by misjudgment, thereby guaranteeing the continuity of key businesses, effectively solving the limitations of the "one-size-fits-all" protection mechanism of the existing intelligent power utilization system, and improving the intelligent level and reliability of low-voltage power management.

Description

Technical Field

[0001] This application relates to the field of low-voltage smart power control, and more specifically, to a low-voltage smart power control method and system based on the Internet of Things. Background Technology

[0002] Low-voltage power management in industrial and civil buildings faces numerous challenges, such as fragmented power consumption data, insufficient real-time data, and untimely fault warnings. These problems can lead to energy waste, equipment damage, and even safety hazards. To address these challenges, IoT-based smart power control systems have been introduced, aiming to improve power monitoring and management through intelligent means.

[0003] However, in practical applications, especially in specific scenarios where power supply continuity requirements are extremely high and high-performance server clusters experience instantaneous high-load surges during normal business operations, existing IoT-based low-voltage smart power control systems have limitations. For example, in environments such as data centers, when high-performance servers perform intensive computing tasks, the current in their power supply lines will experience a significant peak within a short period. Although this peak current is short-lived and its value may be within the design capacity of the power supply line, it may exceed the static current alarm limit preset by the smart power system.

[0004] Existing smart power systems typically employ straightforward control logic. Once the current exceeds a preset limit and persists for a period, the system identifies an "overload" risk and triggers automated protection to cut off power. While this "one-size-fits-all" protection mechanism is effective in ensuring electrical safety, it fails to distinguish between dangerous, sustained overloads caused by short circuits or damaged equipment insulation, and the functional, temporary load spikes required by high-performance server clusters during normal operation. The system misinterprets the latter as the former, leading to unnecessary power outages and severely disrupting the continuity of critical business operations—a far cry from the original purpose of deploying a smart system. Summary of the Invention

[0005] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application discloses a low-voltage smart power control method and system based on the Internet of Things, which aims to solve the problem that existing low-voltage smart power control systems cannot distinguish between functional and temporary load peaks required for normal business and dangerous continuous overloads when dealing with instantaneous high load impacts on high-performance server clusters, resulting in unnecessary power outage protection and seriously interfering with the continuity of critical business.

[0006] In a first aspect, this application discloses a low-voltage smart power control method based on the Internet of Things, comprising the following steps: Identify the power supply circuit of the server cluster; Monitor the electrical parameters of the power supply circuit, including current, voltage, and power; When the current exceeds the preset current and the duration exceeds the preset time, the potential overload event flag is triggered; In response to a potential overload event flag, obtain the expected occurrence time of the computing tasks in the server cluster and the execution status of the computing tasks, where the expected occurrence time is a time period estimated based on historical tasks; Based on electrical parameters, the expected time period, and the execution status of the calculation task, control the power supply circuit to supply power or cut off power.

[0007] Furthermore, based on the above method, the steps for controlling the power supply circuit to supply power or cut off power according to electrical parameters, the expected occurrence time period, and the execution status of the calculation task include: The electrical parameters are matched with the pre-stored computation task electrical parameters to obtain the electrical parameter matching degree between the electrical parameters and the pre-stored computation task electrical parameters. The pre-stored computation task electrical parameters are the electrical parameters of the power supply circuit when the server cluster executes the computation task. When the electrical parameter matching degree is greater than the preset threshold, obtain the trigger time of the potential overload event flag; When the trigger time of the potential overload event flag does not fall within the expected occurrence period, the control power supply circuit is de-energized; When the trigger time of a potential overload event falls within the expected occurrence period, the probability of the server cluster executing computing tasks is determined based on the computing task execution status. When the probability of the server cluster executing a computing task is 1, the power supply circuit is powered.

[0008] In some preferred embodiments, the steps of controlling the power supply circuit to supply power or cut off power based on electrical parameters, the expected time period, and the execution status of the computing task include: The electrical parameters are matched with the pre-stored computation task electrical parameters to obtain the electrical parameter matching degree between the electrical parameters and the pre-stored computation task electrical parameters. The pre-stored computation task electrical parameters are the electrical parameters of the power supply circuit when the server cluster executes the computation task. Obtain the trigger time of the potential overload event flag; Determine the time matching degree based on the trigger time and the expected occurrence period; Based on the execution status of the computing tasks, determine the probability of the server cluster executing computing tasks; The benign load confidence index is calculated based on the electrical parameter matching degree, time matching degree, probability of executing the calculation task, preset electrical parameter matching degree weight, preset time matching degree weight, and preset probability weight of executing the calculation task. The sum of the preset electrical parameter matching degree weight, preset time matching degree weight, and preset probability weight of executing the calculation task is 1. The power supply circuit is controlled to supply power or cut off based on the benign load confidence index.

[0009] Furthermore, the steps of matching the electrical parameters with the pre-stored computational task electrical parameters to obtain the electrical parameter matching degree between the electrical parameters and the pre-stored computational task electrical parameters include: Obtain the current, voltage, and power of the power supply circuit; The current is compared with the pre-stored calculated task current to obtain the current similarity. The voltage is compared with the pre-stored calculated task voltage to obtain the voltage similarity. The power is compared with the power of the pre-stored calculated task to obtain the power similarity; The electrical parameters are calculated based on current similarity, voltage similarity, and power similarity to determine the electrical parameter matching degree between the electrical parameters and the pre-stored electrical parameters of the calculation task.

[0010] As a technological improvement, the steps to compare the current with the pre-stored calculated task current to obtain the current similarity include: Obtain the current curve of the power supply circuit within a preset sliding time window; Extract the current rise rate and peak current from the current curve; The current similarity is obtained by comparing the current rise rate with the current rise rate of the pre-stored calculated task, and by comparing the current peak value with the current peak value of the pre-stored calculated task.

[0011] Based on this, the steps to obtain the current similarity by comparing the current rise rate with the pre-stored calculated task current rise rate, and by comparing the current peak value with the pre-stored calculated task current peak value, include: Compare the current rise rate with the pre-stored calculated task current rise rate, and calculate the current rise rate deviation. Current rise rate deviation = (current rise rate - pre-stored calculated task current rise rate) / pre-stored calculated task current rise rate; Compare the peak current with the pre-stored calculated task current peak, and calculate the peak current deviation. Peak current deviation = (peak current - pre-stored calculated task current peak) / pre-stored calculated task current peak. The geometric mean is used to calculate the current similarity based on the deviation in current rise rate and the deviation in current peak value.

[0012] To improve the solution, the steps to determine the time matching degree based on the trigger time and the expected occurrence period include one of the following: When the time window of the expected occurrence period is less than the threshold of the first time period, and the trigger time falls within the expected occurrence period, the time matching degree is recorded as 1. When the time window of the expected occurrence period is greater than the first time period threshold and less than the second time period threshold, and the trigger time falls within the expected occurrence period, the time matching degree is recorded as 0.5, where the first time period threshold is less than the second time period threshold. If the time window of the expected occurrence period is greater than the threshold of the second time period or the trigger time does not fall within the expected occurrence period, the time matching degree is recorded as 0.

[0013] To optimize the architecture, the steps for determining the probability of a server cluster executing computational tasks based on the task execution status include: When the execution status of the computation task is "in execution", the probability of executing the computation task is determined to be 1. When the computation task status is "not executed", the probability of executing the computation task is determined to be 0. When the status of the computation task is ambiguous, the probability of executing the computation task is determined to be 0.5.

[0014] To enhance functionality, the steps for controlling the power supply circuit to supply or disconnect power based on a benign load confidence index include one of the following: When the confidence index of a benign load is greater than the first preset index, the power supply circuit is controlled to supply power. When the confidence index of a benign load is less than the second preset index, the power supply circuit is cut off, wherein the first preset index is greater than the second preset index. When the confidence index of benign load is less than or equal to the first preset index and greater than or equal to the second preset index, a manual review is requested, and the power supply circuit is controlled to supply power or cut off power based on the result of the manual review.

[0015] Secondly, this application also discloses a low-voltage smart power control system based on the Internet of Things, the system comprising: The identification module is used to identify the power supply circuit of the server cluster; The parameter monitoring module is used to monitor the electrical parameters of the power supply circuit, including current, voltage, and power. The potential overload event triggering module is used to trigger a potential overload event flag when the current exceeds a preset current and the duration exceeds a preset time. The data acquisition module is used to obtain the expected occurrence time and execution status of the computing tasks of the server cluster in response to potential overload event flags, wherein the expected occurrence time is based on the time period estimated from historical tasks; The control module is used to control the power supply circuit to supply power or cut off power based on electrical parameters, the expected time period, and the execution status of the calculation task.

[0016] This application discloses an IoT-based low-voltage smart power control method that identifies the power supply circuits of a server cluster and monitors their current, voltage, and power parameters in real time, enabling timely detection of potential overload events. Unlike existing technologies that simply rely on current thresholds, this application, after triggering a potential overload event flag, further obtains the expected occurrence time and execution status of the server cluster's computing tasks. This judgment mechanism, combined with business scenario information, allows the system to distinguish between dangerous, continuous overloads caused by line faults and functional, temporary load peaks required by the business during normal operation of the high-performance server cluster. By comprehensively considering electrical parameters, expected occurrence time, and computing task execution status, the system can make more accurate power supply or power outage decisions, avoiding unnecessary power outages due to misjudgments in existing technologies. This ensures the continuity of critical business operations, effectively overcomes the limitations of the "one-size-fits-all" protection mechanism in existing smart power systems, and improves the intelligence and reliability of low-voltage power management.

[0017] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0018] The accompanying drawings are used to provide a further understanding of the technical solutions of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions of this application.

[0019] Figure 1 This is a flowchart illustrating an embodiment of the IoT-based low-voltage smart power control method provided in this application. Detailed Implementation

[0020] To make the objectives, technical methods, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0021] It should be noted that the meaning of "multiple" (or "more than") in the description of the embodiments of this application refers to two or more, and "greater than," "less than," "exceeding," etc. are understood to exclude the number itself, while "above," "below," "within," etc. are understood to include the number itself. If "first," "second," etc. are used in the description, they are only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated or the order of the technical features indicated.

[0022] In the description of this application, unless otherwise expressly defined, terms such as "setup," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this application in conjunction with the specific content of the technical solution.

[0023] Based on the above, this application proposes a low-voltage smart power control method and system based on the Internet of Things, aiming to ensure the efficiency and security of data center operation and maintenance.

[0024] See Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of an IoT-based low-voltage smart power control method provided in this application. The IoT-based low-voltage smart power control method provided in this application includes, but is not limited to, steps S110 to S150, which will be described in detail below.

[0025] S110, Identify the power supply circuit of the server cluster; S120. Monitor the electrical parameters of the power supply circuit, including current, voltage and power; S130. When the current exceeds the preset current and the duration exceeds the preset time, a potential overload event flag is triggered. S140. In response to a potential overload event flag, obtain the expected occurrence time of the computing tasks in the server cluster and the execution status of the computing tasks, wherein the expected occurrence time is a time period estimated based on historical tasks. S150: Control the power supply circuit to supply power or cut off power based on electrical parameters, the expected time period, and the execution status of the calculation task.

[0026] A "server cluster" refers to a collection of computing resources consisting of multiple servers that work together to provide high-performance computing services, such as in data centers and cloud computing platforms. A "power supply circuit" refers to the electrical path that provides power to the server cluster, typically including components such as circuit breakers, cables, and sockets. "Electrical parameters" are physical quantities describing the electrical state of the power supply circuit, mainly including current, voltage, and power; real-time monitoring of these parameters is fundamental to determining the power consumption status. "Preset current" and "preset time" are thresholds preset by the system for preliminary assessment of potential overload conditions. A "potential overload event flag" is a signal or state indicating that the system has detected an abnormal current, but it has not yet been definitively determined whether a power outage is necessary. "Expected occurrence period" refers to the anticipated time window for the server cluster to execute a specific computing task; this period is estimated based on historical data and task scheduling information. "Computation task execution status" refers to whether the server cluster is currently executing a computing task and the progress of the task. The implementation environment of this application is typically an IoT-based smart power control system, which uses sensors, communication networks, and a data processing platform to achieve real-time monitoring and intelligent control of the low-voltage power environment.

[0027] The first step is to identify the power supply circuits of the server cluster. This identification can be achieved in several ways. For example, it can be done manually by having maintenance personnel manually input the power supply circuit numbers or identifiers connected to the server cluster during system initialization. Another method is to automatically identify and register the power supply circuits associated with the server cluster by scanning smart meters or sensors on the power supply circuits.

[0028] After identifying the power supply circuit, it is necessary to monitor its electrical parameters. This monitoring can be achieved by installing smart sensors or smart meters on the power supply circuit. These devices can collect current, voltage, and power data of the power supply circuit in real time and upload the data to the control system via an IoT communication module. For example, a split-type current transformer and voltage sensor can be used, installed on the incoming line of the power supply circuit, to collect current and voltage signals in real time, and then calculate the power value using a built-in computing module.

[0029] When the monitored current exceeds a preset current and lasts for a preset duration, the system triggers a potential overload event flag. The preset current and preset time are pre-set by the system administrator based on factors such as the server cluster's rated power, line capacity, and service characteristics. For example, the flag can be set to trigger when the current exceeds 120% of the rated current for more than 5 seconds. This process can be implemented by setting up a real-time monitoring and comparison module in the control system. This module continuously receives electrical parameter data and compares it with preset thresholds.

[0030] In response to potential overload event flags, the system needs to obtain the estimated occurrence time and execution status of computing tasks on the server cluster. The estimated occurrence time is based on historical task data. For example, by analyzing the task scheduling logs of the server cluster over a past period, information such as the average execution time and peak periods for different types of computing tasks can be obtained, thus predicting computing tasks that may occur in a future time period. The execution status of computing tasks can be obtained through an interface with the server cluster's management system. For example, the server cluster's management system can provide an API interface for the smart power control system to query information such as the list of currently running tasks and task progress.

[0031] Finally, based on electrical parameters, the expected time period of occurrence, and the execution status of computing tasks, the system controls the power supply circuit to either supply power or cut off power. For example, when the system detects a potential overload event flag being triggered, if the server cluster is currently in the expected peak computing task period and the computing task execution status shows that high-intensity tasks are running, the system may determine that this is a benign load fluctuation and choose to continue supplying power. Conversely, if the potential overload event flag is triggered, but the server cluster is not in the expected computing task period, or the computing task execution status shows that no high-intensity tasks are running, the system may determine that this is a real overload risk and choose to cut off power to protect the equipment and lines.

[0032] This application significantly improves the intelligence level of power control by incorporating considerations of the estimated occurrence time and execution status of server cluster computing tasks. Specifically, when the system detects that the current in the power supply circuit exceeds a preset value and persists for a period of time, triggering a potential overload event flag, it does not immediately take power-off measures. Instead, the system further obtains the estimated occurrence time of the server cluster's computing tasks and the current execution status of the computing tasks. The estimated occurrence time is estimated based on historical task data, reflecting the time window when the server cluster may experience high loads under normal business conditions. The computing task execution status provides real-time information on whether the server cluster is currently performing high-intensity computing tasks.

[0033] By comprehensively analyzing electrical parameters, the expected time period of occurrence, and the execution status of computing tasks, this application can more accurately determine the nature of potential overload events. For example, if a potential overload event occurs during a period when the server cluster is expected to perform high-intensity computing tasks, and the server cluster is indeed performing computing tasks, the system can determine that this is likely a benign, functional load peak required by the business, rather than a dangerous electrical fault. In this case, the system will choose to continue supplying power to ensure the continuity of critical business operations. Conversely, if a potential overload event occurs outside of the expected task period, or if the server cluster is not performing high-intensity computing tasks, the system will determine that this may be a genuine overload risk and will take power-off measures to protect equipment and line safety.

[0034] This multi-dimensional and intelligent judgment mechanism enables this application to effectively avoid unnecessary power outages caused by misjudgments in traditional systems, greatly improving the reliability of power supply to server clusters and the continuity of business operations. Compared with existing technologies, the core innovation of this application lies in its introduction of a deep understanding of load behavior patterns, that is, intelligently interpreting abnormal fluctuations in electrical parameters by combining the business characteristics of the server cluster (the expected occurrence time and execution status of computing tasks). This method breaks through the limitations of traditional systems that rely solely on static electrical parameter thresholds for judgment, realizing a transformation from "passive response" to "intelligent prediction".

[0035] In some embodiments, the step of controlling the power supply circuit to supply power or cut off power based on the electrical parameters, the expected occurrence time period, and the execution status of the computing task includes: The electrical parameters are matched with the pre-stored computation task electrical parameters to obtain the electrical parameter matching degree between the electrical parameters and the pre-stored computation task electrical parameters. The pre-stored computation task electrical parameters are the electrical parameters of the power supply circuit when the server cluster executes the computation task. When the electrical parameter matching degree is greater than a preset threshold, the trigger time of the potential overload event flag is obtained; When the trigger time of the potential overload event flag does not fall within the expected occurrence period, the power supply circuit is controlled to be de-energized; When the trigger time of the potential overload event flag falls within the expected occurrence time, the probability of the server cluster executing the computing task is determined based on the computing task execution status; When the probability of the server cluster executing a computing task is 1, the power supply circuit is controlled to supply power.

[0036] Specifically, when controlling the power supply circuit to supply or de-energize, the system first matches the currently monitored electrical parameters with the pre-stored electrical parameters for the computational task. These pre-stored parameters refer to the typical electrical parameter patterns exhibited by the power supply circuit under normal conditions when the server cluster is executing a specific computational task, such as specific current, voltage, and power curves or numerical ranges. By matching, the electrical parameter matching degree can be obtained, reflecting the similarity between the current electrical parameters and the pre-stored normal task electrical parameters. When this matching degree is greater than a preset threshold, it indicates that the current power consumption pattern is highly similar to the expected normal computational task pattern. At this point, the system will further acquire the trigger time of potential overload event flags.

[0037] Furthermore, the system determines the potential overload event based on the relationship between its trigger time and the expected occurrence period. If the trigger time does not fall within the expected occurrence period, it means the current potential overload event is not caused by the anticipated computing task, and therefore the power supply circuit is shut down to prevent actual overload damage. Conversely, if the trigger time falls within the expected occurrence period, it indicates that the potential overload event may be related to a computing task that is currently being executed or is about to be executed. In this case, the system determines the probability of the server cluster executing the computing task based on the computing task execution status. For example, if the computing task execution status shows that the task is in progress, the probability of executing the computing task may be determined to be 1. When the probability of executing the computing task is 1, the system determines that the current high load is benign and expected, and thus controls the power supply circuit to continue supplying power.

[0038] This application's solution constructs a more refined decision-making logic by introducing electrical parameter matching degree, comparison of trigger time with expected occurrence period, and judgment of the probability of executing computational tasks. When a potential overload event occurs, the system no longer simply judges based on the instantaneous high value of electrical parameters, but first identifies whether the current load conforms to the characteristics of known benign tasks by matching real-time electrical parameters with pre-stored electrical parameters of normal computational tasks. If the matching degree is high, it further verifies whether the trigger time of the potential overload event falls within the expected task execution period. This multi-dimensional, phased judgment mechanism can effectively distinguish between high load caused by normal computational tasks and genuine abnormal overloads, thereby avoiding accidental power outages to normal tasks and ensuring timely response to abnormal overloads.

[0039] In some preferred embodiments, suppose a server cluster executes a large data analysis task between 2:00 AM and 4:00 AM daily. This task causes a significant increase in the current, voltage, and power of the power supply circuit, but remains within safe limits. The system pre-stores the normal electrical parameter patterns during the execution of this data analysis task. At 2:30 AM one day, the current in the power supply circuit suddenly exceeds the preset current and persists for a period of time, triggering a potential overload event flag. At this time, the system first matches the currently monitored electrical parameters with the pre-stored electrical parameters of the data analysis task, finding a matching degree as high as 0.95, far exceeding the preset threshold of 0.8. Subsequently, the system obtains that the trigger time of the potential overload event flag is 2:30 AM, which clearly falls within the expected occurrence period (2:00 AM to 4:00 AM). Furthermore, the system queries the server cluster's computing task execution status, which shows that the data analysis task is in "running," thus determining the probability of executing the computing task to be 1. Based on these judgments, the system ultimately controls the power supply circuit to continue supplying power, thereby ensuring the smooth completion of the data analysis task and avoiding power outages caused by misjudgments.

[0040] In some embodiments, a more precise method for controlling the power supply circuit to supply or de-energize is proposed, the steps of which include: The electrical parameters are matched with the pre-stored computation task electrical parameters to obtain the electrical parameter matching degree between the electrical parameters and the pre-stored computation task electrical parameters. The pre-stored computation task electrical parameters are the electrical parameters of the power supply circuit when the server cluster executes the computation task. Obtain the trigger time of the potential overload event flag; Determine the time matching degree based on the trigger time and the expected occurrence period; Based on the execution status of the computing tasks, determine the probability of the server cluster executing computing tasks; A positive load confidence index is calculated based on the electrical parameter matching degree, the time matching degree, the probability of executing the calculation task, the preset electrical parameter matching degree weight, the preset time matching degree weight, and the preset probability weight of executing the calculation task, wherein the sum of the preset electrical parameter matching degree weight, the preset time matching degree weight, and the preset probability weight of executing the calculation task is 1. The power supply circuit is controlled to supply power or disconnect power based on the benign load confidence index.

[0041] Specifically, after a potential overload event flag is triggered, to more accurately determine the nature of the current load, it is first necessary to match the currently monitored electrical parameters with the pre-stored electrical parameters of the computing task. These pre-stored electrical parameters refer to the normal, typical electrical parameter patterns exhibited by the power supply circuit when the server cluster executes a specific computing task; for example, the typical change curves or numerical ranges of current, voltage, and power when a specific task starts. By comparing the similarity between the current electrical parameters and these pre-stored patterns, the electrical parameter matching degree can be obtained. This matching degree reflects the degree of agreement between the current load pattern and the known benign task pattern. For example, when the server cluster executes a big data analysis task, its power may increase significantly in a short period and remain elevated for a time. If this pattern is highly consistent with the pre-stored electrical parameter pattern for this task, then the electrical parameter matching degree will be high.

[0042] Meanwhile, obtaining the trigger time of the potential overload event flag is the basis for time matching. This trigger time refers to the specific moment when the system detects that the current exceeds the preset current and continues for more than the preset time point.

[0043] Furthermore, the time matching degree can be determined based on the trigger time and the expected occurrence period. The expected occurrence period is a time window estimated based on historical task data for the server cluster to execute a specific computing task. The time matching degree aims to assess whether the occurrence time of a potential overload event matches the expected task execution time. For example, if the trigger time falls precisely within the expected occurrence period of a high-load task, the time matching degree will be high, indicating that the potential overload may be related to the expected task execution.

[0044] Furthermore, the probability of the server cluster executing a computation task can be determined based on the execution status of the computation task. The execution status of the computation task can indicate whether a task is currently being executed, whether it is about to be executed, or whether it has been completed. For example, if the system explicitly indicates that a high-load task is being executed, the probability of executing the computation task is 1; if the task has not yet started or has been completed, the probability may be 0; if the status is ambiguous or in the preparation stage, the probability may be 0.5.

[0045] Building upon this foundation, this application introduces a benign load confidence index. This index is calculated by comprehensively considering electrical parameter matching degree, time matching degree, and the probability of executing computational tasks, combined with their respective preset weights. The sum of the preset weights for electrical parameter matching degree, time matching degree, and the probability of executing computational tasks is 1, ensuring a reasonable contribution of each index to the overall evaluation. The calculation of the benign load confidence index aims to provide a quantitative and comprehensive evaluation result to determine whether a potential overload event belongs to the benign, expected load of the server cluster, or a truly abnormal, dangerous overload.

[0046] Finally, the power supply circuit is controlled to supply power or cut off power based on the benign load confidence index. For example, when the benign load confidence index is high, it indicates that the current load is likely benign, and the power supply circuit will continue to supply power; when the index is low, it indicates that the current load may be abnormal, and the power supply circuit will be cut off to protect the equipment.

[0047] Through the above technical solution, this application significantly improves the accuracy and robustness of the low-voltage intelligent power control system in judging the load characteristics of server clusters. This solution introduces electrical parameter matching degree, time matching degree, and the probability of executing computational tasks, and combines this with a pre-weighted calculation of a benign load confidence index, achieving refined identification of potential overload events. This enables the system to effectively distinguish between normal, benign loads generated by server clusters performing high-intensity computational tasks and genuine abnormal, dangerous overloads, thereby avoiding unnecessary power outages and ensuring business continuity. Simultaneously, for genuine abnormal overloads, the system can also make timely power-off decisions based on more reliable confidence indicators, effectively protecting equipment safety and reducing operational risks. Compared to basic solutions, this application provides a more intelligent and reliable power control strategy through a multi-dimensional, weighted comprehensive evaluation approach.

[0048] In some preferred embodiments, a specific example is given below. Suppose a server cluster is scheduled to execute a big data processing task between 14:00 and 16:00 on a certain day. This task typically causes a specific increase in the current, voltage, and power of the power supply circuit. At 14:30, the system detects that the current in the power supply circuit exceeds a preset current and persists for more than a preset time, thereby triggering a potential overload event flag.

[0049] At this point, the system will initiate a refined judgment process: 1. Electrical Parameter Matching Degree Calculation: The system acquires the real-time electrical parameters of the current power supply circuit and matches them with the pre-stored electrical parameter patterns of big data processing tasks. Assuming that the current electrical parameter pattern is highly similar to the pre-stored pattern after matching, the calculated electrical parameter matching degree is 0.9.

[0050] 2. Time Match Determination: The trigger time for the potential overload event flag is 14:30. This time point clearly falls within the expected occurrence period (14:00-16:00). Assuming that according to preset rules, when the trigger time falls within the expected occurrence period and the time window of the expected occurrence period is less than the first time period threshold, the time match degree is recorded as 1.

[0051] 3. Determining the probability of executing the computation task: The system queries the execution status of computation tasks on the server cluster and finds that the big data processing task is currently in the "running" state. Therefore, the probability of executing the computation task is determined to be 1.

[0052] 4. Calculation of Beneficial Load Confidence Index: Assuming the preset electrical parameter matching degree weight is 0.4, the preset time matching degree weight is 0.3, and the preset execution calculation task probability weight is 0.3, then the Beneficial Load Confidence Index = (0.9 * 0.4) + (1 * 0.3) + (1 * 0.3) = 0.36 + 0.3 + 0.3 = 0.96.

[0053] 5. Control Decision: Assume the first preset index is 0.8 and the second preset index is 0.2. Since the calculated benign load confidence index of 0.96 is greater than the first preset index of 0.8, the system determines that the current potential overload is a benign load, and therefore controls the power supply circuit to continue supplying power.

[0054] In some embodiments, the step of matching the electrical parameters with the pre-stored computational task electrical parameters to obtain the electrical parameter matching degree between the electrical parameters and the pre-stored computational task electrical parameters includes: Obtain the current, voltage, and power of the power supply circuit; The current is compared with the pre-stored calculated task current to obtain the current similarity. The voltage is compared with the pre-stored calculated task voltage to obtain the voltage similarity. The power is compared with the power of the pre-stored calculated task to obtain the power similarity; The electrical parameters are calculated based on current similarity, voltage similarity, and power similarity to determine the electrical parameter matching degree between the electrical parameters and the pre-stored electrical parameters of the calculation task.

[0055] The acquisition of current, voltage, and power in the power supply circuit involves real-time collection of various electrical parameters of the power supply circuit, including instantaneous current, voltage, and power values, through a parameter monitoring module. These parameters are the foundational data for determining the current load status. The current is compared with the pre-stored computation task current to obtain current similarity, aiming to assess how closely the current current pattern resembles the typical current pattern of the server cluster performing a specific computation task. The pre-stored computation task current represents the current characteristics exhibited by the server cluster during normal execution of a specific computation task, such as its average value, peak value, and trend. Similarly, the voltage is compared with the pre-stored computation task voltage to obtain voltage similarity, aiming to assess how closely the current voltage pattern resembles the typical voltage pattern of the server cluster performing a specific computation task. The pre-stored computation task voltage represents the voltage characteristics exhibited by the server cluster during normal execution of a specific computation task. Finally, the power is compared with the pre-stored computation task power to obtain power similarity, aiming to assess how closely the current power pattern resembles the typical power pattern of the server cluster performing a specific computation task. The pre-stored computation task power represents the power characteristics exhibited by the server cluster during normal execution of a specific computation task. The electrical parameters are calculated based on current similarity, voltage similarity, and power similarity to determine the electrical parameter matching degree between the electrical parameters and the pre-stored electrical parameters of the computational task. The purpose is to comprehensively consider the matching degree of each electrical parameter to obtain an overall electrical parameter matching degree index. This index reflects the degree of consistency between the electrical parameter characteristics of the current power supply circuit and the electrical parameter characteristics of the server cluster when executing the preset computational task.

[0056] The above technical solution enables multi-dimensional and refined matching analysis of the electrical parameters of the power supply circuit, thereby more accurately determining whether potential overload events are caused by benign computing tasks of the server cluster. This refined matching mechanism helps improve the calculation accuracy of benign load confidence indicators, thereby enhancing the decision-making accuracy of intelligent power control, effectively avoiding unnecessary power outages due to misjudgments, and ensuring the stable operation of the server cluster.

[0057] In some embodiments, the step of comparing the current with the pre-stored calculated task current to obtain the current similarity can be further refined.

[0058] The step of comparing the current with the pre-stored calculated task current to obtain the current similarity includes: Obtain the current curve of the power supply circuit within a preset sliding time window; Extract the current rise rate and peak current from the current curve; The current similarity is obtained by comparing the current rise rate with the pre-stored calculated task current rise rate and the current peak value with the pre-stored calculated task current peak value.

[0059] Specifically, when acquiring the current curve of the power supply circuit within a preset sliding time window, the preset sliding time window can be understood as a continuous time interval that slides forward over time, thereby enabling real-time or near-real-time capture of the current change trend of the power supply circuit. By acquiring current data within the preset sliding time window, a current curve reflecting the change of current over time can be generated.

[0060] Furthermore, based on the acquired current curves, key characteristic parameters can be extracted, such as the current rise rate and peak current. The current rise rate refers to the rate of change of the current as it rapidly rises from a lower level to a higher level, reflecting the transient response characteristics at load startup or task initiation. The peak current refers to the maximum current reached within a preset sliding time window, reflecting the maximum demand of the load at a given moment. These characteristic parameters can be extracted using signal processing algorithms, such as differential operations and peak detection algorithms.

[0061] Subsequently, the extracted current rise rate is compared with the pre-stored current rise rate of the computational task, and the current peak value is also compared with the pre-stored current peak value of the computational task. The pre-stored current rise rate and pre-stored current peak value are typical current rise rates and current peak values ​​exhibited by the power supply circuit under normal conditions when the server cluster executes a specific computational task. By comparing these characteristics, the similarity between the current characteristics of the current power supply circuit and the current characteristics of a normal computational task can be quantified, thus obtaining the current similarity.

[0062] The above technical solution overcomes the limitations of relying solely on comparisons of single current values, significantly improving the accuracy and robustness of current similarity assessment. Specifically, by analyzing the current curve within a preset sliding time window and extracting the current rise rate and peak current, the unique current fingerprints of server clusters executing computing tasks can be more effectively identified, such as the rapid current surge at task startup and the stable peak current during task execution. This meticulous feature extraction and comparison mechanism enables the system to more accurately distinguish between current fluctuations caused by normal computing tasks and genuine abnormal overloads, thereby avoiding misjudgments and improving the reliability of intelligent power control.

[0063] In some embodiments, the steps of comparing the current rise rate with the pre-stored calculation task current rise rate and comparing the current peak value with the pre-stored calculation task current peak value to obtain the current similarity include: The current rise rate is compared with the pre-stored calculated task current rise rate, and the current rise rate deviation is calculated as follows: Current rise rate deviation = (Current rise rate - Pre-stored calculated task current rise rate) / Pre-stored calculated task current rise rate; The peak current is compared with the pre-stored calculated task current peak, and the peak current deviation is calculated as follows: peak current deviation = (peak current - pre-stored calculated task current peak) / pre-stored calculated task current peak. The current similarity is calculated using the geometric mean based on the deviation of the current rise rate and the deviation of the current peak value.

[0064] Specifically, the current rise rate deviation refers to the relative difference between the currently monitored current rise rate and the pre-stored current rise rate under normal power supply circuit conditions when the server cluster is performing computing tasks. This deviation is quantified by dividing the difference between the two by the pre-stored current rise rate of the computing task, aiming to standardize the comparison of current rise rates of different orders of magnitude and make them more comparable. The current peak deviation can be understood as the relative difference between the currently monitored current peak value and the pre-stored current peak value under normal power supply circuit conditions when the server cluster is performing computing tasks. This deviation is also calculated by dividing the difference between the two by the pre-stored current peak value of the computing task, aiming to eliminate the influence of absolute numerical differences on the comparison results, thereby more accurately reflecting the degree of matching of peak features. In practical applications, after obtaining the current rise rate deviation and current peak deviation, the geometric mean is used to calculate the current similarity. As an averaging method, the geometric mean can effectively balance the influence of different deviation values, especially suitable for scenarios that comprehensively evaluate multiple ratios or rates of change. Through the geometric mean, the degree of matching of the two key features, current rise rate and current peak value, can be comprehensively considered, thus obtaining a more comprehensive and robust current similarity index.

[0065] This application's solution transforms the comparison of current rise rate and current peak value from a simple qualitative judgment into a precise quantitative analysis by introducing the calculation of current rise rate deviation and current peak value deviation. This deviation calculation method allows current characteristics of different orders of magnitude to be standardized, thus avoiding comparison distortion caused by excessively large differences in absolute values. It is precisely because of this standardized deviation calculation that it becomes possible to subsequently use the geometric mean to comprehensively evaluate the matching degree of these two key characteristics. The geometric mean has unique advantages in processing ratio data; it ensures that even if the deviation of a certain feature is large, it will not completely dominate the similarity result, but will be balanced with the deviations of other features, thus providing a more robust and comprehensive similarity assessment. Therefore, this solution can more accurately reflect the degree of matching between the electrical parameter characteristics of the current power supply circuit and the normal electrical parameter characteristics of the server cluster performing computing tasks, providing a solid data foundation for determining whether a potential overload event is a benign load.

[0066] In some embodiments, the step of determining the time matching degree based on the trigger time and the expected occurrence period includes: When the time window of the expected occurrence period is less than the threshold of the first time period, and the trigger time falls within the expected occurrence period, the time matching degree is recorded as 1. When the time window of the expected occurrence period is greater than the first time period threshold and less than the second time period threshold, and the trigger time falls within the expected occurrence period, the time matching degree is recorded as 0.5, where the first time period threshold is less than the second time period threshold. If the time window of the expected occurrence period is greater than the threshold of the second time period or the trigger time does not fall within the expected occurrence period, the time matching degree is recorded as 0.

[0067] Specifically, the expected occurrence time window refers to the duration of the expected occurrence period for the server cluster to execute computing tasks. The first and second time window thresholds are preset time lengths used to classify the accuracy or certainty of the expected occurrence period. For example, the first time window threshold can be set to a shorter time, such as 5 minutes, indicating a very precise expected occurrence period; the second time window threshold can be set to a longer time, such as 30 minutes, indicating a certain range for the expected occurrence period. When the trigger time of a potential overload event falls precisely within a very short and certain expected occurrence period, it indicates that the event is highly likely to be related to the expected computing task, and therefore the time matching degree is assigned the highest value of 1. When the range of the expected occurrence period is moderate and the trigger time falls within it, a medium matching degree of 0.5 is given, reflecting a moderate degree of certainty. However, if the range of the expected occurrence period is too large, or the trigger time deviates completely from the expected occurrence period, the time matching degree is considered to be 0, indicating that the potential overload event and the expected computing task are not matched in time or have a very low matching degree.

[0068] Through the above technical solution, this application can perform multi-level quantification of time matching degree based on the accuracy of the predicted occurrence period and the actual occurrence time of the potential overload event, rather than simply judging whether there is a match. This refined time matching degree calculation method enables the system to more accurately weigh the impact of time factors when evaluating benign load confidence indicators. For example, it avoids assigning the same time matching degree to events occurring within a broad predicted period and events occurring within a precise predicted period, thereby improving the accuracy of judging whether a potential overload event is a benign load. This helps reduce misjudgments, ensures power supply or power outage control when truly needed, and improves the intelligence level and reliability of low-voltage smart power control methods.

[0069] In some embodiments, the step of determining the probability of the server cluster executing a computing task based on the execution status of the computing task includes: When the execution status of the computation task is "in execution", the probability of executing the computation task is determined to be 1. When the computation task status is "not executed", the probability of executing the computation task is determined to be 0. When the status of the computation task is ambiguous, the probability of executing the computation task is determined to be 0.5.

[0070] The "computation task execution status" refers to the actual situation of the computation tasks that the server cluster is currently processing or planning to process. Specifically, when the server cluster is clearly in an active computation task processing phase, its status is determined to be "in execution"; when the server cluster has not been assigned or is processing any computation tasks, its status is determined to be "not in execution"; and when the computation task status of the server cluster cannot be clearly determined, such as when it is in a task scheduling gap, waiting for resources, or when there is uncertainty, its status is determined to be "fuzzy". The "probability of executing a computation task" can be understood as the likelihood that the server cluster is currently executing a beneficial computation task.

[0071] This application's solution directly correlates the execution status of computational tasks with the probability of executing those tasks, providing a quantitative input based on the actual task status for subsequent calculations of benign load confidence indicators. Specifically, when the server cluster is clearly in an "executing" state, it is assigned the highest probability of executing a computational task (i.e., 1), indicating that the current power consumption is highly likely caused by normal computational tasks, thus enhancing confidence that the current load is benign. Conversely, when the server cluster is in an "not executing" state, it is assigned the lowest probability of executing a computational task (i.e., 0), meaning that the current power consumption is unrelated to normal computational tasks, thus reducing confidence that the current load is benign. For the "fuzzy" state, an intermediate probability value (i.e., 0.5) is assigned to reflect its uncertainty and avoid overconfidence or overly pessimistic judgments. This probability assignment mechanism based on task status allows the system to more accurately combine the actual working conditions of the server cluster when assessing potential overload events.

[0072] Through the above technical solution, this application can objectively and quantitatively determine the probability of a server cluster executing a computing task based on its computing task execution status. This explicit probability assignment mechanism effectively avoids subjectivity and uncertainty in judging whether a server cluster is executing a benign computing task, thereby improving the accuracy and reliability of the benign load confidence index calculation. Therefore, when facing potential overload events, the system can more accurately determine whether the current power consumption belongs to a normal computing task load, and thus make more reasonable and timely power supply or power outage decisions, effectively ensuring the stable operation of the server cluster and avoiding unnecessary power outages or safety hazards caused by misjudgments.

[0073] In some embodiments, the steps of controlling the power supply circuit to supply or cut off power based on the benign load confidence index include one of the following: When the confidence index of a benign load is greater than the first preset index, the power supply circuit is controlled to supply power. When the confidence index of a benign load is less than the second preset index, the power supply circuit is cut off, wherein the first preset index is greater than the second preset index. When the confidence index of benign load is less than or equal to the first preset index and greater than or equal to the second preset index, a manual review is requested, and the power supply circuit is controlled to supply power or cut off power based on the result of the manual review.

[0074] Specifically, the benign load confidence index is a quantitative indicator calculated by comprehensively considering factors such as electrical parameter matching degree, time matching degree, and the probability of executing computational tasks. It is used to assess whether a potential overload event is caused by benign computational tasks in the server cluster. The first and second preset indices are pre-set thresholds used to divide the benign load confidence index into different ranges to guide different control decisions. The first preset index is typically set to a higher value, indicating high confidence in the benign load; the second preset index is typically set to a lower value, indicating insufficient confidence in the benign load or the existence of potential risks. The first preset index is greater than the second preset index, thus creating a decision ambiguity range between the two.

[0075] When the benign load confidence index is greater than the first preset index, it indicates that the system has high confidence that the current load is benign. At this time, the power supply circuit is controlled to power supply state to ensure the normal operation of the server cluster.

[0076] When the confidence index of benign load is less than the second preset index, it indicates that the system is not confident that the current load is benign or believes that there is an overload risk. At this time, the power supply circuit is controlled to be in a power-off state to avoid potential equipment damage or safety accidents.

[0077] When the benign load confidence index falls between the first and second preset indices (i.e., the benign load confidence index is less than or equal to the first preset index and greater than or equal to the second preset index), the system cannot automatically make a clear power supply or power cut-off decision. In this case, the system will request manual review, with maintenance personnel or experts intervening to manually assess and intervene in the current situation. Based on the results of the manual review, the system will ultimately control the power supply circuit to either supply power or cut off power, thereby avoiding misjudgments.

[0078] This application's solution divides the range of the benign load confidence index into clearly defined power supply, power outage, and manual verification zones by introducing a first and a second preset index. When the benign load confidence index clearly indicates a benign load (greater than the first preset index) or a non-benign load (less than the second preset index), the system can automatically and quickly make a power supply or power outage decision, ensuring timely response. However, when the benign load confidence index is in an ambiguous intermediate zone—that is, when the benign load confidence index is less than or equal to the first preset index but greater than or equal to the second preset index—the system no longer blindly makes automatic decisions. Instead, it requests manual verification, incorporating human experience and judgment to comprehensively analyze complex situations, thereby avoiding the risk of misjudgment that may arise from automated decision-making. This hierarchical control strategy effectively compensates for the shortcomings of purely automated decision-making in handling critical situations, improving the accuracy and reliability of decisions.

[0079] Meanwhile, this application proposes a low-voltage smart power control system based on the Internet of Things, which includes: The identification module is used to identify the power supply circuit of the server cluster; The parameter monitoring module is used to monitor the electrical parameters of the power supply circuit, including current, voltage, and power. A potential overload event triggering module is used to trigger a potential overload event flag when the current exceeds a preset current and the duration exceeds a preset time. The data acquisition module is used to obtain the expected occurrence time and execution status of the computing tasks of the server cluster in response to the potential overload event flag, wherein the expected occurrence time is based on the time period estimated from historical tasks. The control module is used to control the power supply circuit to supply power or cut off power based on the electrical parameters, the expected time period, and the execution status of the calculation task.

[0080] This system integrates identification, parameter monitoring, potential overload event triggering, data acquisition, and control modules to achieve refined management of the power supply circuits for server clusters. By monitoring electrical parameters in real time and combining this with information on the server cluster's computing tasks, the system can intelligently determine the nature of potential overload events, thereby avoiding unnecessary power outages caused by misjudgments in traditional systems and effectively ensuring the continuity of critical business operations and power safety.

[0081] Specifically, the identification module can be a standalone hardware unit, such as an embedded controller, which integrates a communication interface and storage unit to store and manage the configuration information of the power supply circuit. Alternatively, the identification module can be a software service deployed in the cloud or on a local server, interacting with field devices via a network interface to achieve automatic discovery and registration of the power supply circuit.

[0082] The parameter monitoring module can use hardware devices such as smart meters, split-type current transformers, and voltage sensors to collect electrical parameter data of the power supply circuit in real time. These hardware devices transmit the collected data to the software part of the parameter monitoring module for processing and storage via wired or wireless communication methods (such as Wi-Fi, LoRa, NB-IoT, etc.).

[0083] The potential overload event triggering module can be a standalone software service or a sub-function integrated into the control module. Its working principle involves continuously receiving real-time electrical parameter data from the parameter monitoring module and comparing it with preset current and time thresholds. When the triggering conditions are met, the module generates a potential overload event flag and sends it to other relevant modules.

[0084] The data acquisition module can interface with the server cluster's management system or task scheduling platform. For example, the data acquisition module can query the server cluster's management system for information such as currently running computing tasks, their estimated start and end times, and task priorities by calling API interfaces. The estimated occurrence time period can be obtained based on analysis and prediction of historical task data, such as training historical task logs using machine learning models to predict the occurrence time of future tasks.

[0085] The control module is the core decision-making unit of the system. It receives electrical parameters from the parameter monitoring module, computational task information from the data acquisition module, and flags from the potential overload event triggering module. Based on this comprehensive information, the control module executes intelligent judgment logic. For example, when a potential overload event flag is triggered, the control module assesses whether it is currently in a peak period of expected computational tasks and whether the server cluster is performing high-intensity computational tasks. Based on the judgment result, the control module sends a power supply or power cut-off command to the actuator of the power supply circuit (such as a smart circuit breaker).

[0086] The IoT-based low-voltage smart power control system proposed in this application aims to solve the misjudgment problem of existing smart power systems when dealing with instantaneous high load surges in server clusters. Traditional systems often simply determine overload and cut off power supply based on whether the current exceeds a preset threshold. This "one-size-fits-all" approach cannot distinguish between functional load peaks during normal business operations and genuine electrical faults.

[0087] This application significantly improves the intelligence level of power control by incorporating consideration of the estimated occurrence time and execution status of server cluster computing tasks. Specifically, through the collaborative work of an identification module, a parameter monitoring module, a potential overload event triggering module, a data acquisition module, and a control module, the system can more accurately determine the nature of potential overload events. For example, when the system detects that the current in the power supply circuit exceeds a preset value and persists for a period of time, triggering a potential overload event flag, the control module will not immediately take power-off measures. Instead, the data acquisition module will further acquire the estimated occurrence time of the server cluster's computing tasks and the current execution status of the computing tasks. The estimated occurrence time is estimated based on historical task data, reflecting the time window when the server cluster may experience high loads under normal business conditions. The computing task execution status provides real-time information on whether the server cluster is currently performing high-intensity computing tasks.

[0088] By comprehensively analyzing electrical parameters, the expected time period, and the execution status of computing tasks, the control module of this application can more accurately determine the nature of potential overload events. For example, if a potential overload event occurs during a period when the server cluster is expected to perform high-intensity computing tasks, and the server cluster is indeed performing computing tasks, the system can determine that this is likely a benign, functional load peak required by the business, rather than a dangerous electrical fault. In this case, the system will choose to continue supplying power to ensure the continuity of critical business operations. Conversely, if a potential overload event occurs outside of the expected task period, or if the server cluster is not performing high-intensity computing tasks, the system will determine that this may be a genuine overload risk and will take power-off measures to protect the equipment and lines.

[0089] This multi-dimensional and intelligent judgment mechanism enables the system of this application to effectively avoid unnecessary power outages caused by misjudgments in traditional systems, greatly improving the reliability of power supply to the server cluster and the continuity of business. Compared with existing technologies, the core innovation of this application lies in its introduction of a deep understanding of load behavior patterns, that is, intelligently interpreting abnormal fluctuations in electrical parameters by combining the business characteristics of the server cluster (the expected occurrence time and execution status of computing tasks). This system breaks through the limitations of traditional systems that rely solely on static electrical parameter thresholds for judgment, realizing a transformation from "passive response" to "intelligent prediction".

[0090] The foregoing has provided a detailed description of the preferred embodiments of this application. However, this application is not limited to the above-described embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined in this application.

Claims

1. A low-voltage smart power control method based on the Internet of Things, characterized in that, Includes the following steps: Identify the power supply circuit of the server cluster; Monitor the electrical parameters of the power supply circuit, including current, voltage, and power; When the current exceeds a preset current and the duration exceeds a preset time, a potential overload event flag is triggered; In response to the potential overload event flag, the expected occurrence time of the computing tasks in the server cluster and the execution status of the computing tasks are obtained, wherein the expected occurrence time is a time period estimated based on historical tasks; Based on the electrical parameters, the expected time period, and the execution status of the calculation task, the power supply circuit is controlled to supply power or cut off power.

2. The method according to claim 1, characterized in that, The step of controlling the power supply circuit to supply power or cut off power based on the electrical parameters, the expected occurrence time period, and the execution status of the calculation task includes: The electrical parameters are matched with the pre-stored computation task electrical parameters to obtain the electrical parameter matching degree between the electrical parameters and the pre-stored computation task electrical parameters. The pre-stored computation task electrical parameters are the electrical parameters of the power supply circuit when the server cluster executes the computation task. When the electrical parameter matching degree is greater than a preset threshold, the trigger time of the potential overload event flag is obtained; When the trigger time of the potential overload event flag does not fall within the expected occurrence period, the power supply circuit is controlled to be de-energized; When the trigger time of the potential overload event flag falls within the expected occurrence period, the probability of the server cluster executing the computing task is determined based on the computing task execution status. When the probability of the server cluster executing a computing task is 1, the power supply circuit is controlled to supply power.

3. The method according to claim 1, characterized in that, The step of controlling the power supply circuit to supply power or cut off power based on the electrical parameters, the expected occurrence time period, and the execution status of the calculation task includes: The electrical parameters are matched with the pre-stored computation task electrical parameters to obtain the electrical parameter matching degree between the electrical parameters and the pre-stored computation task electrical parameters. The pre-stored computation task electrical parameters are the electrical parameters of the power supply circuit when the server cluster executes the computation task. Obtain the trigger time of the potential overload event flag; Determine the time matching degree based on the trigger time and the expected occurrence period; Based on the execution status of the computing tasks, determine the probability of the server cluster executing computing tasks; A positive load confidence index is calculated based on the electrical parameter matching degree, the time matching degree, the probability of executing the calculation task, the preset electrical parameter matching degree weight, the preset time matching degree weight, and the preset probability weight of executing the calculation task, wherein the sum of the preset electrical parameter matching degree weight, the preset time matching degree weight, and the preset probability weight of executing the calculation task is 1. The power supply circuit is controlled to supply power or disconnect power based on the benign load confidence index.

4. The method according to claim 3, characterized in that, The step of matching the electrical parameters with the pre-stored calculation task electrical parameters to obtain the electrical parameter matching degree between the electrical parameters and the pre-stored calculation task electrical parameters includes: Obtain the current, voltage, and power of the power supply circuit; The current is compared with the pre-stored calculated task current to obtain the current similarity. The voltage is compared with the pre-stored calculated task voltage to obtain the voltage similarity. The power is compared with the power of the pre-stored calculated task to obtain the power similarity; The electrical parameter matching degree between the electrical parameters and the pre-stored calculation task electrical parameters is calculated based on the current similarity, voltage similarity, and power similarity.

5. The method according to claim 4, characterized in that, The step of comparing the current with the pre-stored calculated task current to obtain the current similarity includes: Obtain the current curve of the power supply circuit within a preset sliding time window; Extract the current rise rate and peak current from the current curve; The current similarity is obtained by comparing the current rise rate with the pre-stored calculated task current rise rate and the current peak value with the pre-stored calculated task current peak value.

6. The method according to claim 5, characterized in that, The steps of comparing the current rise rate with the pre-stored calculated task current rise rate and comparing the current peak value with the pre-stored calculated task current peak value to obtain the current similarity include: The current rise rate is compared with the pre-stored calculated task current rise rate, and the current rise rate deviation is calculated as follows: Current rise rate deviation = (Current rise rate - Pre-stored calculated task current rise rate) / Pre-stored calculated task current rise rate; The current peak value is compared with the pre-stored calculated task current peak value, and the current peak value deviation is calculated as follows: current peak value deviation = (current peak value - pre-stored calculated task current peak value) / pre-stored calculated task current peak value. The current similarity is calculated using the geometric mean based on the deviation of the current rise rate and the deviation of the current peak value.

7. The method according to claim 3, characterized in that, The step of determining the time matching degree based on the trigger time and the expected occurrence period includes one of the following: When the time window of the expected occurrence period is less than the first time period threshold, and the trigger time falls within the expected occurrence period, the time matching degree is recorded as 1. When the time window of the expected occurrence period is greater than the first time period threshold and less than the second time period threshold, and the trigger time falls within the expected occurrence period, the time matching degree is recorded as 0.5, wherein the first time period threshold is less than the second time period threshold. If the time window of the expected occurrence period is greater than the second time period threshold or the trigger time does not fall within the expected occurrence period, the time matching degree is recorded as 0.

8. The method according to claim 3, characterized in that, The step of determining the probability of the server cluster executing a computing task based on the execution status of the computing task includes: When the execution status of the computation task is "in execution", the probability of executing the computation task is determined to be 1. When the computation task status is "not executed", the probability of executing the computation task is determined to be 0. When the status of the computation task is ambiguous, the probability of executing the computation task is determined to be 0.

5.

9. The method according to claim 3, characterized in that, The step of controlling the power supply circuit to supply power or cut off power based on the benign load confidence index includes one of the following: When the benign load confidence index is greater than the first preset index, the power supply circuit is controlled to supply power. When the benign load confidence index is less than the second preset index, the power supply circuit is controlled to be cut off, wherein the first preset index is greater than the second preset index; When the benign load confidence index is less than or equal to the first preset index and greater than or equal to the second preset index, a manual review is requested, and the power supply circuit is controlled to supply power or cut off power based on the manual review result.

10. A low-voltage smart power control system based on the Internet of Things, characterized in that, The system includes: The identification module is used to identify the power supply circuit of the server cluster; The parameter monitoring module is used to monitor the electrical parameters of the power supply circuit, including current, voltage, and power. A potential overload event triggering module is used to trigger a potential overload event flag when the current exceeds a preset current and the duration exceeds a preset time. The data acquisition module is used to obtain the expected occurrence time and execution status of the computing tasks of the server cluster in response to the potential overload event flag, wherein the expected occurrence time is a time period estimated based on historical tasks. The control module is used to control the power supply circuit to supply power or cut off power based on the electrical parameters, the expected time period, and the execution status of the calculation task.

Images