Power sampling data completion method and device, and computing device

Abstract

Embodiments of the present application provide a power sampling data completion method and device and a computing device. The method comprises: sampling a first power value corresponding to a first time and a second power value corresponding to a second time; wherein the second time is after the first time, and power sampling data between the first time and the second time is lost; determining a target power value corresponding to at least one to-be-completed time in a to-be-completed period according to the first power value, the second power value and an actual power consumption corresponding to the to-be-completed period; wherein the to-be-completed period is all or part of the period between the first time and the second time. Embodiments of the present application calibrate the target power value calculated based on the first power value and the second power value by using the actual power consumption, can anchor the completion result on the real actual power consumption in the scenario of data loss, thereby avoiding the power curve from deviating from the actual curve and the situation of power curve drift.

Description

Technical Field

[0001] This application relates to the field of infrastructure operation and maintenance technology, and in particular to a power sampling data completion method, apparatus and computing device. Background Technology

[0002] In data center infrastructure operation and maintenance scenarios, rack power curves are typically collected by the Data Center Infrastructure Management (DCIM) system. In actual operating environments, due to intermittent sensor failures, network communication interruptions, or system maintenance, the rack power time-series data collected by the DCIM system often suffers from partial omissions.

[0003] To address missing time-series power data for server racks, common industry-standard data completion methods include statistical interpolation and filtering, such as linear interpolation, spline interpolation, EMA (exponential moving average), and Kalman filtering. These methods quickly fill in missing points using values ​​from nearby time points. However, these methods primarily rely on sampled data from near-missing time points for estimation and are suitable for short-term data loss scenarios. In scenarios with long-term data loss, the completed power curve may drift, resulting in a significant deviation from the actual power value. Summary of the Invention

[0004] This application provides a power sampling data completion method, apparatus, and computing device. It uses actual power consumption to calibrate the target power value calculated based on a first power value and a second power value. In scenarios where data is missing, the completion result can be anchored to the actual power consumption, thereby avoiding the completed power curve from deviating from the actual curve and causing power curve drift.

[0005] To achieve the above objectives, the embodiments of this application adopt the following technical solutions: In a first aspect, embodiments of this application provide a power sampling data completion method, the method comprising: acquiring a first power value corresponding to a first time moment and a second power value corresponding to a second time moment; wherein, the second time moment is after the first time moment, and the power sampling data of the power curve on the digital twin corresponding to the first time moment and the second time moment is lost; determining a target power value corresponding to at least one time moment in the time period to be completed based on the first power value, the second power value and the actual power consumption corresponding to the time period to be completed; wherein, the time period to be completed is a complete time period or a partial time period between the first time moment and the second time moment.

[0006] Based on this scheme, firstly, the first and second power values ​​corresponding to the start point (e.g., the first moment) and end point (e.g., the second moment) of the missing sampling data period to be completed are obtained. These first and second power values ​​are used to define the boundaries of the power curve to be completed, providing a basis for subsequent calculations. Then, the actual power consumption of the period to be completed is obtained and used as a global energy consumption constraint within that period. Thus, the missing target power value within the period to be completed is completed based on the first power value, the second power value, and the actual power consumption. The target power value not only depends on the start and end points of the period to be completed but also needs to meet the actual power consumption. By linking the interpolation results within the period to be completed with the actual energy consumption, the cumulative drift error caused by missing data over a period of time can be effectively suppressed. In summary, the power sampling data completion method provided in this application can calibrate the power curve calculated based on the start and end points using actual power consumption. In scenarios with missing data, the completion result can be anchored to the actual power consumption, thereby avoiding deviation of the completed power curve from the actual curve and preventing power curve drift.

[0007] In one possible implementation, determining the target power value corresponding to at least one moment in the time period to be completed, based on the first power value, the second power value, and the actual power consumption corresponding to the time period to be completed, includes: determining the initial power value corresponding to at least one moment in the time period to be completed based on the first power value and the second power value; calibrating the initial power value based on the actual power consumption corresponding to the time period to be completed to obtain the target power value.

[0008] Based on this scheme, the initial power value at the time to be filled is first obtained based on the first power value and the second power value. This can effectively utilize the fixed boundary value of the missing window. Then, the initial power value obtained based only on the first and second times is calibrated using the actual power consumption as a benchmark to obtain the calibrated target power value. This can effectively utilize the global actual power consumption of the time period to be filled and improve the completion accuracy in the data missing scenario.

[0009] In one possible implementation, the target power value is the total power value corresponding to all devices in the rack, including computing devices and non-computing devices. Determining an initial power value corresponding to at least one time point to be completed, based on a first power value and a second power value, includes: determining an initial power value corresponding to at least one time point to be completed based on power sample values ​​of the computing devices at at least one time point to be completed and reference power values ​​of the non-computing devices; determining a first power offset of the rack at the first time point based on the first power value, the power sample values ​​of the computing devices at the first time point, and the reference power value; determining a second power offset of the rack at the second time point based on the second power value, the power sample values ​​of the computing devices at the second time point, and the reference power value; and adjusting the initial power value based on the first power offset and the second power offset.

[0010] Based on this scheme, the initial power value of the server rack at each time point to be completed is first estimated using the real-time power sampling values ​​of the computing equipment and the baseline power value of the non-computing equipment. Then, using the first and second power values ​​of the server rack as anchor points, the first and second power offsets at the first and second times are determined respectively. A linear transition method is used to transfer the first and second power offsets to each time point to be completed within the time period, correcting the offset of the initial power value at each time point to be completed. In this way, not only can the initial power value be accurately aligned with the actual power value at the sampling points without missing data (the first and second times), but the offset can also be transferred to the entire time period to be completed, correcting the power curve offset caused by inaccurate baseline power value estimation or server power sampling errors. This helps to improve the accuracy of power sampling data completion, thereby obtaining a draft of a power curve with aligned endpoints and a reasonable shape.

[0011] In one possible implementation, after determining the initial power value corresponding to at least one time period to be completed based on the first power value and the second power value, the method further includes: if a target event occurs during the time period to be completed, determining the occurrence time of the target event and the power step direction caused by the target event; wherein the power step direction is either upward or downward; and adjusting the initial power value based on the occurrence time and the power step direction.

[0012] Based on this scheme, the initial power value is stepped to correct the time of occurrence of the target event and its surrounding area. This allows the completed target power value to reflect the actual situation such as power supply switching, avoiding the distortion problem that occurs when the traditional smoothing difference occurs at the time of the target event.

[0013] In one possible implementation, calibrating the initial power value based on the actual power consumption corresponding to the time period to be completed includes: determining the predicted power consumption corresponding to the time period to be completed based on the initial power value corresponding to at least one moment to be completed within the time period; and calibrating the initial power value based on the difference between the actual power consumption and the predicted power consumption corresponding to the time period to be completed.

[0014] Based on this scheme, the difference between the predicted power consumption and the actual power consumption is calculated using the actual power consumption as a benchmark. Then, the initial power value corresponding to each time point to be filled is calibrated based on the difference by means of overall scaling, overall translation or local weighting, etc., so that the calibrated predicted power consumption is accurately matched with the actual power consumption. This allows the accuracy of power sampling data completion in the data missing scenario to be improved by utilizing the energy consumption constraints in the time period to be filled.

[0015] In one possible implementation, the initial power value is calibrated based on the difference between the actual power consumption and the predicted power consumption corresponding to the time period to be filled, including: determining the adjustment weight corresponding to at least one time moment to be filled within the time period to be filled; allocating the difference to at least one time moment to be filled based on the adjustment weight to obtain the power adjustment amount corresponding to each time moment to be filled; and adjusting the initial power value corresponding to each time moment to be filled based on the power adjustment amount corresponding to each time moment to be filled.

[0016] Based on this scheme, by using a local weighting method, an adjustment weight is assigned to each moment to be completed. Then, based on the adjustment weight, the difference between the actual power consumption and the predicted power consumption is allocated, which can achieve fine point-by-point calibration of the initial power value and improve the completion accuracy of the power sampling data.

[0017] In one possible implementation, the rate of change of the target power value between any two adjacent moments within the time period to be completed is less than a preset power change threshold.

[0018] Based on this scheme, constraining the rate of power change between adjacent time points can prevent steep changes in the power curve that violate electrical characteristics during the time period to be completed, making the completion result more consistent with the actual situation.

[0019] In one possible implementation, the method further includes adjusting the target power value based on the sum of the power sample value and the reference power value.

[0020] Based on this solution, and taking into account the power topology relationship between the server rack and computing devices, the target power value of the server rack is verified and corrected using the power sampling data of the computing devices. This ensures the consistency between the server rack power and the power on the device side, and improves the accuracy of the target power value.

[0021] Secondly, embodiments of this application also provide a power sampling data completion device, which includes: a data acquisition module, used to acquire a first power value corresponding to a first time moment and a second power value corresponding to a second time moment; wherein the second time moment is after the first time moment, and the power sampling data of the power curve on the digital twin corresponding to the first time moment and the second time moment is lost; a data completion module, used to determine a target power value corresponding to at least one time moment in the time period to be completed based on the first power value, the second power value and the actual power consumption corresponding to the time period to be completed; wherein the time period to be completed is a complete time period or a partial time period between the first time moment and the second time moment.

[0022] In one possible implementation, the data completion module is used to: determine an initial power value corresponding to at least one time period to be completed based on a first power value and a second power value; and calibrate the initial power value based on the actual power consumption corresponding to the time period to be completed to obtain a target power value.

[0023] In one possible implementation, the target power value is the total power value corresponding to all devices in the rack, including computing devices and non-computing devices. The data completion module is used to: determine an initial power value corresponding to at least one time to be completed based on the power sample value of the computing device at at least one time to be completed and the reference power value of the non-computing device; determine a first power offset of the rack at the first time based on the first power value, the power sample value of the computing device at the first time, and the reference power value; determine a second power offset of the rack at the second time based on the second power value, the power sample value of the computing device at the second time, and the reference power value; and adjust the initial power value based on the first power offset and the second power offset.

[0024] In one possible implementation, the data completion module is used to: determine the occurrence time of the target event and the power step direction caused by the target event when the target event occurs within the time period to be completed; wherein the power step direction is either upward or downward; and adjust the initial power value based on the occurrence time and the power step direction.

[0025] In one possible implementation, the data completion module is used to: determine the predicted power consumption corresponding to the time period to be completed based on the initial power value corresponding to at least one time point to be completed within the time period to be completed; and calibrate the initial power value based on the difference between the actual power consumption and the predicted power consumption corresponding to the time period to be completed.

[0026] In one possible implementation, the data completion module is used to: determine the adjustment weight corresponding to at least one time point to be completed within the time period to be completed; based on the adjustment weight, allocate the difference to at least one time point to be completed to obtain the power adjustment amount corresponding to each time point to be completed; and adjust the initial power value corresponding to each time point to be completed according to the power adjustment amount corresponding to each time point to be completed.

[0027] In one possible implementation, the rate of change of the target power value between any two adjacent moments within the time period to be completed is less than a preset power change threshold.

[0028] In one possible implementation, the data completion module is used to adjust the target power value based on the sum of the power sample value and the reference power value.

[0029] Thirdly, embodiments of this application also provide a computing device, including: a processor and a memory; the processor and the memory are coupled; the memory is used to store program instructions; the processor is used to execute the program instructions to perform the method as described in any of the first aspects above.

[0030] Fourthly, embodiments of this application provide a chip for performing the methods described in any of the first aspects above.

[0031] Fifthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a computer, implement the method as described in any of the first aspects.

[0032] In a sixth aspect, embodiments of this application provide a program product including a computer program that, when executed by a processor, implements the method as described in any of the first aspects.

[0033] It is understood that the beneficial effects that the technical solutions provided in the second to sixth aspects above can achieve can be referred to the beneficial effects in the first aspect and any of its implementation methods, and will not be repeated here. Attached Figure Description

[0034] Figure 1 This is a schematic diagram showing missing power timing data for a server rack. Figure 2 This is an application scenario diagram of the power sampling data completion method provided in the embodiments of this application; Figure 3 This is a flowchart of a power sampling data completion method provided in an embodiment of this application; Figure 4 This is a schematic diagram of a power curve before and after completion, provided in an embodiment of this application; Figure 5 This is a flowchart illustrating the determination of a target power value provided in an embodiment of this application; Figure 6 This is a flowchart illustrating the determination of an initial power value provided in an embodiment of this application; Figure 7 This is a flowchart of an embodiment of the present application that provides a power step correction of the initial power value; Figure 8 This is a flowchart of a calibration initial power value provided in an embodiment of this application; Figure 9 This is another flowchart of the calibration initial power value provided in the embodiments of this application; Figure 10 This is a flowchart illustrating the adjustment of a target power value provided in an embodiment of this application; Figure 11 This is another flowchart of the power sampling data completion method provided in the embodiments of this application; Figure 12 This is a schematic diagram of a power sampling data completion device provided in an embodiment of this application; Figure 13 This is a schematic diagram of a computing device provided in an embodiment of this application. Detailed Implementation

[0035] The technical solutions of the embodiments of this application will now be described with reference to the accompanying drawings. To facilitate a clear description of the technical solutions of the embodiments of this application, the use of terms such as "first," "second," etc., in the embodiments of this application is for illustrative purposes and to distinguish the objects being described. There is no particular order between them, nor does it indicate a specific limitation on the number of devices in the embodiments of this application, and they do not constitute any limitation on the embodiments of this application.

[0036] The application background and application scenarios of this application will be explained below.

[0037] The following is combined Figure 1 The application background of the power sampling data completion method provided in this application is illustrated by example.

[0038] Figure 1 This is a schematic diagram showing that the power timing data of a server rack is missing.

[0039] In data center infrastructure operation and maintenance scenarios, rack power curves are typically collected by the Data Center Infrastructure Management (DCIM) system. However, in actual operating environments, due to intermittent sensor failures, network communication interruptions, or system maintenance, the rack power time-series data collected by the DCIM system often suffers from partial omissions. For example... Figure 1 In data centers based on digital twin technology, collected power data can be displayed on a digital twin of the server management platform. However, when the digital twin visualizes the rack power curves, data gaps may occur due to data acquisition anomalies. For example... Figure 1 The left side of T1 shows the power curve before the data was missing, and the right side of T2 shows the power curve after the data was missing. The period from T1 to T2 is the period when the data was missing, and it is also the period when the power sampling data needs to be supplemented.

[0040] To address the aforementioned issues, this application provides a power sampling data completion method, apparatus, and computing device. When completing power sampling data, this method references the actual power consumption during the period to be completed and constrains the completion result by using the actual power consumption during the period to be completed, thereby reducing the cumulative deviation in scenarios with missing data and solving the problems of low completion accuracy and easy drift of power curves in scenarios where data is missing for a period of time using traditional methods.

[0041] The following is combined Figure 2 The following provides an illustrative example of the application scenarios for the power sampling data completion method provided in this application.

[0042] Figure 2 This is an application scenario diagram of the power sampling data completion method provided in the embodiments of this application.

[0043] like Figure 2 As shown, the power sampling data completion method provided in this application can be applied to data centers based on digital twin technology. The data center includes four layers from bottom to top: infrastructure layer, data acquisition layer, digital twin layer, and visualization layer.

[0044] The infrastructure layer includes server racks, which house several power-consuming computing devices (such as servers) and non-computing devices (not shown in the figure).

[0045] The data acquisition layer includes a DCIM system, which can sample the total power of all devices in the rack to obtain rack power time-series data. The rack power time-series data includes the total power value of all devices in the rack at each moment. The data acquisition layer also includes an IT equipment management system, which can sample the power of each server in the rack to obtain power time-series data for each server. The power time-series data for each server includes the power sample value of that server. Theoretically, the rack power time-series data should be equal to the sum of the power sample values ​​of all servers in the rack and the power of non-computing devices. The data acquisition layer also includes a configuration management database (CMDB), which can manage the infrastructure in the data center, such as racks, servers, power supplies, etc., and can manage the topology relationships between infrastructure.

[0046] The digital twin layer includes a power sampling data completion device, capable of acquiring rack power time-series data and the actual power consumption of all devices within the rack during the completion period from the DCIM system; it can also acquire power topology link data from the CMDB; and it can obtain alarm information corresponding to the rack from the data acquisition layer and server power time-series data from the IT equipment management system. In this application scenario, the power sampling data completion device checks for gaps in the rack power time-series data and then completes the rack power sampling data based on the actual power consumption of the rack during the completion period. Furthermore, during the power sampling data completion process, the device can also refer to the power topology links obtained from the CMDB and the server power time-series data obtained from the IT equipment management system. Here, the server power time-series data refers to the sum of the power time-series data of all servers within the rack, which can be acquired individually or directly as a sum. The list of all servers within the rack can be determined based on the power topology links. In this way, compared with the traditional solution that generates a digital twin based on the collected power sampling data, the embodiments of this application can use the completed power sampling data to adjust the digital twin and obtain the completed cabinet power curve.

[0047] The visualization layer includes terminal devices such as laptops, tablets, and displays, which can be used to access the digital twin and display the completed cabinet power curve.

[0048] The power sampling data completion method provided in this application first obtains the first power value and the second power value corresponding to the start point (e.g., the first moment) and end point (e.g., the second moment) of the missing sampling data period to be completed. This first and second power values ​​define the boundaries of the power curve to be completed, providing a basis for subsequent calculations. Then, the actual power consumption of the period to be completed is obtained and used as a global energy consumption constraint within that period. This global energy consumption constraint means that the integral (total power consumption) of the completed power curve within the period to be completed is equal to the actual power consumption within that period. Thus, based on the first power value, the second power value, and the actual power consumption within the period to be completed, the missing target power value is completed. The target power value not only depends on the start and end points of the period to be completed but also needs to satisfy that the actual power consumption within the period to be completed is the same as the predicted power consumption. By associating the interpolation result within the period to be completed with the actual energy consumption of that period, the cumulative drift error due to missing data can be effectively suppressed. In summary, the power sampling data completion method provided in this application uses the actual power consumption during the time period to be completed to calibrate the power curve calculated based on the start and end points. In scenarios where data is missing, the completion result can be anchored to the actual power consumption during the time period to be completed, thereby avoiding the completed power curve from deviating from the actual curve and causing power curve drift.

[0049] The following is combined Figures 3 to 11 This example illustrates a method for completing power sampling data.

[0050] Figure 3 This is a flowchart of a power sampling data completion method provided in an embodiment of this application.

[0051] Figure 4 This is a schematic diagram of a power curve before and after completion, provided in an embodiment of this application.

[0052] like Figure 3 As shown, this embodiment may include the following steps S11-S12: S11: Obtain the first power value corresponding to the first time and the second power value corresponding to the second time; if the first time is less than the second time, the power sampling data of the power curve on the digital twin corresponding to the first time and the second time is lost.

[0053] In step S11, for cases where power sampling data is missing, the first power value corresponding to the first time point and the second power value corresponding to the second time point can be sampled based on the time period when the power sampling data is missing (hereinafter referred to as the missing window). Here, the first and second times are the times when the power sampling data is not missing. For example, the first and second times are valid sampling times when the power sampling data is not missing; there is no missing power sampling data at any time before the first time and at any time after the second time. However, the power sampling data at all sampling times between the first and second times is missing and needs to be supplemented.

[0054] The second time point is located after the first time point. For example... Figure 4 In (1), for example: the sampling period is 30 seconds, before 10:00:00 (e.g. Figure 4 (9:55:00~10:00:00) and after 10:05:00 (e.g.) Figure 4 The power sampling data from 10:05:00 to 10:10:00 is complete, but the sampling data from 10:00:00 to 10:05:00 is missing, with the missing window being (10:00:00, 10:05:00). In this case, t0 = 10:00:00 can be taken as the first time point, and t10 = 10:05:00 as the second time point. Then, the first power value P_rack(t0) corresponding to the first time point t0 and the second power value P_rack(t10) corresponding to the second time point t10 can be obtained respectively. In some implementations, the first power value can be 5KW, and the second power value can be 8KW.

[0055] S12: Based on the first power value, the second power value, and the actual power consumption corresponding to the time period to be completed, determine the target power value corresponding to at least one moment to be completed in the time period to be completed; the time period to be completed is a complete time period or a part of the time period between the first moment and the second moment.

[0056] In step S12, for the electrical equipment requiring power sampling data supplementation (hereinafter referred to as the target equipment), the actual power consumption during the supplementation period can be obtained through devices such as smart meters. For example, for the cabinet in the aforementioned application scenario example, the actual power consumption of all devices in the cabinet during the supplementation period can be directly collected through a smart meter connected to the input side of the cabinet or a power distribution unit (PDU) with high-precision metering function. The supplementation period can be the complete period between the first and second moments, or a partial period between the first and second moments. In this way, even if the power sampling data is missing for a period of time, the actual power consumption during the supplementation period can be obtained accurately and reliably, serving as the basis for power sampling data supplementation and calibration.

[0057] The time period to be completed is located between the first and second time points. For example... Figure 4 In this context, the missing window (10:00:00, 10:05:00) is designated as the time period to be completed, and t1, t2, ..., tN are designated as the times to be completed, where N is an integer greater than or equal to 1. For example... Figure 4 In the case of N=9, the corresponding times t1=10:00:30, t2=10:01:00, ..., t9=10:04:30 are taken as the times to be completed, and then the target power value corresponding to each time t1~t9 is determined. In this way, by determining the target power value of the time to be completed in the time period to be completed, the power sampling data of the time period to be completed can be completed, thereby obtaining the completed power curve as shown in (2) of 4. Of course, a part of the missing window can also be taken as the time period to be completed, such as (10:01:00, 10:04:00) as the time period to be completed. This application does not limit this.

[0058] In this step, based on the first power value, the second power value, and the power consumption corresponding to the time period to be completed, the target power value of at least one moment to be completed within the time period to be completed is determined, and a power curve corresponding to the time period to be completed is formed, such that the starting point of the power curve matches the first power value, the ending point of the power curve matches the second power value, and the area (integral) covered by the power curve within the time period to be completed is equal to the actual power consumption corresponding to the time period to be completed.

[0059] In this embodiment of the application, by referring to the actual power consumption during the power sampling data completion process through the above steps S11-S12, the interpolation result can be associated with the real energy consumption. Therefore, the completion accuracy in the case of missing data can be significantly improved, the cumulative drift error under missing data over a period of time can be suppressed, and the power curve drift can be avoided.

[0060] The steps in the above embodiments will be further described below with reference to the accompanying drawings.

[0061] Figure 5 This is a flowchart of determining a target power value provided in an embodiment of this application.

[0062] like Figure 5 As shown, in one implementation, in the above Figure 3 Based on the illustrated embodiment, step S12 may include the following steps S121-S122: S121: Determine at least one initial power value corresponding to the moment to be completed based on the first power value and the second power value.

[0063] In step S121, the initial power value corresponding to the time to be completed within the time period is first determined based on the first power value and the second power value, combined with the available information within the time period to be completed.

[0064] For example, a continuous power curve can be generated first using interpolation or fitting, and then the initial power values ​​corresponding to discrete moments to be completed can be obtained from the power curve. The moment to be completed can be selected by dividing the time period between the first and second moments into N equal time intervals, with the endpoint of each interval (excluding the first and second moments) serving as the moment to be completed. The value of N depends on the sampling period; for example, with a sampling period of 30 seconds, N=19 moments to be completed can be generated within a 10-minute time period. For instance, a straight line connecting the first and second power values ​​can be generated using linear interpolation, serving as the power curve corresponding to the time period to be completed. For example, for the aforementioned first power value P_rack(t0) = 5kW and second power value P_rack(t10) = 8kW, a straight line from (t0, 5kW) to (t10, 8kW) can be generated on the time axis. Thus, the power value corresponding to each moment to be completed on the power curve can serve as the initial power value for that moment.

[0065] For example, the initial power value corresponding to each discrete moment to be completed can be estimated first based on other related data (such as power topology relationships and the power values ​​of other devices that have topological connections with the target device), and then the discrete initial power values ​​can be connected to form a power curve.

[0066] S122: Based on the actual power consumption corresponding to the time period to be completed, calibrate the initial power value to obtain the target power value.

[0067] In step S122, the initial power value determined in step S121 is calibrated based on the actual power consumption during the time period to be completed, resulting in a calibrated target power value. This target power value ensures that the energy consumption during the time period is close to the actual power consumption. For example, the predicted power consumption during the time period to be completed can be calculated based on the first power value, the second power value, and the initial power value. Then, the predicted power consumption during the time period to be completed is compared with the actual power consumption. If the predicted power consumption is greater than the actual power consumption, the initial power value is considered too high, and therefore the initial power value can be reduced. For example, the difference between the predicted power consumption and the actual power consumption can be calculated, and then a weighted allocation method can be used to adjust the difference point by point to the initial power value at each time period to be completed. The specific adjustment method can be found in steps S1221-S1222 below. Conversely, if the predicted power consumption is less than the actual power consumption, the initial power value is considered too low, and therefore the initial power value can be increased. The adjustment method is similar to the previously described method of reducing the initial power value, and will not be repeated here.

[0068] In one implementation, the predicted power consumption can be calculated based on the initial power values ​​at each discrete moment within the time period to be completed, or it can be calculated based on the continuous power curve corresponding to the time period to be completed. Accordingly, the calibration process for the initial power value can be to adjust the discrete initial power values ​​to obtain calibrated discrete target power values; or it can be to adjust the continuous power curve as a whole, and then recalculate the target power value based on the adjusted power curve.

[0069] In this embodiment, through the above steps S121-S122, the initial power value at the time to be completed is first obtained based on the first power value and the second power value, thus effectively utilizing the fixed boundary value of the missing window; then, the initial power value obtained only based on the first and second times is calibrated using the actual power consumption during the time period to be completed as a benchmark to obtain the calibrated target power value. This effectively utilizes the actual power consumption during the time period to be completed, improving the completion accuracy in data missing scenarios.

[0070] Figure 6 This is a flowchart of determining the initial power value provided in an embodiment of this application.

[0071] like Figure 6 As shown, in one implementation, in the above Figure 5 Based on the illustrated embodiment, step S121 may include the following steps S1211-S1214: S1211: Determine the initial power value corresponding to at least one time to be completed based on the power sample value of the computing device at at least one time to be completed and the reference power value of the non-computing device.

[0072] In step S1211, the target power value that needs to be completed is the power value corresponding to the rack. The target device is the rack, which contains computing devices such as servers. The rack may also contain non-computing devices such as switches and routers.

[0073] At this point, the initial power value of the server rack at the time of completion can be determined based on the power sampling value of the computing device at the time of completion and the reference power value of the non-computing device. The power sampling value of the computing device at the time of completion can be obtained from the IT equipment management system or the out-of-band management interface of the server, and the reference power value of the non-computing device can be predetermined based on the rated power on the nameplate of the non-computing device, historical energy consumption statistics, etc.

[0074] For example, regarding the aforementioned Figure 4 The example shown can determine the power sample values ​​of the servers in the rack at times t1 to t10, resulting in the following sequence S=[S(t0), S(t1), ..., S(t10)]=[4.6, 5.0, 5.0, 4.1, 5.8, 6.0, 6.0, 6.25, 7.0, 7.0, 8.0]kW, where S(ti) in the sequence S corresponds to the power sample value at time ti; and the baseline power value of the non-server load in the rack can be estimated, such as L_rack=0.15kW. Thus, the initial power values ​​of the rack from t1 to t10 can be obtained as R = [R(t0), R(t1), ..., R(t10)] = S + L_rack = [4.75, 5.15, 5.15, 4.25, 5.95, 6.15, 6.15, 6.40, 7.15, 7.15, 8.15] kW. Here, R(ti) in the sequence R corresponds to the initial power value at time ti.

[0075] It should be noted that the embodiments of this application take a single computing device in the rack as an example. In other examples, the rack may contain multiple computing devices. In this case, the power sampling value of the computing device at the time to be completed refers to the sum of the power sampling values ​​of multiple computing devices at the time to be completed.

[0076] S1212: Determine the first power offset of the cabinet at the first moment based on the first power value, the power sampling value of the computing device at the first moment, and the reference power value of the non-computing device.

[0077] In step S1212, the first power offset of the rack at the first moment is determined based on the first power value of all devices in the rack at the first moment, the power sampling value of the computing device in the rack at the first moment, and the reference power value of the non-computing device in the rack, so as to achieve power alignment of the rack, computing device and non-computing device at the first moment.

[0078] For example, regarding the aforementioned Figure 4The example shown can be based on the power sample value 4.6 and the reference power value L of the computing device at the first moment. rack(t) =0.15kW, the theoretical power value of all devices in the rack at the first moment is calculated to be 4.6 + 0.15 = 4.75kW. Then, based on the first power value of all devices in the rack at the first moment P_rack(t0) = 5kW and the theoretical power value of 4.75kW, the first power offset of the rack at the first moment is offset(t0) = +0.25kW.

[0079] S1213: Determine the second power offset of the cabinet at the second moment based on the second power value, the power sample value of the computing device at the second moment, and the reference power value of the non-computing device.

[0080] In step S1213, the second power offset of the cabinet at the second moment is determined based on the second power values ​​of all devices in the cabinet at the second moment, the power sampling value of the computing device in the cabinet at the second moment, and the reference power value of the non-computing device in the cabinet, so as to achieve power alignment of the cabinet, computing device and non-computing device at the second moment.

[0081] For example, regarding the aforementioned Figure 4 The example shown can be based on the power sample value 8.0 and the reference power value L at the second moment of the computing device. rack(t) =0.15kW, the theoretical power value of all devices in the rack at the second moment is calculated to be 8.0 + 0.15 = 8.15kW. Then, based on the second power value P_rack(t10) = 8kW and the theoretical power value 8.15kW of all devices in the rack at the second moment, the second power offset of the rack at the second moment is offset(t10) = -0.15kW.

[0082] S1214: Adjust the initial power value based on the first power offset and the second power offset.

[0083] In step S1214, the initial power values ​​at each time point to be completed within the time period to be completed are adjusted using the first power offset and the second power offset. Specifically, the offset corresponding to each time point to be completed can be calculated using methods such as linear interpolation based on the first power offset and the second power offset, and then the initial power values ​​corresponding to each time point to be completed are adjusted based on the offsets corresponding to each time point to be completed.

[0084] For example, regarding the aforementioned Figure 4In the example shown, after determining the first power offset offset(t0) = +0.25kW and the second power offset offset(t10) = -0.15kW, a linear transition method can be used to construct a decreasing arithmetic sequence in chronological order as the power offset sequence from t1 to t9, such as offset(t1) = +0.21kW, offset(t2) = +0.17kW, ..., offset(t9) = -0.11kW. Subsequently, based on the first power offset at time t0, the power offsets at times t1 to t9, and the second power offset at time t10, the initial power value of the rack in the range t0 to t10 can be adjusted to R=[R(t0), R(t1), ..., R(t10)]=S+L_rack=[4.75, 5.15, 5.15, 4.25, 5.95, 6.15, 6.15, 6.40, 7.15, 7.15, 8.15]kW, resulting in the adjusted initial power value R=[R(t0), R(t1), ..., R(t10)]=[5.00, 5.36, 5.32, ..., 7.04, 8.00]kW.

[0085] This embodiment of the application, through the aforementioned steps S1211-S1214, firstly estimates the initial power values ​​of all devices in the rack at each time point to be completed using the real-time power sampling values ​​of the computing device and the reference power values ​​of the non-computing device. Then, using the first and second power values ​​of all devices in the rack as anchor points, it determines the first and second power offsets at the first and second times respectively, and uses a linear transition method to transfer the first and second power offsets to each time point to be completed within the time period, correcting the offset of the initial power values ​​at each time point to be completed. In this way, it can not only ensure that the initial power values ​​are accurately aligned with the actual power values ​​at the sampling points (first and second times) where no data is missing, but also transfer the offsets to the entire time period to be completed, correcting the power curve offset caused by inaccurate estimation of the reference power value or server power sampling errors, which is beneficial to improving the accuracy of power sampling data completion, thereby obtaining a sketch of a power curve with aligned endpoints and a reasonable shape.

[0086] Figure 7 This is a flowchart of an embodiment of the present application that provides a power step correction of the initial power value.

[0087] like Figure 7 As shown, in one implementation, in the above Figure 5 Based on the illustrated embodiment, steps S121 may be followed by the following steps S123-S124: S123: If a target event occurs during the time period to be completed, determine the time of occurrence of the target event and the direction of the power step caused by the target event.

[0088] In step S123, it is understood that if a specific event occurs during the time period to be completed, the total power of the electrical equipment in the power sampling data to be completed will undergo a sudden change. Therefore, the power curve will show a discontinuous change with a "jump" near the moment the specific event occurs, such as a sudden rise or fall, rather than a slow and smooth transition. Based on this, it is possible to monitor whether a target event occurs during the time period to be completed. If a target event occurs, the corresponding occurrence time and the direction of the power step caused by the target event are determined, thereby correcting the power curve for the target event.

[0089] The target event is a specific event that causes a momentary change in the total power of the electrical equipment, such as power supply switching, maintenance / installation / removal, circuit breaker tripping / closing, and uninterruptible power supply (UPS) switching. In some examples, the target event can be determined based on alarm information, as in the aforementioned cases. Figure 1 The example shown can be obtained by acquiring the alarm information corresponding to the cabinet, parsing the alarm type, alarm event and other information, and then determining whether the target event has occurred within the time period to be completed, and determining the time of occurrence of the target event and the power step direction.

[0090] The power step can be either increasing or decreasing. An increase corresponds to a sudden surge in power, such as when electrical equipment is powered on or when a backup power source is switched on; a decrease corresponds to a sudden drop in power, such as when electrical equipment is powered off or when a circuit breaker trips.

[0091] For example, regarding the aforementioned Figure 4 In the example shown, if the target event is determined to occur around t8 = 10:04:00, the time of occurrence (e.g., t8) and the direction of the power step (upward) can be determined.

[0092] S124: Adjust the initial power value based on the occurrence time and the direction of the power step.

[0093] In step S124, the initial power value in the region near the occurrence time of the target event is corrected according to the occurrence time and power step direction, so that the corrected target power value can reflect the target event and avoid unreasonable smoothing of the target event. For example, regarding the aforementioned Figure 4 In the example shown, after determining the occurrence time (e.g., t8) and the direction of the power step (increase), the target power values ​​at t8 and / or t9 can be adjusted. For example, the power value at t9 can be increased by ΔP to achieve a step-like increase in the target power value during the period from t8 to t9. In some examples, the value of ΔP can be set according to the step amplitude. For instance, if the target event causes a power increase of 2kW, the power value at t9 can be increased by 2kW from its original value.

[0094] In one implementation, the target power value range after the jump can be determined based on the step direction. For example, the target power value range corresponding to the descent direction can be determined to be [0, epsilon), where epsilon is a value close to 0. Thus, if it is determined that a target event occurs within the time period to be completed, and the power step direction caused by the target event is the descent direction, the initial power value corresponding to the time of occurrence can be adjusted to be within the target power value range corresponding to the descent direction.

[0095] In one implementation, the target power value range corresponding to the descent direction may include only 0.

[0096] In one implementation, steps S121-S122 can be executed sequentially, and steps S123-S124 can be executed after step S122. That is, the target power value is first obtained based on the actual power consumption during the time period to be filled, and then the target power value is adjusted based on the occurrence time and the power step direction.

[0097] In one implementation, the rate of change of the target power value between any two adjacent moments within the time period to be completed is less than a preset power change threshold. Specifically, to avoid steep changes in the power curve that violate electrical characteristics, the rate of change of power between adjacent moments can be constrained during the execution of the power sampling data completion method. For example, regarding the aforementioned... Figure 4 The example shown illustrates that while achieving a step-by-step increase in the target power value during the time period from t8 to t9, the rate of change of the target power value between t8 and t9, and between t9 and t10, must be less than the power change threshold. This power change threshold can be preset based on the electrical characteristics of each device within the cabinet or historical power change statistics of the cabinet.

[0098] In one implementation, steps S123-S124 above can be used to first adjust the initial power value based on the target event, and then the adjusted initial power value can be calibrated according to the actual power consumption during the period to be filled.

[0099] In this embodiment of the application, by performing the above steps S123-S124, the initial power value is stepped to correct the occurrence time of the target event and its surrounding area. This enables the completed target power value to reflect the actual situation such as power supply switching, avoiding the problem of distortion in the traditional smoothing difference at the occurrence time of the target event.

[0100] Figure 8 This is a flowchart of a calibration initial power value provided in an embodiment of this application.

[0101] like Figure 8 As shown, in one implementation, in the above Figure 5Based on the illustrated embodiment, step S122 may include the following steps S1221-S1222: S1221: Determine the predicted power consumption corresponding to the time period to be completed based on the initial power value corresponding to at least one time point to be completed within the time period to be completed.

[0102] In step S1221, the predicted power consumption corresponding to the time period to be completed can be calculated based on the first power value, the second power value, and the initial power value at the time to be completed. Then, the predicted power consumption within the time period to be completed is compared with the actual power consumption within the time period to be completed. The predicted power consumption within the time period to be completed can be calculated based on the initial power values ​​of each discrete time point to be completed within the time period. For example, the trapezoidal method can be used to approximate the predicted power consumption. The power consumption in each sampling period is the average power of the two endpoints of that sampling period multiplied by the length of the sampling period. Therefore, the two endpoints located at the boundary of the time period to be completed (such as the first time t0 and the second time t10) belong to only one sampling period and thus only participate in the calculation of the area of ​​one trapezoid, with a weighting coefficient of 0.5. The times to be completed located within the time period to be completed (such as t1~t9) belong to the two sampling periods before and after them, and thus participate in the calculation of the areas of two trapezoids, with a weighting coefficient of 1. Based on this, the specific formula for approximating the predicted power consumption using the trapezoidal method is as follows: Where T is the sampling period, as mentioned above. Figure 4 In the example shown, the sampling period T = 1 / 120 hours.

[0103] S1222: Calibrate the initial power value based on the difference between the actual power consumption and the predicted power consumption corresponding to the time period to be filled.

[0104] In step S1222, the difference E_area between the actual power consumption and the predicted power consumption E_0 during the time period to be supplemented is calculated, and then the initial power value is calibrated based on the difference E_area.

[0105] In one implementation, the initial power value can be calibrated using a global scaling approach to obtain the calibrated target power value. For example, if the difference E_area = 0.2 kWh, the initial power value corresponding to each time point to be completed in t1~t9 can be increased proportionally, so that the corrected predicted power consumption increases by 0.2 kWh, matching the actual power consumption in the time period to be completed.

[0106] In another implementation, the initial power value can be calibrated by shifting the power curve as a whole to obtain the calibrated target power value. For example, if the difference E_area = 0.2 kWh, the initial power value corresponding to each time point to be completed in t1~t9 can be increased by the same amount, so that the corrected predicted power consumption increases by 0.2 kWh, matching the actual power consumption in the time period to be completed.

[0107] In another implementation, local weighting or similar methods can be used to determine the adjustment weights corresponding to each time step to be completed, and then the initial power value can be calibrated based on the adjustment weights to obtain the calibrated target power value.

[0108] Furthermore, if the difference between the actual power consumption and the predicted power consumption during the period to be supplemented is significant, such as exceeding a preset threshold, it can be assumed that the meter data or energy consumption attribution for the period to be supplemented may be abnormal. In this case, the initial power value can be calibrated based on the difference, and a label can be added to the period to be supplemented to trigger the audit process, prompting maintenance personnel to check the meter status, power topology, etc.

[0109] In this embodiment, through the above steps S1221-S1222, the difference between the predicted power consumption and the actual power consumption during the time period to be completed is calculated based on the actual power consumption during the time period to be completed. Then, the initial power value corresponding to each time period to be completed is calibrated based on the difference by means of overall scaling, overall translation, or local weighting. This makes the calibrated predicted power consumption accurately match the actual power consumption during the time period to be completed, thereby improving the accuracy of power sampling data completion in scenarios with missing data by utilizing the energy consumption constraints during the time period to be completed.

[0110] Figure 9 This is another flowchart of the calibration initial power value provided in the embodiments of this application.

[0111] like Figure 9 As shown, in one implementation, in the above Figure 8 Based on the illustrated embodiment, step S1222 may include the following steps S12221-S12223: S12221: Determine the adjustment weight corresponding to at least one time point to be completed within the time period to be completed.

[0112] In step S12221, a local weighting method can be used to determine the adjustment weight corresponding to the time to be completed within the time period to be completed. Then, by adjusting the weight, the initial power value calibration is differentiated, so that the initial power value calibration is more in line with the actual situation.

[0113] In one implementation, the same adjustment weight can be assigned to each moment to be completed within the time period to be completed. In this way, the power adjustment amount obtained at each moment to be completed is equal, which is suitable for scenarios where the power curve has no obvious local fluctuations.

[0114] In one implementation, adjustment weights can be assigned based on the initial power value at each time point to be completed. For example, the adjustment weights corresponding to each time point to be completed can be set to be positively correlated with the initial power value at that time point, so that time points with larger initial power values ​​are assigned more adjustment amounts and time points with smaller initial power values ​​are assigned less adjustment amounts, thereby reflecting the differences in the sensitivity of power adjustment under different load levels.

[0115] In one implementation, considering the persistent impact of the target event on the power step, a higher adjustment weight can be assigned to the moments to be completed after the occurrence of the target event, avoiding unreasonable smoothing of the target event. For example, regarding the aforementioned... Figure 4 The example shown can assign a higher adjustment weight to t8~t9 than other times to be completed, so that interpolation is preferentially assigned to the time period between t8~t10.

[0116] S12222: Based on the adjustment weight, the difference is allocated to at least one time point to be completed, and the power adjustment amount corresponding to each time point to be completed is obtained.

[0117] In step S12222, according to the adjustment weight of each time period to be completed, the difference between the actual power consumption and the predicted power consumption in the time period to be completed is allocated to each time period to be completed, and combined with the sampling period, the difference allocated to each time period to be completed is converted into the corresponding power adjustment amount.

[0118] Specifically, for example, regarding the aforementioned Figure 4 In the example shown, if the difference E_area = 0.2 kWh, then based on the adjusted weights, power adjustment amounts p_1~p_9 can be allocated to the time intervals t1~t9 to be filled, such that... In this way, the initial power value can be adjusted based on the power adjustment amounts p_1~p_9 to obtain... Figure 4 The final power curve in the data.

[0119] S12223: Adjust the initial power value corresponding to each time step to be completed based on the power adjustment amount corresponding to each time step to be completed.

[0120] In step S12223, after determining the power adjustment amount corresponding to each time step to be completed, the initial power value corresponding to the time step to be completed can be adjusted based on the power adjustment amount. Specifically, the initial power value of each time step to be completed can be added to the power adjustment amount to obtain the calibrated target power value.

[0121] In this embodiment of the application, through the above steps S12221-S12223, an adjustment weight is assigned to each time point to be completed by means of local weighting. Then, based on the adjustment weight, the difference between the actual power consumption and the predicted power consumption in the time period to be completed is allocated, which can realize the fine point-by-point calibration of the initial power value and improve the completion accuracy of the power sampling data.

[0122] Figure 10 This is a flowchart of adjusting the target power value provided in an embodiment of this application.

[0123] like Figure 10 As shown, in one implementation, in the above Figure 3 Based on the illustrated embodiment, step S12 may be followed by the following step S13: S13: Adjust the target power value based on the sum of the power sample value and the reference power value.

[0124] In step S13, the completed target power value is corrected based on the power topology relationship between the target device and other devices. For the aforementioned application scenario where the target device is a server rack containing both computing and non-computing devices, considering that the rack power should be close to the sum of the power of the computing and non-computing devices, the target power value can be adjusted based on the power sampling value of the computing device at the time of completion and the reference power value of the non-computing device. For example, a deviation threshold can be preset, and for each time of completion, the sum of the power sampling value and the reference power value is calculated. If the difference between this sum and the target power value exceeds the deviation threshold, the target power value is adjusted until the difference falls back to within the deviation threshold.

[0125] In one implementation, the deviation threshold can be set based on historical experience or the accuracy requirements of the target device, such as a value between 50W and 200W. For example, the deviation threshold can be set to 50W, 100W, 150W, 200W, or other values ​​between 50W and 200W.

[0126] In one implementation, the deviation threshold can be dynamically set according to the proportion of the target power value, such as a value between 2% and 5% of the target power value. For example, the deviation threshold can be set to 2%, 3%, 4%, 5% of the target power value, or other values ​​between 2% and 5% of the target power value.

[0127] In this embodiment of the application, by using the power topology relationship between the cabinet and computing devices, the target power value of the cabinet is verified and corrected by using the power sampling data of the computing devices through the above-described step S13. This can ensure the consistency between the cabinet power and the power on the device side and improve the accuracy of the target power value.

[0128] The following is combined Figure 11This paper introduces the power sampling data completion method from a holistic perspective.

[0129] Figure 11 This is another flowchart of the power sampling data completion method provided in the embodiments of this application.

[0130] like Figure 11 As shown, this embodiment includes the following steps S21-S24: S21: Based on the first power value of the cabinet at the first moment and the second power value at the second moment, obtain the initial power curve corresponding to the cabinet.

[0131] In step S21, for the first time and the second time at both ends of the time period to be filled, the first power value of the cabinet at the first time and the second power value at the second time are obtained respectively. Then, interpolation is performed based on the first power value and the second power value to obtain an initial power curve with a reasonable shape.

[0132] If a target event, such as switching or maintenance, occurs during the time period to be completed, the step near the time of the target event can be retained during the process of determining the initial power curve.

[0133] S22: Based on the actual power consumption corresponding to the time period to be completed, calibrate the initial power curve to obtain the target power curve corresponding to the cabinet.

[0134] In step S22, the initial power curve is corrected based on the actual power consumption of all devices in the cabinet during the period to be supplemented. Specifically, the energy consumption increment of the meter during the period to be supplemented is calculated to obtain the actual power consumption during the period to be supplemented. The predicted power consumption corresponding to the initial power curve is calculated by integrating the initial power curve and obtaining the area enclosed by the initial power curve and the time axis during the period to be supplemented. It is verified whether the predicted power consumption during the period to be supplemented is equal to the actual power consumption during the period to be supplemented. If they are not equal, the initial power curve during the period to be supplemented is adjusted globally or locally to make the adjusted predicted power consumption equal to the actual power consumption.

[0135] S23: Adjust the target power curve based on the occurrence time of the target event and the direction of the power step.

[0136] In step S23, if a target event occurs during the time period to be completed, the target power curve is stepped corrected near the occurrence time based on the occurrence time of the target event and the power step direction.

[0137] S24: Correct the target power curve based on the power sample values ​​of computing devices in the rack and the reference power values ​​of non-computing devices.

[0138] In step S24, topology conservation correction is performed on the target power curve based on the cabinet-related power topology data. Specifically, firstly, time alignment is performed on each data source, and the time-series data, such as the power sampling values ​​of each device (e.g., computing devices like servers), are resampled to a unified sampling grid (e.g., 15-second or 30-second intervals). Then, time offset correction is performed on the power sampling values ​​of each device to eliminate time asynchrony issues between different acquisition systems. Simultaneously, for cabinets with different power supply paths, such as when the cabinet is simultaneously connected to a power supply and a UPS, the power supply paths from different loops (e.g., path A, path B) within the cabinet are merged based on the configuration information in the CMDB to form a total input power perspective for the cabinet.

[0139] Based on this, the power topology relationship of "cabinet - server - PSU (A / B) - PDU / Outlet - loop" is generated according to CMDB. Mismatched and missing topology relationships are repaired, computing devices (such as servers) and non-computing devices (such as switches) in the cabinet are marked, and the baseline power value of non-computing devices is determined.

[0140] Then, for each moment to be completed within the time period to be completed, the conservation residual is calculated as follows: Conservation residual = target power value - (power sample value + reference power value). A deviation threshold is set, and the moment to be completed where the conservation residual exceeds the deviation threshold is identified. The target power value of the moment to be completed is automatically corrected so that the conservation residual falls back to the range limited by the deviation threshold, thereby ensuring that the sum of the rack-level power and the equipment-side power satisfies the energy conservation relationship.

[0141] Furthermore, for a detailed explanation of steps S21-S24, please refer to the foregoing embodiments, which will not be repeated here.

[0142] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention. For example, the steps described in the above embodiments can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in the present invention can be achieved, the present invention does not limit this.

[0143] Corresponding to the embodiments of the aforementioned power sampling data completion method, this application also provides embodiments of a power sampling data completion device.

[0144] Figure 12 This is a schematic diagram of a power sampling data completion device provided in an embodiment of this application.

[0145] like Figure 12As shown, the power sampling data completion device 1200 may include a data acquisition module 1210 and a data completion module 1220.

[0146] The data acquisition module 1210 is used to acquire the first power value corresponding to the first time moment and the second power value corresponding to the second time moment; wherein the second time moment is after the first time moment, and the power sampling data of the power curve on the digital twin corresponding to the first time moment and the second time moment is lost; the data completion module 1220 is used to determine the target power value corresponding to at least one time moment in the time period to be completed based on the first power value, the second power value and the actual power consumption corresponding to the time period to be completed; wherein the time period to be completed is a complete time period or a part of the time period between the first time moment and the second time moment.

[0147] In one possible implementation, the data completion module 1220 is used to: determine an initial power value corresponding to at least one time period to be completed based on a first power value and a second power value; and calibrate the initial power value based on the actual power consumption corresponding to the time period to be completed to obtain a target power value.

[0148] In one possible implementation, the target power value is the total power value corresponding to all devices in the rack, including computing devices and non-computing devices; the data completion module 1220 is used to: determine an initial power value corresponding to at least one time to be completed based on the power sample value of the computing device at at least one time to be completed and the reference power value of the non-computing device; determine a first power offset of the rack at the first time based on the first power value, the power sample value of the computing device at the first time, and the reference power value; determine a second power offset of the rack at the second time based on the second power value, the power sample value of the computing device at the second time, and the reference power value; and adjust the initial power value based on the first power offset and the second power offset.

[0149] In one possible implementation, the data completion module 1220 is used to: determine the occurrence time of the target event and the power step direction caused by the target event when the target event occurs during the time period to be completed; wherein the power step direction is rising or falling; and adjust the initial power value based on the occurrence time and the power step direction.

[0150] In one possible implementation, the data completion module 1220 is used to: determine the predicted power consumption corresponding to the time period to be completed based on the initial power value corresponding to at least one time point to be completed within the time period to be completed; and calibrate the initial power value based on the difference between the actual power consumption and the predicted power consumption corresponding to the time period to be completed.

[0151] In one possible implementation, the data completion module 1220 is used to: determine the adjustment weight corresponding to at least one moment to be completed within the time period to be completed; based on the adjustment weight, allocate the difference to at least one moment to be completed to obtain the power adjustment amount corresponding to each moment to be completed; and adjust the initial power value corresponding to each moment to be completed according to the power adjustment amount corresponding to each moment to be completed.

[0152] In one possible implementation, the rate of change of the target power value between any two adjacent moments within the time period to be completed is less than a preset power change threshold.

[0153] In one possible implementation, the data completion module 1220 is used to: adjust the target power value based on the sum of the power sample value and the reference power value.

[0154] Figure 13 This is a schematic diagram of a computing device provided in an embodiment of this application.

[0155] like Figure 13 As shown, the computing device 1300 includes a processor 1301 and a memory 1302. Exemplarily, the computing device 1300 may also include a communications interface 1303 and a communications bus 1304.

[0156] The processor 1301, memory 1302, and communication interface 1303 communicate with each other via communication bus 1304. The communication interface 1303 may include a transmitter and receiver for communicating with other devices or communication networks, and may be a wired interface (port), such as a fiber distributed data interface (FDDI) or a gigabit Ethernet interface (GE).

[0157] In some embodiments, the processor 1301 is used to execute program 1305, which may specifically perform the relevant steps in the method embodiments described above. Specifically, program 1305 may include program code, which includes computer-executable instructions.

[0158] For example, processor 1301 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement some embodiments of this application. Computing device 1300 may include one or more processors, which may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs. The CPU may be a single-core CPU or a multi-core CPU.

[0159] In some embodiments, memory 1302 is used to store program 1305. Memory 1302 may include high-speed random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device.

[0160] Specifically, program 1305 can be called by processor 1301 to cause computing device 1300 to perform the method operations described in the above embodiments.

[0161] Some embodiments of this application provide a computer-readable storage medium storing at least one executable instruction that, when executed on a computing device 1300, causes the computing device 1300 to perform the power sampling data completion method described in the above embodiments.

[0162] Specifically, the executable instructions can be used to enable the computing device 1300 to perform a power sampling data completion method operation.

[0163] For example, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, and an optical data storage device, etc.

[0164] This application provides a chip device in some embodiments, which is applied to a server. The chip device includes one or more interface circuits and one or more processors. The interface circuits and processors are interconnected via lines. The interface circuits are used to receive signals from the server's memory and send signals to the processors, the signals including computer instructions stored in the memory. When the server processor executes the computer instructions, the server performs various steps in the power sampling data completion method shown in the above-described method embodiments.

[0165] The beneficial effects that the readable storage medium provided in some embodiments of this application can achieve can be referred to the beneficial effects in the corresponding power sampling data completion method provided above, and will not be repeated here.

[0166] It should be noted that in this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0167] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0168] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution apparatus, device, or equipment (such as a computer-based device, a processor-included device, or other device that can fetch and execute instructions from, an instruction execution apparatus, device, or equipment).

[0169] For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by or in conjunction with an instruction execution means, apparatus, or device.

[0170] More specific examples of computer-readable media (a non-exhaustive list) include the following: electrical connections having one or more wires (electronic devices), portable computer disks (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM).

[0171] Furthermore, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory. It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof.

[0172] In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution device. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc. The embodiments described above are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made based on the technical solutions of this application should be included within the scope of protection of this application.

Claims

1. A method for power sampling data completion, characterized in that, The method includes: Obtain the first power value corresponding to the first time point and the second power value corresponding to the second time point; wherein, the second time point is after the first time point, and the power sampling data of the power curve on the digital twin corresponding to the first time point and the second time point is lost; Based on the first power value, the second power value, and the actual power consumption corresponding to the time period to be filled, a target power value corresponding to at least one moment in the time period to be filled is determined; wherein, the time period to be filled is a complete time period or a part of the time period between the first moment and the second moment.

2. The method according to claim 1, characterized in that, The step of determining the target power value corresponding to at least one moment in the time period to be filled based on the first power value, the second power value, and the actual power consumption corresponding to the time period to be filled includes: Based on the first power value and the second power value, determine the initial power value corresponding to the at least one time point to be completed; Based on the actual power consumption corresponding to the time period to be filled, the initial power value is calibrated to obtain the target power value.

3. The method according to claim 2, characterized in that, The target power value is the total power value corresponding to all devices in the rack, and the devices in the rack include computing devices and non-computing devices; determining the initial power value corresponding to the at least one time point to be completed based on the first power value and the second power value includes: The initial power value corresponding to the at least one time to be completed is determined based on the power sample value of the computing device at the at least one time to be completed and the reference power value of the non-computing device; Based on the first power value, the power sampling value of the computing device at the first moment, and the reference power value, the first power offset of the cabinet at the first moment is determined; Based on the second power value, the power sample value of the computing device at the second moment, and the reference power value, the second power offset of the cabinet at the second moment is determined; The initial power value is adjusted based on the first power offset and the second power offset.

4. The method according to claim 2 or 3, characterized in that, After determining the initial power value corresponding to the at least one time point to be completed based on the first power value and the second power value, the method further includes: If a target event occurs during the time period to be completed, determine the time of occurrence of the target event and the power step direction caused by the target event; wherein the power step direction is either upward or downward. The initial power value is adjusted based on the occurrence time and the power step direction.

5. The method according to any one of claims 2 to 4, characterized in that, The step of calibrating the initial power value based on the actual power consumption corresponding to the time period to be filled includes: Based on the initial power value corresponding to at least one time point to be completed within the time period to be completed, determine the predicted power consumption corresponding to the time period to be completed; The initial power value is calibrated based on the difference between the actual power consumption and the predicted power consumption for the time period to be filled.

6. The method according to claim 5, characterized in that, The step of calibrating the initial power value based on the difference between the actual power consumption and the predicted power consumption corresponding to the time period to be filled includes: Determine the adjustment weight corresponding to at least one moment in the time period to be completed; Based on the adjustment weight, the difference is allocated to each of the at least one time moments to be completed, to obtain the power adjustment amount corresponding to each of the time moments to be completed; Adjust the initial power value corresponding to each of the time periods to be completed according to the power adjustment amount corresponding to each of the time periods to be completed.

7. The method according to any one of claims 1 to 6, characterized in that, The rate of change of the target power value between any two adjacent moments within the time period to be completed is less than a preset power change threshold.

8. The method according to claim 3, characterized in that, The method further includes: The target power value is adjusted based on the sum of the power sample value and the reference power value.

9. A power sampling data completion device, characterized in that, The device includes: The data acquisition module is used to acquire the first power value corresponding to the first time moment and the second power value corresponding to the second time moment; wherein the second time moment is after the first time moment, and the power sampling data of the power curve on the digital twin corresponding to the first time moment and the second time moment is lost; The data completion module is used to determine a target power value corresponding to at least one moment in the time period to be completed based on the first power value, the second power value, and the actual power consumption corresponding to the time period to be completed; wherein the time period to be completed is a complete time period or a part of the time period between the first moment and the second moment.

10. A computing device, characterized in that, The computing device includes a memory and a processor; the memory and the processor are coupled; the memory is used to store computer program code, the computer program code including computer instructions, which, when executed by the processor, cause the computing device to perform the power sampling data completion method as described in any one of claims 1 to 8.

Images