A data center energy efficiency monitoring method based on digital twinning

Abstract

The application discloses a kind of data center energy efficiency monitoring methods based on digital twinning, it is related to energy efficiency monitoring technical field, solve the problem that the energy efficiency of data center cannot be dynamically analyzed, method includes: the physical space data of data center and real-time multi-source dynamic data are handled;According to historical multi-source dynamic data, neural network is trained, and digital twin prediction model is obtained by training;According to the real-time multi-source dynamic data, the predicted power data of data center is obtained, and the real-time equipment temperature and real-time ambient temperature of different equipment in data center are analyzed;According to the predicted power data, the standard interval of real-time power data is analyzed, and the real-time power data of data center is comprehensively analyzed according to the standard interval;The connection mode between different equipment in data center is optimized using digital twin prediction model, and the energy efficiency of data center is dynamically analyzed.

Description

Technical Field

[0001] This invention belongs to the field of energy efficiency monitoring technology, specifically a data center energy efficiency monitoring method based on digital twins. Background Technology

[0002] Currently, with the deepening of digital transformation, the scale and computing density of data centers continue to grow, and their energy consumption has become a key bottleneck restricting the sustainable development of the industry. In order to improve energy efficiency, existing technologies have evolved from the initial manual inspection and static report management to the deployment of data center infrastructure management (DCIM) systems. These systems achieve automated collection and centralized monitoring of the operating parameters of infrastructure such as power, environment, and cabinets by deploying a large number of sensors in the data center. Operation and maintenance personnel can view key performance indicators such as power usage efficiency (PUE), temperature, and humidity in real time through the dashboard of the central control console, and manage alarms based on the static thresholds preset by the system. This has improved the visualization level and fault response speed of data center operation and maintenance to a certain extent.

[0003] However, most existing energy efficiency monitoring methods in data centers operate on a "passive response" mode, meaning that the system can only trigger an alarm after an abnormal event actually occurs and exceeds a preset threshold. This completely lacks the ability to predict the future state of the system. At the same time, there is a lack of a unified dynamic model to simulate the physical connections and causal relationships between different devices in the data center, making it impossible to extrapolate future evolution trends based on real-time data.

[0004] Therefore, this invention proposes a data center energy efficiency monitoring method based on digital twins. Summary of the Invention

[0005] The purpose of this invention is to propose a data center energy efficiency monitoring method based on digital twins, so as to solve the problem mentioned in the background art of the inability to dynamically analyze the energy efficiency of data centers.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A data center energy efficiency monitoring method based on digital twins, the specific method is as follows:

[0008] Step S1: Process the physical space data and real-time multi-source dynamic data of the data center;

[0009] Step S2: Train the neural network based on historical multi-source dynamic data to obtain a digital twin prediction model;

[0010] Step S3: Obtain the predicted power data of the data center based on real-time multi-source dynamic data, and analyze the real-time equipment temperature and real-time ambient temperature of different devices in the data center.

[0011] Step S4: Based on the predicted power data analysis, obtain the standard range of real-time power data, and perform a comprehensive analysis of the real-time power data of the data center based on the standard range.

[0012] Step S5: Optimize the connection methods between different devices in the data center using a digital twin prediction model.

[0013] Furthermore, the processing in step S1 includes the following sub-steps:

[0014] Step S11: Obtain the physical space data of the data center, and construct a three-dimensional spatial model of the data center in a three-dimensional coordinate system based on the physical space data. The coordinates of different devices and sensors in the three-dimensional spatial model are used as actual three-dimensional coordinates. The physical space data refers to the geometric dimensions and spatial layout of different devices and sensors inside the data center in the physical world.

[0015] Step S12: Collect real-time multi-source dynamic data from the data center at different time points, add timestamps to the multi-source dynamic data according to the time corresponding to different time points, and select the time period between any two time points and record it as the evaluation period.

[0016] Among them, real-time multi-source dynamic data refers to the total real-time power consumption of the data center from the power grid and the real-time individual power consumption of different devices during the evaluation period;

[0017] Step S13: Sum the real-time individual power consumption of different devices to obtain the total real-time power consumption of all devices in the data center during the evaluation period.

[0018] Furthermore, the processing in step S1 also includes the following sub-steps:

[0019] Step S14: Divide the real-time total power consumption by the real-time total power consumption to obtain the real-time power consumption ratio of the data center during the evaluation period. At the same time, divide the real-time individual power consumption by the real-time total power consumption to obtain the real-time device power consumption ratio of different devices during the evaluation period.

[0020] Step S15: Subtract the total real-time power consumption from the total real-time power consumption to obtain the real-time remaining power of the data center during the evaluation period. Multiply the real-time remaining power consumption by the real-time device power consumption ratio and add it to the real-time individual power consumption, then divide by the real-time individual power consumption to obtain the real-time power utilization efficiency of different devices during the evaluation period.

[0021] Step S16: Record the real-time power ratio, real-time device power consumption of different devices, and real-time power utilization efficiency as the real-time power data of the data center.

[0022] Furthermore, the training process in step S2 includes the following sub-steps:

[0023] Step S21: Obtain the actual three-dimensional coordinates of different sensors in the three-dimensional space model, bind the historical multi-source dynamic data collected by different sensors with the corresponding actual three-dimensional coordinates, and arrange the historical multi-source dynamic data collected by different sensors in ascending order of time to obtain the historical monitoring dataset.

[0024] Step S22: Record the historical multi-source dynamic data in the historical monitoring dataset as different monitoring data, and obtain the standard text length and theoretical data range of different monitoring data;

[0025] Step S23: Obtain the actual text length of the monitoring data, compare the actual text length with the standard text length, and compare the monitoring data with the theoretical data range;

[0026] If the actual text length of all monitoring data is equal to the standard text length, and all monitoring data are within the theoretical data range, then proceed to step S24.

[0027] If the actual text length of any monitoring data is not equal to the standard text length, or if any monitoring data does not fall within the theoretical data range, then the corresponding monitoring data will be removed.

[0028] Furthermore, the training process in step S2 also includes the following sub-steps:

[0029] Step S24: Divide the historical monitoring dataset into a training set and a test set according to a fixed ratio, input the training set into the neural network for training to obtain a digital twin prediction model, and input the test set into the digital twin prediction model to obtain prediction results for different test sets;

[0030] Step S25: Obtain the standard results of the test set, record the prediction results that are the same as the standard results as the accurate results, and count the number of accurate results in the prediction results and record it as the accurate number.

[0031] Step S26: Obtain the number of samples input into the test set of the digital twin prediction model, and calculate the actual prediction accuracy of the digital twin prediction model using the mean absolute error formula.

[0032] Step S27: Obtain the standard prediction accuracy of the digital twin prediction model and compare the actual prediction accuracy with the standard prediction accuracy;

[0033] If the actual prediction accuracy is greater than or equal to the standard prediction accuracy, no action will be taken.

[0034] If the actual prediction accuracy is less than the standard prediction accuracy, the digital twin prediction model will be repeatedly trained until the actual prediction accuracy reaches the standard prediction accuracy.

[0035] Furthermore, the analysis process in step S3 includes the following sub-steps:

[0036] Step S31: Obtain real-time multi-source dynamic data from the data center at different time points, and input the real-time multi-source dynamic data into the digital twin prediction model for prediction. The prediction process is as follows:

[0037] Acquire historical multi-source dynamic data that is the same as the real-time multi-source dynamic data, and record the real-time power data calculated under the conditions of historical multi-source dynamic data as historical power data.

[0038] The minimum value of the historical power data is obtained by iterating and comparing different historical power data, and the maximum value of the historical power data is recorded as the predicted power data; among them, the predicted power data includes the predicted power ratio of the data center, the predicted power consumption of different devices, and the predicted power utilization efficiency.

[0039] Step S32: Collect the real-time ambient temperature of the environment where the equipment is located in the data center and the real-time equipment temperature of different equipment, bind the real-time equipment temperature with the actual three-dimensional coordinates of the corresponding equipment in the three-dimensional space model, and obtain the standard equipment temperature range of different equipment.

[0040] Step S33: Add the left endpoint value and the right endpoint value of the standard equipment temperature range, sum them up, and take the average value to obtain the temperature reference value;

[0041] Step S34: Subtract the temperature reference value from the real-time device temperature and take the absolute value to obtain the real-time temperature difference between different devices. Sort the real-time device temperatures of different devices at the same time node in descending order according to the real-time temperature difference to obtain the real-time temperature array in the data center at the corresponding time node.

[0042] Step S35: Compare the real-time device temperature of different devices in the data center with the standard device temperature range;

[0043] If the real-time device temperature of any device in the data center is not within the standard device temperature range, proceed to step S36.

[0044] If the real-time device temperature of all devices in the data center is within the standard device temperature range, proceed to step S37.

[0045] Step S36: Collect devices whose real-time device temperature does not fall within the standard device temperature range and record them as abnormal devices. Analyze the matching relationship between the abnormal devices and the real-time temperature array.

[0046] Step S37: Analyze the adjustability of the real-time ambient temperature within the data center.

[0047] Furthermore, the analysis process of the matching relationship between the abnormal equipment group and the real-time temperature array in step S36 is as follows:

[0048] Step S361: Obtain the actual three-dimensional coordinates corresponding to the temperature of the first real-time device in the real-time temperature array within the data center, and obtain the actual three-dimensional coordinates corresponding to different devices in the abnormal device group and record them as the abnormal three-dimensional coordinate group.

[0049] Step S362: Compare the actual three-dimensional coordinates corresponding to the first real-time device temperature with the abnormal three-dimensional coordinate group;

[0050] If the actual three-dimensional coordinates corresponding to the temperature of the first real-time device belong to the abnormal three-dimensional coordinate group, then check the abnormal device group;

[0051] If the actual three-dimensional coordinates corresponding to the first real-time device temperature do not belong to the abnormal three-dimensional coordinate group, then the actual three-dimensional coordinates corresponding to the second real-time device temperature in the real-time temperature array are compared with the abnormal three-dimensional coordinate group in turn, until the actual three-dimensional coordinates corresponding to the real-time device temperature in the real-time temperature array belong to the abnormal three-dimensional coordinate group, and all real-time device temperatures before the corresponding real-time device temperature are recorded as abnormal device temperatures.

[0052] Step S363: Obtain the actual three-dimensional coordinates associated with the abnormal device temperature, and examine different devices within the data center based on the actual three-dimensional coordinates;

[0053] Specifically, the analysis process of the adjustability of the real-time ambient temperature within the data center in step S37 is as follows:

[0054] Step S371: Obtain the historical ambient temperature of the device when the historical device power consumption is the same as the real-time device power consumption and the historical power utilization efficiency is the same as the real-time power utilization efficiency. At the same time, obtain the historical device temperature of different devices under the historical ambient temperature.

[0055] Step S372: Compare the historical equipment temperature of different devices in the data center with the standard equipment temperature range;

[0056] If the historical device temperatures of all devices in the data center are within the standard device temperature range, then the corresponding historical ambient temperatures will be retained.

[0057] If any device in the data center has a historical device temperature that is not within the standard device temperature range, then the corresponding historical ambient temperature will be removed.

[0058] Step S373: Subtract the real-time ambient temperature from the different historical ambient temperatures to obtain the ambient temperature difference.

[0059] When all ambient temperature differences are less than or equal to zero, no operation is performed.

[0060] If any ambient temperature difference is greater than zero, proceed to step S374;

[0061] Step S374: Sum the different historical ambient temperatures and take the average value to obtain the average historical ambient temperature in the data center, and adjust the real-time ambient temperature in the data center to the average historical ambient temperature.

[0062] Further, step S4 includes the following sub-steps:

[0063] Step S41: Obtain the predicted power data and real-time power data of the data center, and subtract the corresponding predicted power data from the real-time power data to obtain the real-time deviation data of the data center at different time points.

[0064] Among them, the real-time deviation data are the real-time energy ratio deviation, the real-time equipment power consumption deviation, and the real-time energy utilization efficiency deviation.

[0065] Step S42: Calculate the real-time average value of the real-time energy ratio deviation within a fixed time window;

[0066] Step S43: Calculate the real-time standard deviation of the real-time energy ratio deviation.

[0067] Furthermore, step S4 also includes the following sub-steps:

[0068] Step S44: Multiply the confidence coefficient and the real-time standard deviation, then add the result to the real-time average to obtain the first value; add the opposite of the product of the confidence coefficient and the real-time standard deviation to the real-time average to obtain the second value.

[0069] Step S45: Construct a standard interval with the first value as the left endpoint and the second value as the right endpoint, and compare the real-time power ratio deviation of the data center with the standard interval.

[0070] If the real-time power ratio deviation of the data center is within the standard range, no action will be taken.

[0071] If the real-time power ratio of the data center is greater than the right end of the standard range, the real-time temperature array in the data center at the corresponding time point is extracted, and an alarm is issued immediately.

[0072] If the real-time power ratio of the data center is less than the left end of the standard range, the digital twin prediction model is optimized.

[0073] Furthermore, the optimization process in step S5 includes the following sub-steps:

[0074] Step S51: Obtain real-time power data of the data center, change the connection method between different devices in the three-dimensional spatial model, collect real-time multi-source dynamic data of the data center at the corresponding time, and input the real-time multi-source dynamic data into the digital twin prediction model to obtain the simulated prediction data of the data center.

[0075] Among them, the simulated prediction data includes the simulated power ratio of the data center after changing the device connection method, the simulated power consumption of different devices, and the simulated power utilization efficiency.

[0076] Step S52: If the analog power ratio of the data center is less than the real-time power ratio, the power consumption of the analog devices of different devices is less than the power consumption of the real-time devices, and the analog power utilization efficiency is less than the real-time power utilization efficiency, then the connection method of the corresponding devices is recorded as the optimization scheme.

[0077] If any of the following conditions are not met: the analog power ratio of the data center is less than the real-time power ratio, the power consumption of the analog devices of different devices is less than the power consumption of the real-time devices, or the analog power utilization efficiency is less than the real-time power utilization efficiency, then no operation will be performed.

[0078] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0079] 1. This invention first uses the physical space data of different devices in the data center to construct a three-dimensional spatial model, and binds the real-time multi-source dynamic data collected by different devices to the actual three-dimensional coordinates. At the same time, the historical multi-source dynamic data collected by different sensors are arranged in ascending order of time to obtain the historical monitoring dataset. This invention changes the problem of data isolation and lack of spatial correlation in traditional monitoring systems.

[0080] 2. This invention also utilizes real-time multi-source dynamic data to calculate real-time power data within the evaluation period, and uses historical multi-source dynamic data to train a neural network to obtain a digital twin prediction model. Then, the digital twin prediction model is used to obtain predicted power data. Next, a standard interval is constructed based on the predicted power data and the real-time power data, and the standard interval is compared with the real-time power data to analyze the real-time power data of the data center. This invention realizes dynamic analysis of real-time power data.

[0081] 3. This invention obtains different simulated electrical energy data by changing the connection method between different devices in a three-dimensional spatial model, and obtains corresponding optimization schemes based on the different simulated electrical energy data. This invention achieves safe and efficient evaluation of the effects and potential risks of different optimization schemes. Attached Figure Description

[0082] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0083] Figure 1 This is a flowchart of the method of the present invention;

[0084] Figure 2 This is a flowchart of step S2 in the present invention;

[0085] Figure 3 This is a flowchart of step S4 in the present invention;

[0086] Figure 4 This is a schematic diagram of the electronic device in this invention. Detailed Implementation

[0087] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0088] Example 1: Please refer to Figures 1-3 As shown, the technical solution provided by this invention is: a data center energy efficiency monitoring method based on digital twins. This method is applicable to constructing a three-dimensional spatial model using physical space data of different devices in the physical space of a data center, and binding actual three-dimensional coordinates to real-time multi-source dynamic data collected by different devices. Simultaneously, real-time power data within the evaluation period is calculated using the real-time multi-source dynamic data, and a digital twin prediction model is trained using historical multi-source dynamic data. The predicted power data is then obtained using the digital twin prediction model. Next, a standard interval is constructed based on the predicted power data and the real-time power data, and the standard interval is compared with the real-time power data to analyze the energy efficiency of the data center. Furthermore, different simulated power data are obtained by changing the connection method between different devices in the three-dimensional spatial model, and corresponding optimization schemes are derived based on the different simulated power data. The method includes the following sub-steps:

[0089] Step S1: Process the physical space data and real-time multi-source dynamic data of the data center;

[0090] In this embodiment, the processing in step S1 includes the following sub-steps:

[0091] Step S11: Obtain the physical space data of the data center, and construct a three-dimensional spatial model of the data center in a three-dimensional coordinate system based on the physical space data. The coordinates of different devices and sensors in the three-dimensional spatial model are used as actual three-dimensional coordinates. Specifically, the physical space data is the geometric dimensions and spatial layout of different devices and sensors inside the data center in the physical world.

[0092] In this embodiment, the physical space inside the data center can be mapped using technologies such as 3D laser scanning and photogrammetry, or a 3D spatial model identical to that in the physical world can be constructed in the digital space by importing the building information model and CAD design drawings corresponding to the data center.

[0093] Step S12: Collect real-time multi-source dynamic data from the data center at different time points, add timestamps to the multi-source dynamic data according to the time corresponding to different time points, and select the time period between any two time points and record it as the evaluation period.

[0094] Among them, real-time multi-source dynamic data refers to the total real-time power consumption of the data center from the power grid and the real-time individual power consumption of different devices during the evaluation period;

[0095] In the specific implementation process, real-time multi-source dynamic data can also be environmental and meteorological data collected by IoT sensors such as water immersion, temperature and humidity, pressure difference, wind speed, liquid level, smoke detection, airflow and outdoor weather stations in the data center. It can also be device-level and system-level power consumption data from smart distribution units (PDUs), main incoming line meters, UPS sub-meters and row-head cabinet meters. It also includes IT load data such as CPU utilization and network throughput returned by servers, switches and virtualization management platforms. In this embodiment, real-time total power and real-time individual power consumption are taken as examples.

[0096] Step S13: Sum the real-time individual power consumption of different devices to obtain the total real-time power consumption of all devices in the data center during the evaluation period.

[0097] Step S14: Divide the real-time total power consumption by the real-time total power consumption to obtain the real-time power consumption ratio of the data center during the evaluation period. At the same time, divide the real-time individual power consumption by the real-time total power consumption to obtain the real-time device power consumption ratio of different devices during the evaluation period.

[0098] Step S15: Subtract the total real-time power consumption from the total real-time power consumption to obtain the real-time remaining power of the data center during the evaluation period. Multiply the real-time remaining power consumption by the real-time device power consumption ratio and add it to the real-time individual power consumption, then divide by the real-time individual power consumption to obtain the real-time power utilization efficiency of different devices during the evaluation period.

[0099] Step S16: Record the real-time power ratio, real-time device power consumption of different devices, and real-time power utilization efficiency as the real-time power data of the data center.

[0100] Step S2: Train the neural network based on historical multi-source dynamic data to obtain a digital twin prediction model;

[0101] In this embodiment, the training process in step S2 includes the following sub-steps:

[0102] Step S21, please refer to Figure 2 As shown, the actual three-dimensional coordinates of different sensors in the three-dimensional space model are obtained, and the historical multi-source dynamic data collected by different sensors are bound to the corresponding actual three-dimensional coordinates. The historical multi-source dynamic data collected by different sensors are arranged in ascending order of time to obtain the historical monitoring dataset.

[0103] For example, if the actual three-dimensional coordinates of sensor 1 in the three-dimensional space model are (10, 10, 10), and sensor 1 collects a set of data at 10:00 on September 23, 2025, then the actual three-dimensional coordinates (10, 10, 10) are bound to the corresponding data, and the corresponding data is data with time and space attributes.

[0104] Step S22: Record the historical multi-source dynamic data in the historical monitoring dataset as different monitoring data, and obtain the standard text length and theoretical data range of different monitoring data;

[0105] Step S23: Obtain the actual text length of the monitoring data, compare the actual text length with the standard text length, and compare the monitoring data with the theoretical data range;

[0106] If the actual text length of all monitoring data is equal to the standard text length, and all monitoring data are within the theoretical data range, then proceed to step S24.

[0107] If the actual text length of any monitoring data is not equal to the standard text length, or if any monitoring data does not fall within the theoretical data range, then the corresponding monitoring data will be removed.

[0108] For example, if the monitoring data is time and the standard text length is 8 characters, for instance, the theoretical data range is [00:00:00, 23:59:59], and there are two sets of monitoring data, 25:00:01 and 01:25, then these two sets of data will be removed.

[0109] In the specific implementation process, historical monitoring datasets can also be processed in different ways. For example, the same physical quantity in different units can be converted into the same unit. It should be explained that the purpose of processing historical monitoring datasets is to improve the convergence speed and stability of model training.

[0110] Step S24: Divide the historical monitoring dataset into a training set and a test set according to a fixed ratio, input the training set into the neural network for training to obtain a digital twin prediction model, and input the test set into the digital twin prediction model to obtain prediction results for different test sets;

[0111] In the specific implementation process, the choice of neural network includes, but is not limited to, recurrent neural networks (RNN), long short-term memory networks (LSTM), or Transformer networks based on attention mechanisms;

[0112] Step S25: Obtain the standard results of the test set, record the prediction results that are the same as the standard results as the accurate results, and count the number of accurate results in the prediction results and record it as the accurate number.

[0113] It should be explained that the standard results of the test set are predetermined during the data collection phase and reflect the true state of the samples.

[0114] Step S26: Obtain the number of samples input into the test set of the digital twin prediction model, and calculate the actual prediction accuracy of the digital twin prediction model using the mean absolute error formula.

[0115] Step S27: Obtain the standard prediction accuracy of the digital twin prediction model and compare the actual prediction accuracy with the standard prediction accuracy;

[0116] If the actual prediction accuracy is greater than or equal to the standard prediction accuracy, no action will be taken.

[0117] If the actual prediction accuracy is less than the standard prediction accuracy, the digital twin prediction model will be repeatedly trained until the actual prediction accuracy reaches the standard prediction accuracy.

[0118] In this embodiment, the weights between different neurons in the neural network can be corrected and updated using the backpropagation algorithm, thereby enabling the actual prediction accuracy of the digital twin prediction model to reach the standard prediction accuracy.

[0119] Step S3: Obtain the predicted power data of the data center based on real-time multi-source dynamic data, and analyze the real-time equipment temperature and real-time ambient temperature of different devices in the data center.

[0120] In this embodiment, the analysis process in step S3 includes the following sub-steps:

[0121] Step S31: Obtain real-time multi-source dynamic data from the data center at different time points, and input the real-time multi-source dynamic data into the digital twin prediction model for prediction. The prediction process is as follows:

[0122] Acquire historical multi-source dynamic data that is the same as the real-time multi-source dynamic data, and record the real-time power data calculated under the conditions of historical multi-source dynamic data as historical power data.

[0123] The minimum value of the historical power data is obtained by iterating and comparing different historical power data, and the minimum value of the historical power data is recorded as the predicted power data.

[0124] Among them, the predicted power data includes the predicted power ratio of the data center, the predicted power consumption of different devices, and the predicted power utilization efficiency.

[0125] It should be explained that the predicted power data is the theoretical data that the data center should achieve under the current real-time multi-source dynamic data conditions, while the real-time power data represents the current actual data of the data center. In this embodiment, the smaller the real-time power data, the better.

[0126] Step S32: Collect the real-time ambient temperature of the environment where the equipment is located in the data center and the real-time equipment temperature of different equipment, bind the real-time equipment temperature with the actual three-dimensional coordinates of the corresponding equipment in the three-dimensional space model, and obtain the standard equipment temperature range of different equipment.

[0127] In this embodiment, the standard equipment temperature range can be obtained from the user manuals of different devices in the data center;

[0128] Step S33: Add the left endpoint value and the right endpoint value of the standard equipment temperature range, sum them up, and take the average value to obtain the temperature reference value;

[0129] Step S34: Subtract the temperature reference value from the real-time device temperature and take the absolute value to obtain the real-time temperature difference between different devices. Sort the real-time device temperatures of different devices at the same time node in descending order according to the real-time temperature difference to obtain the real-time temperature array in the data center at the corresponding time node.

[0130] It should be explained that each element in the real-time temperature array is bound to an actual three-dimensional coordinate;

[0131] Step S35: Compare the real-time device temperature of different devices in the data center with the standard device temperature range;

[0132] If the real-time device temperature of any device in the data center is not within the standard device temperature range, proceed to step S36.

[0133] If the real-time device temperature of all devices in the data center is within the standard device temperature range, proceed to step S37.

[0134] Step S36: Collect devices whose real-time device temperature does not fall within the standard device temperature range and record them as abnormal devices. Analyze the matching relationship between the abnormal devices and the real-time temperature array.

[0135] In this embodiment, the analysis process in step S36 includes the following sub-steps:

[0136] Step S361: Obtain the actual three-dimensional coordinates corresponding to the temperature of the first real-time device in the real-time temperature array within the data center, and obtain the actual three-dimensional coordinates corresponding to different devices in the abnormal device group and record them as the abnormal three-dimensional coordinate group.

[0137] Step S362: Compare the actual three-dimensional coordinates corresponding to the first real-time device temperature with the abnormal three-dimensional coordinate group;

[0138] If the actual three-dimensional coordinates corresponding to the temperature of the first real-time device belong to the abnormal three-dimensional coordinate group, then check the abnormal device group;

[0139] If the actual three-dimensional coordinates corresponding to the first real-time device temperature do not belong to the abnormal three-dimensional coordinate group, it means that there are still abnormal devices in addition to the abnormal device group. Then, the actual three-dimensional coordinates corresponding to the second real-time device temperature in the real-time temperature array are compared with the abnormal three-dimensional coordinate group in turn until the actual three-dimensional coordinates corresponding to the real-time device temperature in the real-time temperature array belong to the abnormal three-dimensional coordinate group. All real-time device temperatures before the corresponding real-time device temperature are recorded as abnormal device temperatures.

[0140] Step S363: Obtain the actual three-dimensional coordinates associated with the abnormal device temperature, and examine different devices within the data center based on the actual three-dimensional coordinates;

[0141] It should be explained that the real-time temperature array represents the deviation distance between the real-time device temperature of different devices and the temperature reference value. The real-time device temperature corresponding to different devices in the abnormal device group must be at the beginning of the real-time temperature array. If the real-time device temperature corresponding to the beginning of the real-time temperature array is not the real-time device temperature corresponding to the device in the abnormal device group, it means that there are other undetected abnormal temperature devices.

[0142] Step S37: Analyze the adjustability of the real-time ambient temperature within the data center;

[0143] In this embodiment, the analysis process in step S37 includes the following sub-steps:

[0144] Step S371: Obtain the historical ambient temperature of the device when the historical device power consumption is the same as the real-time device power consumption and the historical power utilization efficiency is the same as the real-time power utilization efficiency. At the same time, obtain the historical device temperature of different devices under the historical ambient temperature.

[0145] Step S372: Compare the historical equipment temperature of different devices in the data center with the standard equipment temperature range;

[0146] If the historical device temperatures of all devices in the data center are within the standard device temperature range, then the corresponding historical ambient temperatures will be retained.

[0147] If any device in the data center has a historical device temperature that is not within the standard device temperature range, then the corresponding historical ambient temperature will be removed.

[0148] Step S373: Subtract the real-time ambient temperature from the different historical ambient temperatures to obtain the ambient temperature difference.

[0149] When all ambient temperature differences are less than or equal to zero, it indicates that the real-time ambient temperature in the current data center is not adjustable, so no operation is performed.

[0150] When any ambient temperature difference value is greater than zero, it indicates that the real-time ambient temperature in the current data center is adjustable, and then proceed to step S374.

[0151] Step S374: Sum the different historical ambient temperatures and take the average value to obtain the average historical ambient temperature in the data center, and adjust the real-time ambient temperature in the data center to the average historical ambient temperature.

[0152] It should be explained that when the real-time device temperature of all devices in the data center is within the normal range, adjusting the real-time ambient temperature to the average historical ambient temperature can reduce the cooling effect of the air conditioner, thereby reducing energy consumption while ensuring that the real-time device temperature is within the normal range.

[0153] Step S4: Based on the predicted power data analysis, obtain the standard range of real-time power data, and perform a comprehensive analysis of the real-time power data of the data center based on the standard range.

[0154] In this embodiment, step S4 includes the following sub-steps:

[0155] Step S41: Obtain the predicted power data and real-time power data of the data center, and subtract the corresponding predicted power data from the real-time power data to obtain the real-time deviation data of the data center at different time points.

[0156] Among them, the real-time deviation data are the real-time energy ratio deviation, the real-time equipment power consumption deviation, and the real-time energy utilization efficiency deviation.

[0157] It should be explained that neither the real-time power data nor the predicted power data is unique. In the specific implementation process, it is necessary to subtract the predicted power ratio from the real-time power ratio, subtract the predicted power consumption from the real-time device power consumption, and subtract the predicted power utilization efficiency from the real-time power utilization efficiency to obtain the corresponding real-time deviation data. In this embodiment, the real-time deviation data is taken as the real-time power ratio deviation.

[0158] Step S42: Calculate the real-time average value PJZ of the real-time energy ratio deviation within a fixed time window W using the formula:

[0159] Where W is a fixed time window, t is a time node, n is the index of discrete moments within the fixed time window, and DNPn is the real-time power ratio deviation of the data center at different discrete moments within the fixed time window.

[0160] It should be explained that the fixed time window W is a preset value, and n is used to iterate through the W time points calculated backward from the current time t. For example, if the current time t=100s and the fixed time window W=10s, then n takes the values ​​91, 92, 93, ..., 100 in sequence. DNPn corresponds to the real-time power ratio deviation of the data center at the 91st, 92nd, 93rd, ..., 100th seconds. The real-time power ratio deviation of the data center at the 91st, 92nd, 93rd, ..., 100th seconds is added together and the average value is taken to obtain the real-time average value of the power ratio deviation within the fixed time window.

[0161] Step S43, using the formula The real-time standard deviation BZC of the real-time energy ratio deviation is calculated.

[0162] Step S44: Multiply the confidence coefficient and the real-time standard deviation, then add the result to the real-time average to obtain the first value; add the opposite of the product of the confidence coefficient and the real-time standard deviation to the real-time average to obtain the second value.

[0163] In this embodiment, the confidence coefficient is 3, representing a 99.7% confidence interval;

[0164] Step S45: Construct a standard interval with the first value as the left endpoint and the second value as the right endpoint, and compare the real-time power ratio deviation of the data center with the standard interval.

[0165] If the real-time power ratio deviation of the data center is within the standard range, no action will be taken.

[0166] If the real-time power ratio of the data center is greater than the right end of the standard range, the real-time temperature array in the data center at the corresponding time point is extracted, and an alarm is issued immediately.

[0167] If the real-time power ratio of the data center is less than the left end of the standard range, the digital twin prediction model is optimized.

[0168] In the specific implementation process, when the real-time power ratio of the data center is greater than the right end of the standard range, the real-time temperature array at the corresponding time node can be used to view different devices in the data center. When the real-time power ratio of the data center is less than the left end of the standard range, the real-time multi-source dynamic data of the data center at the current time node is input into the digital twin prediction model again to further optimize the digital twin prediction model.

[0169] In this embodiment, taking the real-time power ratio deviation as an example, the standard interval corresponding to the real-time power ratio deviation is calculated. The calculation method and comparison method of the standard interval are the same for the real-time device power consumption deviation and the real-time power utilization efficiency deviation, which will not be repeated here.

[0170] Step S5: Optimize the connection methods between different devices in the data center using a digital twin prediction model;

[0171] In this embodiment, the optimization process in step S5 includes the following sub-steps:

[0172] Step S51: Obtain real-time power data of the data center, change the connection method between different devices in the three-dimensional spatial model, collect real-time multi-source dynamic data of the data center at the corresponding time, and input the real-time multi-source dynamic data into the digital twin prediction model to obtain the simulated prediction data of the data center.

[0173] Among them, the simulated prediction data includes the simulated power ratio of the data center after changing the device connection method, the simulated power consumption of different devices, and the simulated power utilization efficiency.

[0174] In the specific implementation process, the environmental data within the three-dimensional spatial model can also be changed to obtain the simulated prediction data of the data center under the corresponding conditions;

[0175] Step S52: If the analog power ratio of the data center is less than the real-time power ratio, the power consumption of the analog devices of different devices is less than the power consumption of the real-time devices, and the analog power utilization efficiency is less than the real-time power utilization efficiency, then the connection method of the corresponding devices is recorded as the optimization scheme.

[0176] If any of the following conditions are not met: the analog power ratio of the data center is less than the real-time power ratio, the power consumption of the analog devices of different devices is less than the power consumption of the real-time devices, or the analog power utilization efficiency is less than the real-time power utilization efficiency, then no operation will be performed.

[0177] In the specific implementation process, the best optimization scheme is obtained by comparing and analyzing the simulated prediction data under different optimization schemes.

[0178] Example 2: Figure 4 As shown, this embodiment provides an electronic device that may include a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other via the communication bus. The processor can call logical instructions in the memory to execute a data center energy efficiency monitoring method based on digital twins. This method includes: processing physical space data and real-time multi-source dynamic data of the data center; training a neural network based on historical multi-source dynamic data to obtain a digital twin prediction model; obtaining predicted power data of the data center based on real-time multi-source dynamic data, and analyzing the real-time device temperature and real-time ambient temperature of different devices within the data center; obtaining a standard range of real-time power data based on the predicted power data analysis, and performing a comprehensive analysis of the real-time power data of the data center based on the standard range; and optimizing the connection methods between different devices within the data center using the digital twin prediction model.

[0179] Furthermore, when the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0180] On the other hand, this application also provides a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer can execute a data center energy efficiency monitoring method based on digital twins provided by the above methods. The method includes: processing physical space data and real-time multi-source dynamic data of the data center; training a neural network based on historical multi-source dynamic data to obtain a digital twin prediction model; obtaining predicted power data of the data center based on real-time multi-source dynamic data, and analyzing the real-time equipment temperature and real-time ambient temperature of different devices in the data center; obtaining a standard range of real-time power data based on the predicted power data analysis, and performing a comprehensive analysis of the real-time power data of the data center based on the standard range; and optimizing the connection method between different devices in the data center using the digital twin prediction model.

[0181] In another aspect, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, performs the aforementioned data center energy efficiency monitoring method based on digital twins. This method includes: processing physical space data and real-time multi-source dynamic data of the data center; training a neural network based on historical multi-source dynamic data to obtain a digital twin prediction model; obtaining predicted power data of the data center based on real-time multi-source dynamic data, and analyzing the real-time device temperature and real-time ambient temperature of different devices within the data center; analyzing the predicted power data to obtain a standard range of real-time power data, and performing a comprehensive analysis of the real-time power data of the data center based on the standard range; and optimizing the connection methods between different devices within the data center using the digital twin prediction model.

[0182] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0183] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0184] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A data center energy efficiency monitoring method based on digital twins, characterized in that, The methods include: Step S1: Process the physical space data and real-time multi-source dynamic data of the data center; Step S2: Train the neural network based on historical multi-source dynamic data to obtain a digital twin prediction model; Step S3: Obtain the predicted power data of the data center based on real-time multi-source dynamic data, and analyze the real-time equipment temperature and real-time ambient temperature of different devices in the data center. Step S4: Based on the predicted power data analysis, obtain the standard range of real-time power data, and perform a comprehensive analysis of the real-time power data of the data center based on the standard range. Step S4 includes the following sub-steps: Step S41: Obtain the predicted power data and real-time power data of the data center, and subtract the corresponding predicted power data from the real-time power data to obtain the real-time deviation data of the data center at different time points. Among them, the real-time deviation data are the real-time energy ratio deviation, the real-time equipment power consumption deviation, and the real-time energy utilization efficiency deviation. Step S42: Calculate the real-time average value of the real-time energy ratio deviation within a fixed time window; Step S43: Calculate the real-time standard deviation of the real-time energy ratio deviation; Step S44: Multiply the confidence coefficient and the real-time standard deviation, then add the result to the real-time average to obtain the first value; add the opposite of the product of the confidence coefficient and the real-time standard deviation to the real-time average to obtain the second value. Step S45: Construct a standard interval with the first value as the left endpoint and the second value as the right endpoint, and compare the real-time power ratio deviation of the data center with the standard interval. If the real-time power ratio deviation of the data center is within the standard range, no action will be taken. If the real-time power ratio of the data center is greater than the right end of the standard range, the real-time temperature array in the data center at the corresponding time point is extracted, and an alarm is issued immediately. If the real-time power ratio of the data center is less than the left end of the standard range, the digital twin prediction model is optimized. Step S5: Optimize the connection methods between different devices in the data center using a digital twin prediction model.

2. The data center energy efficiency monitoring method based on digital twin according to claim 1, characterized in that, The processing procedure in step S1 includes the following sub-steps: Step S11: Obtain the physical space data of the data center, and construct a three-dimensional spatial model of the data center in a three-dimensional coordinate system based on the physical space data. The coordinates of different devices and sensors in the three-dimensional spatial model are used as actual three-dimensional coordinates. The physical space data refers to the geometric dimensions and spatial layout of different devices and sensors inside the data center in the physical world. Step S12: Collect real-time multi-source dynamic data of the data center at different time nodes, and add timestamps to the multi-source dynamic data according to the time corresponding to different time nodes. Select the time period between any two time nodes and record it as the evaluation period. Among them, the real-time multi-source dynamic data is the real-time total power energy obtained by the data center from the power grid and the real-time individual power consumption of different devices during the evaluation period. Step S13: Sum the real-time individual power consumption of different devices to obtain the total real-time power consumption of all devices in the data center during the evaluation period.

3. The data center energy efficiency monitoring method based on digital twin according to claim 2, characterized in that, The processing in step S1 further includes the following sub-steps: Step S14: Divide the real-time total power consumption by the real-time total power consumption to obtain the real-time power consumption ratio of the data center during the evaluation period. At the same time, divide the real-time individual power consumption by the real-time total power consumption to obtain the real-time device power consumption ratio of different devices during the evaluation period. Step S15: Subtract the total real-time power consumption from the total real-time power consumption to obtain the real-time remaining power of the data center during the evaluation period. Multiply the real-time remaining power consumption by the real-time device power consumption ratio and add it to the real-time individual power consumption, then divide by the real-time individual power consumption to obtain the real-time power utilization efficiency of different devices during the evaluation period. Step S16: Record the real-time power ratio, real-time device power consumption of different devices, and real-time power utilization efficiency as the real-time power data of the data center.

4. The data center energy efficiency monitoring method based on digital twin according to claim 3, characterized in that, The training process in step S2 includes the following sub-steps: Step S21: Obtain the actual three-dimensional coordinates of different sensors in the three-dimensional space model, bind the historical multi-source dynamic data collected by different sensors with the corresponding actual three-dimensional coordinates, and arrange the historical multi-source dynamic data collected by different sensors in ascending order of time to obtain the historical monitoring dataset. Step S22: Record the historical multi-source dynamic data in the historical monitoring dataset as different monitoring data, and obtain the standard text length and theoretical data range of different monitoring data; Step S23: Obtain the actual text length of the monitoring data, compare the actual text length with the standard text length, and compare the monitoring data with the theoretical data range; If the actual text length of all monitoring data is equal to the standard text length, and all monitoring data are within the theoretical data range, then proceed to step S24. If the actual text length of any monitoring data is not equal to the standard text length, or if any monitoring data does not fall within the theoretical data range, then the corresponding monitoring data will be removed.

5. The data center energy efficiency monitoring method based on digital twin according to claim 4, characterized in that, The training process in step S2 further includes the following sub-steps: Step S24: Divide the historical monitoring dataset into a training set and a test set according to a fixed ratio, input the training set into the neural network for training to obtain a digital twin prediction model, and input the test set into the digital twin prediction model to obtain prediction results for different test sets; Step S25: Obtain the standard results of the test set, record the prediction results that are the same as the standard results as the accurate results, and count the number of accurate results in the prediction results and record it as the accurate number. Step S26: Obtain the number of samples input into the test set of the digital twin prediction model, and calculate the actual prediction accuracy of the digital twin prediction model using the mean absolute error formula. Step S27: Obtain the standard prediction accuracy of the digital twin prediction model and compare the actual prediction accuracy with the standard prediction accuracy; If the actual prediction accuracy is greater than or equal to the standard prediction accuracy, no action will be taken. If the actual prediction accuracy is less than the standard prediction accuracy, the digital twin prediction model will be repeatedly trained until the actual prediction accuracy reaches the standard prediction accuracy.

6. The data center energy efficiency monitoring method based on digital twin according to claim 5, characterized in that, The analysis process in step S3 includes the following sub-steps: Step S31: Obtain real-time multi-source dynamic data from the data center at different time points, and input the real-time multi-source dynamic data into the digital twin prediction model for prediction. The prediction process is as follows: Acquire historical multi-source dynamic data that is the same as the real-time multi-source dynamic data, and record the real-time power data calculated under the conditions of historical multi-source dynamic data as historical power data. The minimum value of the historical power data is obtained by iterating and comparing different historical power data, and the maximum value of the historical power data is recorded as the predicted power data. Among them, the predicted power data includes the predicted power ratio of the data center, the predicted power consumption of different devices, and the predicted power utilization efficiency. Step S32: Collect the real-time ambient temperature of the environment where the equipment is located in the data center and the real-time equipment temperature of different equipment, bind the real-time equipment temperature with the actual three-dimensional coordinates of the corresponding equipment in the three-dimensional space model, and obtain the standard equipment temperature range of different equipment. Step S33: Add the left endpoint value and the right endpoint value of the standard equipment temperature range, sum them up, and take the average value to obtain the temperature reference value; Step S34: Subtract the temperature reference value from the real-time device temperature and take the absolute value to obtain the real-time temperature difference between different devices. Sort the real-time device temperatures of different devices at the same time node in descending order according to the real-time temperature difference to obtain the real-time temperature array in the data center at the corresponding time node. Step S35: Compare the real-time device temperature of different devices in the data center with the standard device temperature range; If the real-time device temperature of any device in the data center is not within the standard device temperature range, proceed to step S36. If the real-time device temperature of all devices in the data center is within the standard device temperature range, proceed to step S37. Step S36: Collect devices whose real-time device temperature does not fall within the standard device temperature range and classify them as abnormal devices. Analyze the matching relationship between the abnormal devices and the real-time temperature array. Step S37: Analyze the adjustability of the real-time ambient temperature within the data center.

7. A data center energy efficiency monitoring method based on digital twins according to claim 6, characterized in that, The specific process of analyzing the matching relationship between the abnormal equipment group and the real-time temperature array in step S36 is as follows: Step S361: Obtain the actual three-dimensional coordinates corresponding to the temperature of the first real-time device in the real-time temperature array within the data center, and obtain the actual three-dimensional coordinates corresponding to different devices in the abnormal device group and record them as the abnormal three-dimensional coordinate group. Step S362: Compare the actual three-dimensional coordinates corresponding to the first real-time device temperature with the abnormal three-dimensional coordinate group; If the actual three-dimensional coordinates corresponding to the temperature of the first real-time device belong to the abnormal three-dimensional coordinate group, then check the abnormal device group; If the actual three-dimensional coordinates corresponding to the first real-time device temperature do not belong to the abnormal three-dimensional coordinate group, then the actual three-dimensional coordinates corresponding to the second real-time device temperature in the real-time temperature array are compared with the abnormal three-dimensional coordinate group in turn, until the actual three-dimensional coordinates corresponding to the real-time device temperature in the real-time temperature array belong to the abnormal three-dimensional coordinate group, and all real-time device temperatures before the corresponding real-time device temperature are recorded as abnormal device temperatures. Step S363: Obtain the actual three-dimensional coordinates associated with the abnormal device temperature, and examine different devices within the data center based on the actual three-dimensional coordinates; Specifically, the analysis process of the adjustability of the real-time ambient temperature within the data center in step S37 is as follows: Step S371: Obtain the historical ambient temperature of the device when the historical device power consumption is the same as the real-time device power consumption and the historical power utilization efficiency is the same as the real-time power utilization efficiency. At the same time, obtain the historical device temperature of different devices under the historical ambient temperature. Step S372: Compare the historical equipment temperature of different devices in the data center with the standard equipment temperature range; If the historical device temperatures of all devices in the data center are within the standard device temperature range, then the corresponding historical ambient temperatures will be retained. If any device in the data center has a historical device temperature that is not within the standard device temperature range, then the corresponding historical ambient temperature will be removed. Step S373: Subtract the real-time ambient temperature from the different historical ambient temperatures to obtain the ambient temperature difference. When all ambient temperature differences are less than or equal to zero, no operation is performed. If any ambient temperature difference is greater than zero, proceed to step S374; Step S374: Sum the different historical ambient temperatures and take the average value to obtain the average historical ambient temperature in the data center, and adjust the real-time ambient temperature in the data center to the average historical ambient temperature.

8. The data center energy efficiency monitoring method based on digital twin according to claim 1, characterized in that, The optimization process in step S5 includes the following sub-steps: Step S51: Obtain real-time power data of the data center, change the connection method between different devices in the three-dimensional spatial model, collect real-time multi-source dynamic data of the data center at the corresponding time, and input the real-time multi-source dynamic data into the digital twin prediction model to obtain the simulated prediction data of the data center. Among them, the simulated prediction data includes the simulated power ratio of the data center after changing the device connection method, the simulated power consumption of different devices, and the simulated power utilization efficiency. Step S52: If the analog power ratio of the data center is less than the real-time power ratio, the power consumption of the analog devices of different devices is less than the power consumption of the real-time devices, and the analog power utilization efficiency is less than the real-time power utilization efficiency, then the connection method of the corresponding devices is recorded as the optimization scheme. If any of the following conditions are not met: the analog power ratio of the data center is less than the real-time power ratio, the power consumption of the analog devices of different devices is less than the power consumption of the real-time devices, or the analog power utilization efficiency is less than the real-time power utilization efficiency, then no operation will be performed.

Images