Data center refrigeration system energy consumption prediction method, device, equipment and medium

Abstract

The invention belongs to the technical field of system energy consumption monitoring, and particularly discloses a data center refrigeration system energy consumption prediction method and device, equipment and a medium, and the method comprises the steps: obtaining energy consumption information of a system in a plurality of first historical time periods before a current time period; based on the energy consumption information of each first historical time period, determining dynamic energy consumption information of the system in each first historical time period and static energy consumption information of the system in the current time period; predicting the dynamic energy consumption information of the system in the current time period based on the dynamic energy consumption information of the system in the plurality of target historical time periods and the multi-source parameter information corresponding to each target historical time period; the plurality of target historical time periods are a target number of first historical time periods adjacent to the current time period in the plurality of first historical time periods; and determining the energy consumption information of the system in the current time period based on the static energy consumption information of the current time period and the dynamic energy consumption information of the current time period. According to the invention, the precision of energy consumption prediction of the refrigeration system of the data center can be effectively improved, and the prediction effect is improved.

Description

Technical Field

[0001] The present application belongs to the technical field of system energy consumption monitoring, and more specifically, to a method, device, equipment and medium for predicting energy consumption of a data center refrigeration system. Background Art

[0002] With the continuous development of technologies such as cloud computing and machine learning, the scale of enterprise data centers continues to expand. The operation of data centers requires a safe and reliable computer room environment. A large amount of heat is generated during the operation of the server. In order to ensure the continuous and reliable operation of computer servers, it is usually necessary to install a central air conditioning refrigeration system to cool the equipment. Therefore, more electricity will be consumed, and energy conservation and emission reduction has become an important goal of various companies.

[0003] In order to scientifically control the energy consumption level of data centers, it is crucial to predict the energy consumption of data center cooling systems in advance. In the existing technology, for enterprise data center cooling systems, most of them still estimate the system energy consumption by human experience, and the prediction effect is difficult to guarantee. At present, machine learning algorithms are also used to monitor system energy consumption. However, traditional machine learning energy consumption prediction methods usually only consider the relationship between time series and system energy consumption, and do not fully consider the inherent data characteristics of energy consumption data itself, resulting in low accuracy in energy consumption prediction of data center cooling systems.

[0004] Therefore, how to better predict the energy consumption of data center cooling systems has become a technical problem that needs to be solved urgently in the industry. Summary of the invention

[0005] In view of the defects of the prior art, the present application aims to solve the problem that the accuracy of energy consumption prediction of data center refrigeration systems in the prior art is not high.

[0006] To achieve the above objectives, in a first aspect, the present application provides a method for predicting energy consumption of a data center refrigeration system, comprising:

[0007] Obtaining energy consumption information of the system in a plurality of first historical periods before the current period;

[0008] Based on the energy consumption information of each of the first historical time periods, determining the dynamic energy consumption information of the system in each of the first historical time periods and the static energy consumption information in the current time period;

[0009] Based on the dynamic energy consumption information of the system in multiple target historical time periods and the multi-source parameter information corresponding to each of the target historical time periods, predict the dynamic energy consumption information of the system in the current time period; the multiple target historical time periods are the first historical time periods of the target number of the multiple first historical time periods that are adjacent to the current time period; the multi-source parameter information includes time parameter information, meteorological parameter information and operation mode information;

[0010] Based on the static energy consumption information of the current time period and the dynamic energy consumption information of the current time period, the energy consumption information of the system in the current time period is determined.

[0011] Optionally, determining the dynamic energy consumption information of the system in each of the first historical time periods and the static energy consumption information in the current time period based on the energy consumption information of each of the first historical time periods includes:

[0012] Performing cluster analysis on the energy consumption information of each of the first historical periods according to the trend of data changes within a preset time period to determine a plurality of clusters;

[0013] Decomposing the energy consumption information of each of the first historical periods in each of the clusters respectively, and determining the dynamic energy consumption information and the static energy consumption information of each of the first historical periods in each of the clusters;

[0014] According to the dynamic energy consumption information of each of the first historical periods in each of the clusters, obtaining the dynamic energy consumption information of the system in each of the first historical periods;

[0015] Based on the static energy consumption information of each of the first historical time periods in each of the clusters, the static energy consumption information of the system in the current time period is determined.

[0016] Optionally, the respectively performing data decomposition on the energy consumption information of each of the first historical periods in each of the clusters to determine the dynamic energy consumption information and the static energy consumption information of each of the first historical periods in each of the clusters includes:

[0017] Averaging the energy consumption information of the periodic time periods corresponding to the first historical time periods in each of the clusters to obtain static energy consumption information of the first historical time periods in each of the clusters;

[0018] The dynamic energy consumption information of each of the first historical periods in each of the clusters is obtained according to the energy consumption information and the static energy consumption information of each of the first historical periods in each of the clusters.

[0019] Optionally, the determining the static energy consumption information of the system in the current period based on the static energy consumption information of each of the first historical periods in each of the clusters includes:

[0020] Determine the historical time period corresponding to the current time period;

[0021] Determine, from each of the clusters, a target cluster to which the historical time period corresponding to the current time period belongs;

[0022] Based on the static energy consumption information of each of the first historical time periods in the target cluster, the static energy consumption information of the system in the current time period is determined.

[0023] Optionally, the predicting the dynamic energy consumption information of the system in the current period based on the dynamic energy consumption information of the system in multiple target historical periods and the multi-source parameter information corresponding to each of the target historical periods includes:

[0024] Inputting the dynamic energy consumption information of each of the target historical time periods and the multi-source parameter information corresponding to each of the target historical time periods into a dynamic energy consumption prediction model to obtain the dynamic energy consumption information of the system in the current time period;

[0025] The dynamic energy consumption prediction model is trained based on dynamic energy consumption information samples of multiple second historical time periods, and corresponding multi-source parameter information and dynamic energy consumption information labels; the dynamic energy consumption information samples of multiple second historical time periods are determined based on the original dynamic energy consumption information sample data set.

[0026] Optionally, before inputting the dynamic energy consumption information of each of the target historical time periods and the multi-source parameter information corresponding to each of the target historical time periods into the dynamic energy consumption prediction model to obtain the dynamic energy consumption information of the system in the current time period, the method further includes:

[0027] Performing data enhancement on the original dynamic energy consumption information sample data set to obtain dynamic energy consumption information samples for each of the second historical periods;

[0028] Taking each dynamic energy consumption information sample of the second historical period, and the corresponding multi-source parameter information and dynamic energy consumption information label as a group of training samples, and obtaining multiple groups of the training samples;

[0029] For any one group of training samples among the multiple groups of training samples, inputting the training samples into a dynamic energy consumption prediction model, and outputting prediction information corresponding to the training samples;

[0030] Using a preset loss function, based on the prediction information corresponding to the training sample and the dynamic energy consumption information label corresponding to the training sample, a loss value is calculated, and based on the loss value, the model parameters of the dynamic energy consumption prediction model are adjusted until the loss value is less than a preset threshold or the number of training times reaches a preset number of times;

[0031] The model parameters obtained when the loss value is less than the preset threshold or the number of training times reaches the preset number of times are used as the model parameters of the trained dynamic energy consumption prediction model.

[0032] Optionally, performing data enhancement on the original dynamic energy consumption information sample data set to obtain the dynamic energy consumption information sample of each second historical period includes:

[0033] Performing variational inference analysis on each original dynamic energy consumption information sample in the original dynamic energy consumption information sample data set to generate a new sample corresponding to each original dynamic energy consumption information sample to form a new sample data set;

[0034] According to the meteorological parameter information of the system in the current time period, configuring the target data ratio of the original dynamic energy consumption information sample data set and the new sample data set;

[0035] Based on the original dynamic energy consumption information sample data set and the new sample data set at the target data ratio, a dynamic energy consumption information sample for each second historical period is generated.

[0036] In a second aspect, the present application provides a data center refrigeration system energy consumption prediction device, comprising:

[0037] An acquisition module, used to acquire energy consumption information of the system in a plurality of first historical time periods before the current time period;

[0038] A first processing module, configured to determine the dynamic energy consumption information of the system in each of the first historical time periods and the static energy consumption information in the current time period based on the energy consumption information of each of the first historical time periods;

[0039] A prediction module, configured to predict the dynamic energy consumption information of the system in the current period based on the dynamic energy consumption information of the system in multiple target historical periods and the multi-source parameter information corresponding to each of the target historical periods; the multiple target historical periods are the target number of historical periods adjacent to the current period in the multiple first historical periods; the multi-source parameter information includes time parameter information, meteorological parameter information and operation mode information;

[0040] The second processing module is used to determine the energy consumption information of the system in the current time period based on the static energy consumption information of the current time period and the dynamic energy consumption information of the current time period.

[0041] In a third aspect, the present application provides an electronic device / image signal generator / network device / transmitter / terminal / base station / industrial computer, comprising: at least one memory for storing programs; at least one processor for executing programs stored in the memory, when the program stored in the memory is executed, the processor is used to execute the method described in the first aspect or any possible implementation of the first aspect.

[0042] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method described in the first aspect or any possible implementation of the first aspect.

[0043] In a fifth aspect, the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method described in the first aspect or any possible implementation of the first aspect.

[0044] In general, the above technical solutions conceived by the present application have at least the following beneficial effects compared with the prior art:

[0045] The present application provides a method, device, equipment and medium for predicting energy consumption of a data center refrigeration system. By making full use of the characteristics of periodic changes and dynamic random changes in energy consumption data in historical data, energy consumption information of the system in multiple historical periods before the current period is obtained, data is decomposed based on the energy consumption information of each historical period, and the dynamic energy consumption information of the system in each historical period and the static energy consumption information in the current period are determined, thereby extracting the dynamic part with large time series correlation and the static part with small time series correlation. The dynamic energy consumption information of the system in multiple target historical periods and the corresponding multi-source parameter information are used to predict the dynamic energy consumption information of the system in the current period, and then the energy consumption information of the system in the current period is obtained according to the static energy consumption information of the current period and the predicted dynamic energy consumption information of the current period, which can effectively improve the accuracy of energy consumption prediction of the data center refrigeration system and improve the energy consumption prediction effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flow chart of a method for predicting energy consumption of a data center refrigeration system provided in an embodiment of the present application;

[0047] Figure 2 It is a schematic diagram of static energy consumption information in the energy consumption prediction method for a data center refrigeration system provided in an embodiment of the present application;

[0048] Figure 3 It is a schematic diagram of dynamic energy consumption information in the energy consumption prediction method for a data center refrigeration system provided in an embodiment of the present application;

[0049] Figure 4 It is a flow chart of model training in the method for predicting energy consumption of a data center refrigeration system provided in an embodiment of the present application;

[0050] Figure 5 It is a structural schematic diagram of a data center refrigeration system energy consumption prediction device provided in an embodiment of the present application;

[0051] Figure 6It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0053] The terms "first" and "second" in the specification and claims herein are used to distinguish different objects rather than to describe a specific order of objects. For example, the first historical period and the second historical period are used to distinguish different historical periods rather than to describe a specific order of historical periods.

[0054] In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0055] In the description of the embodiments of the present application, unless otherwise specified, "multiple" means two or more than two. For example, multiple target historical periods refer to two or more target historical periods, etc.

[0056] Next, the technical solutions provided in the embodiments of the present application are introduced.

[0057] Figure 1 is a flow chart of a method for predicting energy consumption of a data center refrigeration system provided in an embodiment of the present application, such as Figure 1 As shown, the method includes: step 110, step 120, step 130 and step 140.

[0058] Wherein, step 110, obtaining energy consumption information of the system in a plurality of first historical periods before the current period;

[0059] Step 120, based on the energy consumption information of each first historical period, determining the dynamic energy consumption information of the system in each first historical period and the static energy consumption information in the current period;

[0060] Step 130, based on the dynamic energy consumption information of the system in multiple target historical periods and the multi-source parameter information corresponding to each target historical period, predict the dynamic energy consumption information of the system in the current period; the multiple target historical periods are the first historical periods of the target number of adjacent to the current period in multiple first historical periods; the multi-source parameter information includes time parameter information, meteorological parameter information and operation mode information;

[0061] Step 140: Determine the energy consumption information of the system in the current period based on the static energy consumption information of the current period and the dynamic energy consumption information of the current period.

[0062] Specifically, the time period described in the embodiment of the present application refers to the minimum time unit for predicting system energy consumption, which can be in units of hours or days, etc.

[0063] The first historical period described in the embodiment of the present application refers to the historical period before the current period of the system. In the embodiment of the present application, the number of first historical periods can be selected according to actual forecasting requirements, such as multiple first historical periods can be each period of the two weeks before the current period, or each period of the three weeks before the current period, etc.

[0064] It is understandable that the operating energy consumption of the data center cooling system will be affected by factors such as changes in meteorological parameters, switching of operating modes, and changes in user loads. Changes in meteorological conditions, switching of operating modes, and fluctuations in user loads will all cause energy consumption to show certain daily, seasonal, and annual dynamic periodic changes, and this change also has a certain degree of randomness. Of course, a portion of fixed sticky user loads exhibits relatively constant static energy consumption characteristics.

[0065] It should be noted that, in the embodiment of the present application, the energy consumption information includes dynamic energy consumption information and static energy consumption information; the specific indicator of the energy consumption information can be the total active power of the system.

[0066] The dynamic energy consumption information described in the embodiments of the present application refers to the fluctuating energy consumption part of the system energy consumption that has dynamic and random changes. The dynamic energy consumption information is closely related to the actual operating conditions and activities of the data center refrigeration system, resulting in irregular energy consumption changes, such as fluctuations in the computing load of the server, refrigeration demand caused by fluctuations in external ambient temperature, and the impact of personnel activities, which lead to fluctuating energy consumption in the data center refrigeration system.

[0067] In the embodiments of the present application, the dynamic energy consumption information can be regarded as energy consumption composed of outdoor weather fluctuations, fluctuating user loads with a certain degree of randomness, and the like.

[0068] The static energy consumption information described in the embodiments of the present application refers to the energy consumption part of the system energy consumption that has relatively fixed periodic changes. In the data center refrigeration system, the static energy consumption information is usually related to the basic operation and periodic changes of the data center refrigeration system. For example, to maintain the basic operation and safety of the equipment, this part of the energy consumption is relatively fixed and will not change significantly due to factors such as meteorological parameters and personnel. In addition, the cooling demand of the data center shows periodic changes, especially in different time periods of the day or between working days and non-working days. There will be obvious energy consumption trends (high energy consumption during the day, low energy consumption at night, etc.), which reflects the trend and pattern of the energy consumption of the data center refrigeration system over a long period of time.

[0069] In the embodiments of the present application, the static energy consumption information can be regarded as the basic energy consumption part with periodic changes, which can specifically include two parts: one is the minimum lower limit energy consumption of the data center refrigeration system, including the basic energy consumption required for the basic operation of data center equipment and infrastructure; the other is the periodic changes in energy consumption caused by periodic changes in weather parameters and periodic changes in personnel activity patterns.

[0070] The target historical period described in the embodiment of the present application refers to a historical period adjacent to the current period among multiple first historical periods.

[0071] The target quantity described in the embodiment of the present application refers to the number of target historical time periods. When the time period is in hours, the target quantity can be 24, 36, etc. In other words, the dynamic energy consumption information of the first 24 hours of the current time period and the multi-source parameter information corresponding to each hour can be used to predict the dynamic energy consumption information of the system in the current hour.

[0072] The multi-source parameter information described in the embodiment of the present application includes time parameter information, meteorological parameter information and operation mode information, wherein the time parameter information may specifically include month, week, and hour; the meteorological parameter information may specifically include outdoor dry-bulb temperature, outdoor relative humidity, wind speed, and solar radiation intensity; and the operation mode information includes the cooling mode of the system.

[0073] In an embodiment of the present application, in step 110, energy consumption information of the system in multiple first historical time periods before the current time period may be acquired through data monitoring.

[0074] In an embodiment of the present application, in step 120, by making full use of the periodic changes and dynamic random changes of energy consumption data in historical data, data decomposition is performed based on the energy consumption information of each first historical time period, and the dynamic part of the energy consumption data with large time series correlation and the static part of the data with small time series correlation are extracted, thereby determining the dynamic energy consumption information of the system in each first historical time period and the static energy consumption information in the current time period.

[0075] Figure 2 is a schematic diagram of static energy consumption information in the data center refrigeration system energy consumption prediction method provided in an embodiment of the present application, such as Figure 2 As shown, from the 1st hour to the 641st hour of the observation period, the static energy consumption information shows periodic changes. This part of the energy consumption is relatively fixed and will not change significantly due to factors such as meteorological parameters and human activities.

[0076] Figure 3 is a schematic diagram of dynamic energy consumption information in the data center refrigeration system energy consumption prediction method provided in an embodiment of the present application, such as Figure 3 As shown, from the 1st hour to the 641st hour of the observation period, the dynamic energy consumption information presents irregular changes, and this part of the energy consumption has large volatility.

[0077] Furthermore, in an embodiment of the present application, in step 130, the dynamic energy consumption information of the target historical period of the target data is obtained from the dynamic energy consumption information of each first historical period, and the multi-source parameter information corresponding to each target historical period is obtained, including time parameter information, meteorological parameter information and operation mode information, and then a linear regression algorithm or a trained deep neural network model is used to predict the dynamic energy consumption information of the system in the current period based on the dynamic energy consumption information of the system in each target historical period and the multi-source parameter information corresponding to each target historical period.

[0078] Furthermore, in an embodiment of the present application, in step 140, data is added based on the static energy consumption information of the current time period and the dynamic energy consumption information of the current time period, so that the energy consumption information of the system in the current time period can be obtained.

[0079] The energy consumption prediction method for a data center refrigeration system in an embodiment of the present application fully utilizes the characteristics of periodic changes and dynamic random changes in energy consumption data in historical data, obtains energy consumption information of the system in multiple historical periods before the current period, performs data decomposition based on the energy consumption information of each historical period, determines the dynamic energy consumption information of the system in each historical period and the static energy consumption information in the current period, thereby extracting the dynamic part with large time series correlation and the static part with small time series correlation, thereby using the dynamic energy consumption information of the system in multiple target historical periods and the corresponding multi-source parameter information to predict the dynamic energy consumption information of the system in the current period, and then obtains the energy consumption information of the system in the current period based on the static energy consumption information of the current period and the predicted dynamic energy consumption information of the current period, which can effectively improve the accuracy of energy consumption prediction of the data center refrigeration system and improve the energy consumption prediction effect.

[0080] Based on the content of the above embodiment, as an optional embodiment, based on the energy consumption information of each first historical period, determining the dynamic energy consumption information of the system in each first historical period and the static energy consumption information in the current period includes:

[0081] Performing cluster analysis on the energy consumption information of each first historical period according to the trend of data changes within a preset time period to determine a plurality of clusters;

[0082] Decomposing the energy consumption information of each first historical period in each cluster respectively, and determining the dynamic energy consumption information and the static energy consumption information of each first historical period in each cluster;

[0083] According to the dynamic energy consumption information of each first historical period in each cluster, the dynamic energy consumption information of the system in each first historical period is obtained;

[0084] Based on the static energy consumption information of each first historical period in each cluster, the static energy consumption information of the system in the current period is determined.

[0085] Specifically, the preset time period described in the embodiment of the present application may be a time period in days or a time period in weeks; it may be specifically determined based on the time span of the energy consumption data collected during the historical period.

[0086] Optionally, in the case of missing energy consumption data, cluster analysis can be performed on a daily basis and judged according to the day of the week. According to the trend of data changes, since the overall energy consumption level on weekdays is higher than that on weekends, it can be divided into two clusters according to the weekdays or weekends, one cluster for Monday to Friday and another cluster for weekends.

[0087] Optionally, when the amount of energy consumption data is sufficient, for example, a full year of energy consumption data is available, it can be analyzed on a weekly basis and subdivided into three clusters, namely, an energy consumption pattern with high energy consumption during the week and low energy consumption during the weekend, an energy consumption pattern with low energy consumption during the week and high energy consumption during the weekend, and an energy consumption pattern with little difference in energy consumption during the week.

[0088] Furthermore, in an embodiment of the present application, according to the time span of the energy consumption information of each first historical period, cluster analysis is performed on the energy consumption information of each first historical period according to the trend of data changes within a preset time period, and multiple clusters can be determined, so that the energy consumption information of each first historical period can be divided into each cluster. For example, when the preset time period adopts a time period in days, according to the day of the week to which each first historical period belongs, the energy consumption information of all first historical periods is clustered into two clusters, that is, one cluster from Monday to Friday and another cluster on weekends. Each cluster represents a specific operating condition or time distribution.

[0089] In an embodiment of the present application, for each cluster, methods such as mean and variance, Fourier decomposition, wavelet decomposition or empirical mode decomposition (EMD) can be used to perform data decomposition on the energy consumption information of each first historical period in each cluster, so as to identify the dynamic part and the static part of the daily energy consumption data, and then determine the dynamic energy consumption information and the static energy consumption information of each first historical period in each cluster.

[0090] It is easy to understand that, in the embodiment of the present application, the dynamic energy consumption information of the system in each first historical period can be directly obtained based on the dynamic energy consumption information of each first historical period in each cluster.

[0091] Furthermore, in an embodiment of the present application, based on the static energy consumption information of each first historical period in each cluster, the static energy consumption information corresponding to each cluster can be determined. Thus, the static energy consumption information of the system in the current period can be finally determined based on the cluster to which the current period belongs.

[0092] The method of the embodiment of the present application performs cluster analysis and data decomposition on the energy consumption information of each historical time period, extracts dynamic energy consumption information with a high correlation with the time series and static energy consumption information with a low correlation with the time series, thereby achieving effective decomposition of the energy consumption information of each time period of the system, improving the accuracy of the dynamic and static information and static energy consumption information obtained by decomposition, and providing reliable technical data for subsequent energy consumption prediction.

[0093] Based on the content of the above embodiment, as an optional embodiment, data decomposition is performed on the energy consumption information of each first historical period in each cluster to determine the dynamic energy consumption information and the static energy consumption information of each first historical period in each cluster, including:

[0094] Averaging the energy consumption information of the corresponding periodic time periods in each first historical time period in each cluster to obtain the static energy consumption information of each first historical time period in each cluster;

[0095] According to the energy consumption information and static energy consumption information of each first historical period in each cluster, the dynamic energy consumption information of each first historical period in each cluster is obtained.

[0096] Specifically, in the embodiment of the present application, the mean value method may be used to solve the static energy consumption information of each first historical period in each cluster.

[0097] The specific calculation process is as follows:

[0098] E=E 周期性 +E 波动性 ;

[0099]

[0100] In the formula, E represents the overall energy consumption; E 周期性 Indicates the periodic energy consumption, that is, the static energy consumption information; E 波动性 It represents fluctuating energy consumption, that is, dynamic energy consumption information; n represents the number of time periods corresponding to the energy consumption.

[0101] In the embodiment of the present application, the energy consumption information of the corresponding periodic time periods in each first historical time period in each cluster is averaged to obtain the static energy consumption information of each first historical time period in each cluster. For example, in the case of dividing the clusters into two types according to the middle of the week or the weekend, the energy consumption information of the corresponding periodic time periods in each first historical time period in each cluster is averaged, that is, for the same cluster, the energy consumption information of the corresponding periodic time periods in each time period, such as the energy consumption information of the same time period in Monday, is averaged respectively, and the obtained average value is the static energy consumption information corresponding to the same time period in Monday.

[0102] For another example, the energy consumption information belonging to the same period on Saturday is averaged, and the average value obtained is the static energy consumption information corresponding to the same period on Saturday. Thus, the static energy consumption information of each first historical period in each cluster can be obtained.

[0103] It can be understood that the static energy consumption information of each first historical period in the same cluster is the same, and is data obtained by averaging the energy consumption information of each historical period in the cluster.

[0104] Furthermore, in an embodiment of the present application, by subtracting the energy consumption information of each first historical period in each cluster from the corresponding static energy consumption information, the dynamic energy consumption information of each first historical period in each cluster can be obtained.

[0105] The method of the embodiment of the present application fully utilizes the changing trend of the system energy consumption data in different time periods, clusters the energy consumption data of the system historical period, and averages the energy consumption information of the corresponding periodic period in each cluster to extract the static energy consumption information of each periodic period, which can effectively improve the accuracy and reliability of the extraction of static energy consumption information.

[0106] Based on the content of the above embodiment, as an optional embodiment, based on the static energy consumption information of each first historical period in each cluster, determining the static energy consumption information of the system in the current period includes:

[0107] Determine the historical time period corresponding to the current period;

[0108] Determine the target cluster to which the historical time period corresponding to the current time period belongs from each cluster;

[0109] Based on the static energy consumption information of each first historical period in the target cluster, the static energy consumption information of the system in the current period is determined.

[0110] Specifically, the target cluster described in the embodiment of the present application refers to the cluster whose historical time periods in each cluster belong to the historical time period corresponding to the current time period. For example, if the current time period and the historical time period in a cluster belong to the same Monday or the same hour in Monday, then the cluster can be determined to be the target cluster.

[0111] In the embodiment of the present application, after obtaining the static energy consumption information of each first historical period in each cluster, the historical time period corresponding to the current period is first determined. For example, in the case of dividing the clusters into two types according to the middle of the week or the weekend, if the current period is judged to be Monday or a certain hour in Monday, the historical time period corresponding to the current period can be determined to be the historical Monday or the hour in Monday.

[0112] Furthermore, in the embodiment of the present application, the target cluster to which the historical time period corresponding to the current time period belongs is determined from each cluster, that is, the cluster corresponding to Monday can be determined from each cluster based on the current time period belonging to Monday. Furthermore, the static energy consumption information of each first historical time period in the target cluster, that is, the static energy consumption information obtained by averaging calculation, can be used to determine the static energy consumption information of the system in the current time period.

[0113] The method of the embodiment of the present application makes full use of the periodic law of the static part data of the system energy consumption, determines the target cluster to which the historical time period corresponding to the current time period belongs, and determines the static energy consumption information of the current time period based on the static energy consumption information of the corresponding target cluster, thereby ensuring the data reliability of the static energy consumption information of the current time period.

[0114] In a specific embodiment of the present application, an energy consumption prediction model is trained using the dynamic energy consumption information of four weeks in a specific monitoring period of a certain year to predict the dynamic energy consumption information of the fifth week. The predicted dynamic energy consumption information can then be added to the static energy consumption information calculated in the previous four weeks to obtain the predicted energy consumption information of the system in the fifth week.

[0115] Furthermore, when the data center is running for the sixth week, the above steps are repeated using the data from the first five weeks to calculate the corresponding static energy consumption information and dynamic energy consumption information. The correlation coefficient R is calculated between the static energy consumption information calculated using the data from the first five weeks and the static energy consumption information from the first four weeks. 2 , as shown in the following formula:

[0116]

[0117] In the formula, SS 4 Represents the residual sum of squares containing four weeks of static energy consumption information, SS5 Represents the residual sum of squares containing five weeks of static energy consumption information.

[0118] After calculation, the R 2 is 0.9994, satisfying R 2 If the requirement is greater than 0.9, the newly calculated static energy consumption information is considered correct. Repeat the above prediction process to obtain the predicted energy consumption for the sixth week. Similarly, the next prediction is completed using the same method as above.

[0119] Based on the content of the above embodiment, as an optional embodiment, based on the dynamic energy consumption information of the system in multiple target historical time periods and the multi-source parameter information corresponding to each target historical time period, the dynamic energy consumption information of the system in the current time period is predicted, including:

[0120] The dynamic energy consumption information of each target historical period and the multi-source parameter information corresponding to each target historical period are input into the dynamic energy consumption prediction model to obtain the dynamic energy consumption information of the system in the current period;

[0121] The dynamic energy consumption prediction model is trained based on dynamic energy consumption information samples of multiple second historical periods, as well as corresponding multi-source parameter information and dynamic energy consumption information labels; the dynamic energy consumption information samples of multiple second historical periods are determined based on the original dynamic energy consumption information sample data set.

[0122] Specifically, the second historical period described in the embodiment of the present application refers to the historical period in which the system generates energy consumption data. In the embodiment of the present application, the number of second historical periods can be selected according to the actual forecasting requirements, such as multiple second historical periods can be each period of the system in the previous month, or each period of the previous year, etc. It can be understood that multiple first historical periods can be included in multiple second historical periods, and the two can also be equal.

[0123] The original dynamic energy consumption information sample data set described in the embodiment of the present application refers to a sample data set composed of the original dynamic energy consumption information data of each second historical period, wherein the original dynamic energy consumption information data refers to the dynamic energy consumption information data before data enhancement or expansion processing.

[0124] It should be noted that the original dynamic energy consumption information of each second historical period can be obtained by performing data analysis on the original energy consumption information of each second historical period in accordance with the aforementioned data clustering and decomposition method.

[0125] The dynamic energy consumption prediction model described in the embodiment of the present application is obtained by training based on dynamic energy consumption information samples of multiple second historical time periods, as well as corresponding multi-source parameter information and dynamic energy consumption information labels. It is used to learn the spatial characteristics and temporal characteristics of the system dynamic energy consumption information, and outputs high-quality prediction results by extracting and identifying features of the dynamic energy consumption information of multiple target historical time periods and the corresponding multi-source parameter information.

[0126] It should be noted that in the embodiment of the present application, the dynamic energy consumption prediction model can be constructed based on a deep neural network. Among them, the deep neural network can specifically adopt a long short-term memory (Long Short-Term Memory, LSTM) network model, or a joint model of a convolutional neural network model (Convolutional Neural Network, CNN) and an LSTM network model, or other deep neural networks for future energy consumption data prediction, which are not specifically limited in this application.

[0127] In an embodiment of the present application, the model training samples are composed of multiple groups of dynamic energy consumption information samples carrying corresponding multi-source parameter information and dynamic energy consumption information labels.

[0128] In the embodiment of the present application, the dynamic energy consumption information label is predetermined according to the dynamic energy consumption information sample and corresponds to the dynamic energy consumption information sample one by one. That is, each dynamic energy consumption information sample in the training sample is predetermined to carry a corresponding dynamic energy consumption information label.

[0129] Furthermore, in the embodiment of the present application, by performing data enhancement and expansion on the original dynamic energy consumption information sample data set, a large number of dynamic energy consumption information samples of historical periods are obtained, and these sample data are used to train and evaluate the dynamic energy consumption prediction model until a trained dynamic energy consumption prediction model is obtained. Then, the dynamic energy consumption information of each target historical period and the multi-source parameter information corresponding to each target historical period are input into the trained dynamic energy consumption prediction model, and the dynamic energy consumption information of the system in the current period can be predicted after the dynamic energy consumption prediction model performs data feature extraction and recognition.

[0130] In a specific embodiment of the present application, CNN and LSTM models are used as dynamic energy consumption prediction models to process spatial features and temporal features in input data respectively. The structure of the CNN model specifically includes two layers of convolutional neural networks for processing spatial features in the data, and then the data is processed by the LSTM model to process the temporal dependencies, and finally the prediction results are output through the fully connected layer.

[0131] Among them, the CNN model uses convolutional layers to extract and reduce spatial feature dimensions, such as distribution characteristics of outdoor dry-bulb temperature and relative humidity; while the LSTM model captures long-term dependencies in time series data through its gating mechanism, effectively predicting the energy consumption of the data center cooling system in the future.

[0132] In the prediction process, a sliding window model can be used to further enhance the extraction of time series data. In the embodiment of the present application, 25 grids are used as the window size, that is, the energy consumption data of the 25th hour is predicted using the previous 24 hours. The specific structure of the prediction model is as follows:

[0133] The first layer is the first convolution layer of the CNN model, and the convolution kernel size is 3. The process is expressed as follows:

[0134] (x*h)[n]=x[n]·h[0]+x[n-1]·h[1]+x[n-2]·h[2];

[0135] The second layer is the second convolution layer of the CNN model, and the convolution kernel size is 4. The process is expressed as follows:

[0136]

[0137] Among them, x[n] represents the sequence of dynamic energy consumption information of the input target historical period, h[n] represents the sequence of convolution kernels, and n represents the index of the output sequence.

[0138] The third layer is the LSTM layer. The process is expressed as follows:

[0139] Forget gate: f t =σ(W f ·x t +U f ·h t-1 +b f );

[0140] Input gate: i t =σ(W i ·x t +U i ·h t-1 +b i );

[0141] Output gate: o t =σ(W o ·x t +U o ·h t-1 +b o );

[0142] Where t represents the time step, x t Indicates the input value at the current moment, h t-1represents the hidden state at the previous moment, σ represents the sigmoid activation function, and W, U, and b represent weight coefficients.

[0143] The fourth layer is a fully connected layer. As the output, the process is expressed as follows:

[0144] y=W·x+b;

[0145] Among them, W represents the weight matrix, b represents the bias vector, x represents the input vector, and y represents the output vector.

[0146] Using CNN to process energy consumption data of data center cooling systems has the following advantages:

[0147] 1) CNN model effectively extracts local correlation features: The energy consumption data of the data center cooling system is a typical time series data. The convolution kernel in the CNN model can effectively capture the local spatial correlation in the data. Through the convolution operation, the LSTM model can identify and extract local features in the input integrated data, thereby better predicting the system energy consumption.

[0148] 2) Dimensionality reduction and feature integration: Convolutional layers can reduce the spatial dimensions of the input data by striding, pooling, or simply reducing the size of the convolution kernel, thereby reducing the complexity of the model.

[0149] 3) Improve nonlinear features: There are various complex nonlinear relationships in the energy consumption data of the data center cooling system, such as the nonlinear effect of temperature change on energy consumption. Usually, variables can only reflect the trend of energy consumption changes, but cannot specifically reflect the numerical value of energy consumption changes. Therefore, the ReLu activation function can be applied to reduce the dimension of the data and extract high-level features from the data.

[0150] The method of the embodiment of the present application fully extracts the spatial characteristics and temporal characteristics of the dynamic energy consumption information of the data center refrigeration system, and uses the trained dynamic energy consumption prediction model to extract and identify the dynamic energy consumption information of each target historical time period and the corresponding multi-source parameter information to predict the dynamic energy consumption information of the current time period. This can effectively improve the accuracy and effect of the system dynamic energy consumption information prediction, which is conducive to improving the accuracy of the energy consumption prediction of the data center refrigeration system.

[0151] Based on the content of the above embodiment, as an optional embodiment, before inputting the dynamic energy consumption information of each target historical period and the multi-source parameter information corresponding to each target historical period into the dynamic energy consumption prediction model to obtain the dynamic energy consumption information of the system in the current period, the method further includes:

[0152] Performing data enhancement on the original dynamic energy consumption information sample data set to obtain dynamic energy consumption information samples for each second historical period;

[0153] The dynamic energy consumption information samples of each second historical period, as well as the corresponding multi-source parameter information and dynamic energy consumption information labels are used as a group of training samples, and multiple groups of training samples are obtained;

[0154] For any group of training samples among the multiple groups of training samples, the training samples are input into the dynamic energy consumption prediction model, and prediction information corresponding to the training samples is output;

[0155] Using a preset loss function, based on the prediction information corresponding to the training sample and the dynamic energy consumption information label corresponding to the training sample, a loss value is calculated, and based on the loss value, the model parameters of the dynamic energy consumption prediction model are adjusted until the loss value is less than a preset threshold or the number of training times reaches a preset number;

[0156] The model parameters obtained when the loss value is less than a preset threshold or the number of training times reaches a preset number are used as the model parameters of the trained dynamic energy consumption prediction model.

[0157] Specifically, the preset loss function described in the embodiment of the present application refers to a loss function pre-set in the dynamic energy consumption prediction model, which is used for model evaluation.

[0158] The preset threshold described in the embodiment of the present application refers to a threshold pre-set by the model, which is used to obtain the minimum loss value and complete model training.

[0159] The preset number of times described in the embodiment of the present application refers to the maximum number of pre-set model iteration training times.

[0160] In an embodiment of the present application, before the dynamic energy consumption information of each target historical period and the multi-source parameter information corresponding to each target historical period are input into the dynamic energy consumption prediction model, the dynamic energy consumption prediction model needs to be trained to obtain a trained dynamic energy consumption prediction model.

[0161] In the embodiment of the present application, the historical energy consumption data samples of the data center refrigeration system are first obtained, and the dynamic energy consumption information of each historical period in the sample is obtained according to the aforementioned data clustering and decomposition method, thereby forming an original dynamic energy consumption information sample data set. Furthermore, in order to overcome the shortage and insufficient richness of training data samples, the obtained original dynamic energy consumption information sample data set can be enhanced to generate new data samples, and finally the dynamic energy consumption information samples of each second historical period for model training can be obtained.

[0162] Furthermore, in an embodiment of the present application, the dynamic energy consumption information samples of each second historical period, and the corresponding multi-source parameter information and dynamic energy consumption information labels are used as a group of training samples, and multiple groups of training samples are obtained.

[0163] Then, after obtaining multiple sets of training samples, the multiple sets of training samples are sequentially input into the dynamic energy consumption prediction model, and the dynamic energy consumption prediction model is trained using the multiple sets of training samples, that is:

[0164] The dynamic energy consumption information samples of the historical period in each group of training samples, as well as the corresponding multi-source parameter information and dynamic energy consumption information labels are input into the dynamic energy consumption prediction model, and the prediction information corresponding to each group of training samples is output.

[0165] On this basis, the preset loss function is used to substitute the prediction information corresponding to the training sample and the dynamic energy consumption information label corresponding to the training sample into the preset loss function for calculation to obtain the loss value. Further, after the loss value is calculated, the training process ends. Based on the loss value, the model parameters of the dynamic energy consumption prediction model are adjusted to update the weight parameters of each layer of the model in the dynamic energy consumption prediction model, and then the next training is carried out, and the model training is repeated iteratively.

[0166] During the training process, if the training results for a certain group of training samples meet the preset training termination conditions, such as the corresponding calculated loss value is less than the preset threshold, or when the current number of iterations reaches the preset number, the model loss value can be controlled within the convergence range, then the model training ends. At this time, the model parameters obtained when the loss value is less than the preset threshold or the number of training times reaches the preset number are used as the model parameters of the trained dynamic energy consumption prediction model, thereby obtaining a trained dynamic energy consumption prediction model.

[0167] The method of the embodiment of the present application uses multiple groups of training samples consisting of dynamic energy consumption information samples from different historical periods, as well as corresponding multi-source parameter information and dynamic energy consumption information labels, to repeatedly iteratively train the dynamic energy consumption prediction model, and control the loss value of the prediction model within the convergence range, thereby facilitating the improvement of the accuracy of the system dynamic energy consumption prediction results output by the model, and improving the accuracy of the energy consumption prediction of the data center refrigeration system.

[0168] Based on the content of the above embodiment, as an optional embodiment, data enhancement is performed on the original dynamic energy consumption information sample data set to obtain dynamic energy consumption information samples for each second historical period, including:

[0169] Performing variational inference analysis on each original dynamic energy consumption information sample in the original dynamic energy consumption information sample data set, generating new samples corresponding to each original dynamic energy consumption information sample, and forming a new sample data set;

[0170] According to the meteorological parameter information of the system in the current time period, configure the target data ratio of the original dynamic energy consumption information sample data set and the new sample data set;

[0171] Based on the original dynamic energy consumption information sample data set and the new sample data set under the target data ratio, a dynamic energy consumption information sample for each second historical period is generated.

[0172] Specifically, in an embodiment of the present application, in order to expand the model training sample data, in the process of data enhancement of the original dynamic energy consumption information sample data set, first, a variational autoencoder (VAE) can be used to perform variational inference analysis on each original dynamic energy consumption information sample in the original dynamic energy consumption information sample data set. The specific process is as follows:

[0173] The existing small amount of data center cooling system fluctuation energy consumption sample data, that is, the original dynamic energy consumption information sample data, is input into the encoder of the VAE model. Under the constraint of the conditional information, the latent variables are obtained through variational inference, and then the enhanced data is generated through the decoder of the VAE model in combination with the conditional information. In this process, the equal proportion generation rule can be introduced to generate new samples corresponding to each original dynamic energy consumption information sample according to the various conditions in the conditional information, forming a new sample data set. For example, if the conditional information contains four weeks of July data and one week of August data, the ratio of the generated data corresponding to July and August should also be four times.

[0174] In an embodiment of the present application, the variational inference technology is used to generate new data samples to expand the original data set. At the same time, since the conditional information selected for the generated data is a time variable and an operating mode label, it is ensured that the generated data is very consistent with the actual data distribution, timing information and operating mode, and is suitable for solving the energy consumption prediction of the data center cooling system under data shortage.

[0175] In the embodiments of the present application, considering the hourly increase in the amount of energy consumption data samples of the actual data center refrigeration system that can be collected, and the seasonal development and change of energy consumption data, it is necessary to enhance the data under real-time seasonal conditions, while at the same time weakening but retaining the information characteristics of the data under historical seasonal conditions (the historical data with original seasonal characteristics in the above-mentioned directional generated data and the original dynamic energy consumption information sample data set). By setting a certain periodic weight allocation mechanism, a continuous enhancement strategy for time series periodicity is implemented.

[0176] Specifically, the weight allocation mechanism dynamically adjusts the relative contributions of the above generated data and the original historical data in each seasonal cycle. For example, when seasonal weather changes are involved (such as summer just turning into autumn, etc.), the weight of the generated data can be increased to reflect the latest trends and patterns of real-time data center energy consumption. On the contrary, in the stable stage (all in summer weather, etc.), the weight of the generated data can be reduced to retain more information features of the historical data, thereby weakening the impact of the generated data.

[0177] During the specific implementation process, the target data ratio of the original dynamic energy consumption information sample data set and the new sample data set can be dynamically configured based on the system's meteorological parameter information in the current time period, such as outdoor dry-bulb temperature and outdoor relative humidity, to determine whether there are seasonal changes in the current time period.

[0178] For example, based on the outdoor dry-bulb temperature and outdoor relative humidity information in the current time period, if it is judged that seasonal changes occur in the current time period, the learning weight of the newly generated sample data can be increased, and the learning weight of the original historical data can be reduced. That is, the target data ratio can be adjusted from 5:5 to 3:7. When there is no seasonal change in the next period of time (such as two weeks or three weeks), the learning weight is gradually adjusted, and the target data ratio is adjusted to 4:6, and finally the learning weight mode in which the generated data is the same as the original historical data is restored.

[0179] Furthermore, under the above-mentioned weight distribution mechanism, the target data ratio is dynamically adjusted, that is, the sample data ratio between the original dynamic energy consumption information sample data and the new sample data used for model training is dynamically adjusted. Finally, based on the original dynamic energy consumption information sample data set and the new sample data set under the target data ratio, the dynamic energy consumption information samples of each second historical period are obtained according to each dynamic energy consumption information sample data and the corresponding historical period, so as to be used for the subsequent training of the dynamic energy consumption prediction model.

[0180] The method of the embodiment of the present application takes into account the hourly increase in the sample volume of energy consumption data of the real data center refrigeration system that can be collected, as well as the seasonal development and change of the energy consumption data. It is necessary to perform data enhancement under real-time seasonal conditions, while at the same time weakening but retaining the data information characteristics under the original historical seasonal conditions. By setting a certain periodic weight distribution mechanism, the dynamic energy consumption prediction model can be effectively enhanced in terms of seasonal changes, thereby further improving the accuracy of the model prediction.

[0181] Figure 4 is a flow chart of the model training in the method for predicting energy consumption of a data center refrigeration system provided in an embodiment of the present application, such as Figure 4As shown, in an embodiment of the present application, the data in a small amount of original dynamic energy consumption information sample data set of the data center refrigeration system is input into the encoder of the VAE model, and the latent variables are obtained by variational inference under the constraints of the conditional information being the time variable and the operation mode label, and then the enhanced data is generated through the decoder of the VAE model in combination with the conditional information. In this process, an equal proportion generation rule can be introduced, and a new sample corresponding to each original dynamic energy consumption information sample is generated in this proportion according to each condition in the conditional information to form a new sample data set. Based on the original dynamic energy consumption information sample data set and the new sample data set, the dynamic energy consumption information sample data set used to train the dynamic energy consumption prediction model is reconstructed to obtain the dynamic energy consumption information sample of each second historical period.

[0182] Each data center fluctuation energy consumption sample, i.e., dynamic energy consumption information sample, carries multi-source characteristic parameter information of time parameters, meteorological parameters, and cooling mode. The time parameter information may specifically include month, week, and hour; the meteorological parameter information may specifically include outdoor dry-bulb temperature, outdoor relative humidity, wind speed, and solar radiation.

[0183] Furthermore, CNN and LSTM models are used as dynamic energy consumption prediction models, and the dynamic energy consumption information samples of each second historical period and the multi-source parameter information and dynamic energy consumption information labels they carry are used to train the dynamic energy consumption prediction model. The ReLu activation function is used in advance to reduce the dimension of the sample data and extract the high-level features therein. The CNN model specifically includes two layers of convolutional neural networks, which are used to process the spatial features in the sample data. Subsequently, the sample data is processed by the LSTM model to process the temporal dependencies, and finally the prediction results of the dynamic energy consumption information are output through the linear layer.

[0184] In an embodiment of the present application, in the process of energy consumption prediction, a sliding window model is used to further enhance the extraction of time series data, with 25 grids as the window size, that is, the energy consumption data of the 25th hour is predicted using the previous 24 hours, and a total of 168 hours (one week) of predicted fluctuating energy consumption data is output.

[0185] Table 1

[0186]

[0187] In a specific embodiment of the present application, in the scenario where the energy consumption data of the data center cooling system is insufficient, a time span of one week, one month, and then one quarter is selected as the data length of the training set. In this setting, training is performed every other week. Using the determination coefficient R 2 To measure the degree of fit of the prediction model results compared to the true value, that is, the model's ability to explain the variance of the observed data.

[0188] Table 1 shows the R of the prediction results of the traditional prediction method using only the LSTM model and the above prediction method of the embodiment of the present application. 2 From Table 1, it can be found that the prediction method proposed in this application has a R 2 Both are closer to 1 than the traditional LSTM-based method and are not less than 0.994. The prediction method proposed in this application has a better degree of fit.

[0189] Figure 5 is a schematic diagram of the structure of the data center refrigeration system energy consumption prediction device provided in the embodiment of the present application, such as Figure 5 As shown, including:

[0190] An acquisition module 510 is used to acquire energy consumption information of the system in a plurality of first historical periods before the current period;

[0191] The first processing module 520 is used to determine the dynamic energy consumption information of the system in each first historical period and the static energy consumption information in the current period based on the energy consumption information of each first historical period;

[0192] The prediction module 530 is used to predict the dynamic energy consumption information of the system in the current period based on the dynamic energy consumption information of the system in multiple target historical periods and the multi-source parameter information corresponding to each target historical period; the multiple target historical periods are the target number of historical periods adjacent to the current period in the multiple first historical periods; the multi-source parameter information includes time parameter information, meteorological parameter information and operation mode information;

[0193] The second processing module 540 is used to determine the energy consumption information of the system in the current period based on the static energy consumption information of the current period and the dynamic energy consumption information of the current period.

[0194] It should be understood that the above-mentioned device is used to execute the method in the above-mentioned embodiment. The implementation principle and technical effect of the corresponding program module in the device are similar to those described in the above-mentioned method. The working process of the device can refer to the corresponding process in the above-mentioned method, which will not be repeated here.

[0195] The energy consumption prediction device for a data center refrigeration system in an embodiment of the present application fully utilizes the characteristics of periodic changes and dynamic random changes in energy consumption data in historical data, obtains energy consumption information of the system in multiple historical periods before the current period, performs data decomposition based on the energy consumption information of each historical period, determines the dynamic energy consumption information of the system in each historical period and the static energy consumption information in the current period, thereby extracting the dynamic part with large time series correlation and the static part with small time series correlation, thereby using the dynamic energy consumption information of the system in multiple target historical periods and the corresponding multi-source parameter information to predict the dynamic energy consumption information of the system in the current period, and then obtains the energy consumption information of the system in the current period based on the static energy consumption information of the current period and the predicted dynamic energy consumption information of the current period, which can effectively improve the accuracy of energy consumption prediction of the data center refrigeration system and improve the energy consumption prediction effect.

[0196] Based on the method in the above embodiment, the embodiment of the present application provides an electronic device, such as Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other via the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the method in the above embodiment.

[0197] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application.

[0198] Based on the method in the above embodiment, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method in the above embodiment.

[0199] Based on the method in the above embodiment, an embodiment of the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method in the above embodiment.

[0200] It is understandable that the processor in the embodiment of the present application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. The general-purpose processor can be a microprocessor or any conventional processor.

[0201] The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC.

[0202] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions may be transmitted from a website site, a computer, a server or a data center to another website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state disk (SSD)), etc.

[0203] It should be understood that the various numerical numbers involved in the embodiments of the present application are only used for the convenience of description and are not used to limit the scope of the embodiments of the present application.

[0204] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for predicting energy consumption of a data center refrigeration system, characterized in that: include: Obtaining energy consumption information of the system in a plurality of first historical periods before the current period; Based on the energy consumption information of each of the first historical time periods, determining the dynamic energy consumption information of the system in each of the first historical time periods and the static energy consumption information in the current time period; Based on the dynamic energy consumption information of the system in multiple target historical time periods and the multi-source parameter information corresponding to each of the target historical time periods, predict the dynamic energy consumption information of the system in the current time period; the multiple target historical time periods are the first historical time periods of the target number of the multiple first historical time periods that are adjacent to the current time period; The multi-source parameter information includes time parameter information, meteorological parameter information and operation mode information; Based on the static energy consumption information of the current time period and the dynamic energy consumption information of the current time period, the energy consumption information of the system in the current time period is determined.

2. The method for predicting energy consumption of a data center refrigeration system according to claim 1, characterized in that: The determining, based on the energy consumption information of each of the first historical time periods, the dynamic energy consumption information of the system in each of the first historical time periods and the static energy consumption information in the current time period includes: Performing cluster analysis on the energy consumption information of each of the first historical periods according to the trend of data changes within a preset time period to determine a plurality of clusters; Decomposing the energy consumption information of each of the first historical periods in each of the clusters respectively, and determining the dynamic energy consumption information and the static energy consumption information of each of the first historical periods in each of the clusters; According to the dynamic energy consumption information of each of the first historical periods in each of the clusters, obtaining the dynamic energy consumption information of the system in each of the first historical periods; Based on the static energy consumption information of each of the first historical time periods in each of the clusters, the static energy consumption information of the system in the current time period is determined.

3. The method for predicting energy consumption of a data center refrigeration system according to claim 2, characterized in that: The step of respectively performing data decomposition on the energy consumption information of each of the first historical periods in each of the clusters to determine the dynamic energy consumption information and the static energy consumption information of each of the first historical periods in each of the clusters includes: Averaging the energy consumption information of the periodic time periods corresponding to the first historical time periods in each of the clusters to obtain static energy consumption information of the first historical time periods in each of the clusters; The dynamic energy consumption information of each of the first historical periods in each of the clusters is obtained according to the energy consumption information and the static energy consumption information of each of the first historical periods in each of the clusters.

4. The method for predicting energy consumption of a data center refrigeration system according to claim 2, characterized in that: The determining, based on the static energy consumption information of each of the first historical time periods in each of the clusters, the static energy consumption information of the system in the current time period includes: Determine the historical time period corresponding to the current time period; Determine, from each of the clusters, a target cluster to which the historical time period corresponding to the current time period belongs; Based on the static energy consumption information of each of the first historical time periods in the target cluster, the static energy consumption information of the system in the current time period is determined.

5. The method for predicting energy consumption of a data center cooling system according to any one of claims 1 to 4, characterized in that: The predicting the dynamic energy consumption information of the system in the current period based on the dynamic energy consumption information of the system in multiple target historical periods and the multi-source parameter information corresponding to each of the target historical periods includes: Inputting the dynamic energy consumption information of each of the target historical time periods and the multi-source parameter information corresponding to each of the target historical time periods into a dynamic energy consumption prediction model to obtain the dynamic energy consumption information of the system in the current time period; The dynamic energy consumption prediction model is trained based on dynamic energy consumption information samples of multiple second historical time periods, and corresponding multi-source parameter information and dynamic energy consumption information labels; the dynamic energy consumption information samples of multiple second historical time periods are determined based on the original dynamic energy consumption information sample data set.

6. The method for predicting energy consumption of a data center refrigeration system according to claim 5, characterized in that: Before inputting the dynamic energy consumption information of each of the target historical time periods and the multi-source parameter information corresponding to each of the target historical time periods into the dynamic energy consumption prediction model to obtain the dynamic energy consumption information of the system in the current time period, the method further includes: Performing data enhancement on the original dynamic energy consumption information sample data set to obtain dynamic energy consumption information samples for each of the second historical periods; Taking each dynamic energy consumption information sample of the second historical period, and the corresponding multi-source parameter information and dynamic energy consumption information label as a group of training samples, and obtaining multiple groups of the training samples; For any one group of training samples among the multiple groups of training samples, inputting the training samples into a dynamic energy consumption prediction model, and outputting prediction information corresponding to the training samples; Using a preset loss function, based on the prediction information corresponding to the training sample and the dynamic energy consumption information label corresponding to the training sample, a loss value is calculated, and based on the loss value, the model parameters of the dynamic energy consumption prediction model are adjusted until the loss value is less than a preset threshold or the number of training times reaches a preset number of times; The model parameters obtained when the loss value is less than the preset threshold or the number of training times reaches the preset number of times are used as the model parameters of the trained dynamic energy consumption prediction model.

7. The method for predicting energy consumption of a data center refrigeration system according to claim 6, characterized in that: The step of performing data enhancement on the original dynamic energy consumption information sample data set to obtain the dynamic energy consumption information sample of each second historical period includes: Performing variational inference analysis on each original dynamic energy consumption information sample in the original dynamic energy consumption information sample data set to generate a new sample corresponding to each original dynamic energy consumption information sample to form a new sample data set; According to the meteorological parameter information of the system in the current time period, configuring the target data ratio of the original dynamic energy consumption information sample data set and the new sample data set; Based on the original dynamic energy consumption information sample data set and the new sample data set at the target data ratio, a dynamic energy consumption information sample for each second historical period is generated.

8. A data center refrigeration system energy consumption prediction device, characterized in that: include: An acquisition module, used to acquire energy consumption information of the system in a plurality of first historical time periods before the current time period; A first processing module, configured to determine the dynamic energy consumption information of the system in each of the first historical time periods and the static energy consumption information in the current time period based on the energy consumption information of each of the first historical time periods; A prediction module, configured to predict the dynamic energy consumption information of the system in the current period based on the dynamic energy consumption information of the system in a plurality of target historical periods and the multi-source parameter information corresponding to each of the target historical periods; the plurality of target historical periods are the target number of historical periods adjacent to the current period in the plurality of first historical periods; The multi-source parameter information includes time parameter information, meteorological parameter information and operation mode information; The second processing module is used to determine the energy consumption information of the system in the current time period based on the static energy consumption information of the current time period and the dynamic energy consumption information of the current time period.

9. An electronic device, characterized in that: include: at least one memory for storing a computer program; At least one processor is used to execute the program stored in the memory. When the program stored in the memory is executed, the processor is used to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program runs on a processor, the processor is caused to execute the method according to any one of claims 1 to 7.

Images