CN118297285B - Energy efficiency optimization method and system for data center - Google Patents
Energy efficiency optimization method and system for data center Download PDFInfo
- Publication number
- CN118297285B CN118297285B CN202410707432.XA CN202410707432A CN118297285B CN 118297285 B CN118297285 B CN 118297285B CN 202410707432 A CN202410707432 A CN 202410707432A CN 118297285 B CN118297285 B CN 118297285B
- Authority
- CN
- China
- Prior art keywords
- trend
- data sequence
- data
- window
- seasonal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000005457 optimization Methods 0.000 title claims abstract description 23
- 230000008859 change Effects 0.000 claims abstract description 149
- 238000012216 screening Methods 0.000 claims abstract description 7
- 230000001932 seasonal effect Effects 0.000 claims description 98
- 230000000875 corresponding effect Effects 0.000 claims description 80
- 238000013507 mapping Methods 0.000 claims description 16
- 230000000737 periodic effect Effects 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 6
- 230000002596 correlated effect Effects 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 12
- 238000012545 processing Methods 0.000 abstract description 2
- 238000000354 decomposition reaction Methods 0.000 description 14
- 238000005265 energy consumption Methods 0.000 description 10
- 238000001816 cooling Methods 0.000 description 6
- 230000001105 regulatory effect Effects 0.000 description 6
- 230000001276 controlling effect Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000003064 k means clustering Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000003442 weekly effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention relates to the technical field of data processing, in particular to an energy efficiency optimization method and system for a data center, comprising the following steps: and decomposing the equipment load data sequence by using a plurality of trend windows with different sizes to obtain trend items and season items, dividing each trend item into a plurality of trend data sequence segments, acquiring a load rapid change rate index of each trend data sequence segment, thus obtaining a trend rapid change index in the trend item corresponding to each trend window, dividing each season item into a plurality of season data sequence segments, acquiring a season periodicity index in the season item corresponding to each trend window, screening out an optimal trend window by taking the trend item corresponding to the optimal trend window as a regulation index, and optimizing the energy efficiency of the data center. According to the method, the optimal trend window is selected through screening, the optimal trend item is obtained, and the optimal trend item is used as a regulation and control index, so that the energy efficiency optimization effect of the data center is improved.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to an energy efficiency optimization method and system for a data center.
Background
The cooling system of a data center is one of the major parts of the overall data center energy consumption, and the cooling system needs to operate around the clock to maintain proper temperature and humidity. The general data center can be flexibly deployed and expanded according to actual equipment load requirements, so that energy consumption operation is minimized, and energy consumption of a cooling system of the data center is optimized. The change in the trend of the energy consumption of the cooling system can generally be determined from historical energy consumption data in order to achieve a dynamic adjustment of the cooling system in order to reduce the energy consumption.
The existing problems are as follows: the periodic change of the energy consumption data of the cooling center influences the judgment of the energy consumption change trend, and the STL decomposition algorithm can eliminate the periodic change and determine the change trend of the energy consumption data. However, when the STL decomposition algorithm is used for decomposing the equipment load change data, the size of the trend window is generally obtained only by using an empirical value, and under the condition of various load changes, the unsuitable size of the trend window can reduce the decomposition effect of the trend item, so that the energy efficiency optimization effect of the data center is reduced.
Disclosure of Invention
The invention provides an energy efficiency optimization method and system for a data center, which are used for solving the existing problems.
The invention discloses an energy efficiency optimization method and system for a data center, which adopts the following technical scheme:
An embodiment of the present invention provides an energy efficiency optimization method for a data center, the method including the steps of:
Acquiring a device load data sequence of a data center; decomposing the equipment load data sequence by using a plurality of trend windows with different sizes respectively to obtain trend items and season items corresponding to each trend window;
Dividing each trend item into a plurality of trend data sequence segments; according to the trend data change in each trend data sequence segment, obtaining a load rapid change rate index of each trend data sequence segment;
Clustering all trend data sequence segments divided by each trend item according to the load rapid change rate index of each trend data sequence segment to obtain a plurality of clustering clusters; according to the difference between the quantity distribution of the trend data sequence segments among the clusters and the load rapid change rate index, obtaining a trend rapid change index in a trend item corresponding to each trend window;
Dividing each season term into a plurality of season data sequence segments; the seasonal data in the seasonal data sequence segment is a fixed trend change; according to the data quantity in each seasonal data sequence segment and the difference between the seasonal data, obtaining a seasonal periodic index in a seasonal item corresponding to each trend window;
According to the trend rapid change index and the seasonal periodical index corresponding to each trend window, the optimal trend window is screened out from all trend windows; and optimizing the energy efficiency of the data center by taking a trend item corresponding to the optimal trend window as a regulation index.
Further, the load rapid change rate index of each trend data sequence segment comprises the following specific steps:
In each trend data sequence segment divided by the corresponding trend item of each trend window, the average value of the differences of all adjacent trend data is recorded as the adjacent difference of each trend data sequence segment;
Determining the overall difference of each trend data sequence segment according to the maximum value and the minimum value in each trend data sequence segment;
Obtaining a load rapid change rate index of each trend data sequence segment according to the adjacent difference and the overall difference of each trend data sequence segment; the adjacent differences are positively correlated to the overall differences.
Further, the overall difference comprises the following specific steps:
And calculating the difference value of the maximum value and the minimum value in each trend data sequence segment, and recording the ratio of the difference value to the minimum value as the integral difference of each trend data sequence segment.
Further, the trend rapid change index in the trend item corresponding to each trend window comprises the following specific steps:
The cluster with the least number of the sequence segments containing the trend data is marked as a target cluster; the cluster with the largest number of the sequence segments containing the trend data is marked as a reference cluster;
according to the difference between the load rapid change rate indexes of the trend data sequence segments in the target cluster and the reference cluster, obtaining trend change difference in trend items corresponding to each trend window;
and obtaining a trend rapid change index in the trend item corresponding to each trend window according to the number of the trend data sequence segments in the target cluster and the reference cluster and the trend change difference in the trend item corresponding to each trend window.
Further, the trend change difference in the trend item corresponding to each trend window comprises the following specific steps:
The average value of the load rapid change rate indexes of all trend data sequence segments in the target cluster is recorded as a first average value;
The average value of the load rapid change rate indexes of all the trend data sequence segments in the reference cluster is recorded as a second average value;
And recording the difference between the first mean value and the second mean value as the trend change difference in the trend item corresponding to each trend window.
Further, the trend rapid change index in the trend item corresponding to each trend window comprises the following specific steps:
Acquiring an inverse proportion mapping value of the ratio of the number of the trend data sequence segments in the target cluster to the number of the trend data sequence segments in the reference cluster, and determining a trend rapid change index in a trend item corresponding to each trend window according to the trend change difference of the inverse proportion mapping value and the trend item corresponding to each trend window; and the inverse proportion mapping value is positively correlated with the trend change difference in the trend item corresponding to each trend window.
Further, the seasonal periodic index in the seasonal item corresponding to each trend window comprises the following specific steps:
obtaining the change amplitude of each seasonal data sequence segment according to the difference between the seasonal data in each seasonal data sequence segment;
The number of the seasonal data sequence segments with the change amplitude larger than a preset judging threshold value is recorded as the number of target periods; counting the data quantity in each seasonal data sequence segment, taking the average value of the data quantity in all the seasonal data sequence segments as a standard period duration, and recording the average value of the difference between the data quantity in all the seasonal data sequence segments and the standard period duration as the period duration difference of the seasonal item mark corresponding to each trend window; and carrying out inverse proportion mapping on the product of the period duration difference and the target period number, and determining a seasonal periodicity index in a seasonal term corresponding to each trend window.
Further, the changing amplitude of each seasonal data sequence segment comprises the following specific steps:
the difference between the first and last season data in each of the sequence of season data segments is noted as the magnitude of change in each of the sequence of season data segments.
Further, the step of screening the optimal trend window from all the trend windows comprises the following specific steps:
According to the trend rapid change index and the seasonal periodical index corresponding to each trend window, determining a load change capturing index corresponding to each trend window; the trend rapid change index and the seasonal periodical index are positively correlated;
And (3) in the load change capturing indexes corresponding to all the trend windows, marking the trend window corresponding to the maximum load change capturing index as an optimal trend window.
The invention also provides an energy efficiency optimization system for the data center, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program stored in the memory so as to realize the steps of the energy efficiency optimization method for the data center.
The technical scheme of the invention has the beneficial effects that:
In the embodiment of the invention, a device load data sequence of a data center is acquired; decomposing the equipment load data sequence by using a plurality of trend windows with different sizes respectively to obtain trend items and season items corresponding to each trend window; dividing each trend item into a plurality of trend data sequence segments; acquiring a load rapid change rate index of each trend data sequence segment, thereby acquiring a trend rapid change index in a trend item corresponding to each trend window; the analysis among the segments of the trend items determines the effect of the trend items decomposed under different trend windows, and ensures the accuracy of selecting the optimal trend window. Dividing each season term into a plurality of season data sequence segments; the method comprises the steps of obtaining seasonal periodic indexes in seasonal items corresponding to each trend window, determining the effect of the seasonal items decomposed under different trend windows through analysis among segments of the seasonal items, further guaranteeing the accuracy of selecting the optimal trend windows, guaranteeing the credibility of regulation indexes, improving the optimization effect, and screening the optimal trend windows from all trend windows by combining the decomposition effects of the trend items and the seasonal items; and the trend item corresponding to the optimal trend window is used as a regulation and control index, so that the credibility of the regulation and control index is ensured, and the energy efficiency of the data center is optimized. The optimal trend window is screened out, so that the optimal trend item is obtained, and the optimal trend item is used as a regulation index, so that the energy efficiency optimization effect of the data center is improved.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of steps of an energy efficiency optimization method for a data center of the present invention;
Fig. 2 is a flowchart of the acquisition of the control command in the present embodiment.
Detailed Description
In order to further describe the technical means and effects adopted by the present invention to achieve the preset purposes, the following detailed description refers to the specific implementation, structure, characteristics and effects of an energy efficiency optimization method and system for a data center according to the present invention with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following specifically describes a specific scheme of an energy efficiency optimization method and system for a data center provided by the invention with reference to the accompanying drawings.
Referring to fig. 1, a flowchart illustrating a method for optimizing energy efficiency for a data center according to an embodiment of the present invention includes the following steps:
Step S001: acquiring a device load data sequence of a data center; and decomposing the equipment load data sequence by using a plurality of trend windows with different sizes respectively to obtain trend items and season items corresponding to each trend window.
What needs to be described is: according to the embodiment, the trend item of the equipment load data sequence is obtained, the temperature of the data center is regulated and controlled according to the equipment load change trend, and the energy efficiency requirement is reduced. The equipment load condition of the data center in the current year is collected in an equipment data storage database of the data center, the collection frequency is half an hour, and an equipment load data sequence of the data center is obtained, and is described by way of example.
Presetting a plurality of trend windows with different sizes, and decomposing the equipment load data sequence by using an SLT decomposition algorithm according to each preset trend window to obtain trend items, season items and residual items of equipment load data sequence decomposition.
What needs to be described is: STL (Seasonal and Trend decomposition using Loess) the decomposition algorithm is a well known technique, a specific method is not described herein, and the text name is a time series decomposition algorithm, which can decompose a time series data sequence into a trend term and a season term, and a residual term. Two key parameters required by the algorithm are trend window and seasonal window size, respectively. The trend window size is used for fitting trend changes of the equipment load, the seasonal window size determines periodic changes of the equipment load, the trend window sizes of a plurality of different sizes are preset in the embodiment, and are respectively 1 day, 2 days, 3 days, 4 days and 5 days, and the duration of the seasonal window size is set to 7 days, which is described by way of example, because weekly use modes and load fluctuation conditions can be effectively identified through weekly data analysis. The usage rules of weekends and weekdays are often different, and the seasonal window size set to one week can enable the model to adapt to the periodic change better, so that the seasonal window size in STL decomposition can be determined as the equipment load data of one week acquired by the data center.
Further to be described is: in this embodiment, the trend change of the device load needs to be determined, so different trend window sizes can be selected to decompose the device load data sequence. If the trend window size is chosen too small, the trend term may be too sensitive to random fluctuations in the data, affecting the accuracy of the seasonal term. If the trend window size is too large, the trend term may be too smooth, so that important changes in the trend are ignored, and the extraction of the season term is affected.
Step S002: dividing each trend item into a plurality of trend data sequence segments; and obtaining the load rapid change rate index of each trend data sequence segment according to the trend data change in each trend data sequence segment.
What needs to be described is: the present embodiment needs to capture the case of rapid change in the device load, so that the case that there is rapid change in the device load can be shown in the trend item that needs to be acquired. The equipment of the data center can generate a condition of relatively large rapid load change during holidays or emergency, and the load change is generally caused by the use habit and service expansion condition of people at ordinary times, so that the frequency of the data segment with the rapid load change in the whole is relatively small, and the data segment is obviously different from other changed data segments, and the trend item is segmented.
In one embodiment of the invention, an APCA data segmentation algorithm is used to divide the trend term corresponding to each trend window into a plurality of trend data sequence segments. Wherein APCA (ADAPTIVE PIECEWISE Constant Approximation, adaptive segmentation constant approximation) is a data segmentation algorithm, which is a well-known technique, and the specific method is not described here.
What needs to be described is: for the case of rapid changes in the device load, the amplitude of the change between adjacent data in the segment will be relatively large and the difference between the maximum and minimum values in the segment will be large, which represents the overall difference in the segment. Thus, the adjacent differences of the trend data sequence segments are used for determining whether the internal load of the segment is excessively changed, the larger the value is, the more rapid the load change is indicated, and the whole difference of the trend data sequence segments is used for indicating that the load change is relatively rapid when the load state is relatively low, namely, the whole difference is larger when the load change range is relatively large. The greater the load rapid rate index, the more likely the segment is to be in the case of rapid changes in equipment load.
Preferably, in one embodiment of the present invention, the method for acquiring the load rapid change rate indicator includes:
And calculating the absolute value of the difference value of two adjacent trend data in any trend data sequence segment divided by the trend item corresponding to any trend window, and recording the average value of the absolute values of the difference values of all adjacent trend data as the adjacent difference of the trend data sequence segment. And determining the overall difference of the trend data sequence segments according to the maximum value and the minimum value in the trend data sequence segments. And (3) recording the product of the adjacent difference and the integral difference of the trend data sequence segment as a load rapid change rate index of the trend data sequence segment.
What needs to be described is: because the load of the device is 100% at maximum, but generally exceeds 90%, the device expands, so that the rapid change of the load generally exists in a condition of low load state, and the maximum and minimum loads in the segments are also larger through rapid increase of the load, thereby representing the overall difference of the trend data sequence segments.
Preferably, in one embodiment of the present invention, the method for acquiring the overall difference of the trend data sequence segments includes:
And counting the maximum value and the minimum value in each trend data sequence segment, and recording the ratio of the difference value of subtracting the minimum value from the maximum value to the minimum value as the integral difference of each trend data sequence segment.
Step S003: clustering all trend data sequence segments divided by each trend item according to the load rapid change rate index of each trend data sequence segment to obtain a plurality of clustering clusters; and obtaining the trend rapid change index in the trend item corresponding to each trend window according to the difference between the quantity distribution of the trend data sequence segments among the clusters and the load rapid change rate index.
What needs to be described is: the trend change degree of the trend data sequence segments corresponding to the similar load rapid change rate index is similar, so that the trend data sequence segments can be divided into different types, namely the trend data sequence segments with severe trend change or gentle trend change are divided into one type, and the difference between different trend changes is analyzed.
In one embodiment of the invention, in a trend item corresponding to any trend window, clustering operation is performed on all trend data sequence segments by using a K-means clustering algorithm according to the load rapid change rate index of each trend data sequence segment, so as to obtain a plurality of clusters. What needs to be described is: the K-means clustering algorithm is a known technique, and a specific method is not described herein, and the number of clusters preset in the algorithm in this embodiment is 2, which is described as an example.
What needs to be described is: the trend data sequence segments can be divided into severe or gentle segments, and as the trend of severe change is less in the trend item, and the difference between the trend of severe change and the trend of gentle change is larger, the trend change difference is larger, the trend rapid change index in the trend item is larger, the number of the trend data sequence segments of severe change is smaller, the trend rapid change index in the trend item is larger, and when the trend rapid change index is larger, the trend item is more accurate and reliable.
Preferably, in one embodiment of the present invention, the method for acquiring the trend rapid change index in the trend item includes:
and (5) marking the cluster with the least number of the sequence segments containing the trend data as a target cluster. And (5) marking the cluster with the largest number of the sequence segments containing the trend data as a reference cluster.
What needs to be described is: since the trend of the severe change is less in the trend item, the target cluster is often a segment of the severe change, whereas the reference cluster is often a segment of the gentle change. When the number of the trend data sequence segments in all the clusters is the same, one cluster is arbitrarily selected as a target cluster, and the other cluster is a reference cluster.
And obtaining the trend change difference in the trend item corresponding to any trend window according to the difference between the load rapid change rate indexes of the trend data sequence segments in the target cluster and the reference cluster.
Calculating an inverse proportion mapping value of the ratio of the number of the trend data sequence segments in the target cluster to the number of the trend data sequence segments in the reference cluster, and recording the product of the inverse proportion mapping value and the trend change difference in the trend item corresponding to any trend window as a trend rapid change index in the trend item corresponding to any trend window.
What needs to be described is: a preferred embodiment of the inverse ratio map of the above ratios is: the inverse mapping is performed with 1 minus the ratio, since the ratio is between 0 and 1, another example is: inputting the ratio toInverse proportion mapping in whichThe present embodiment uses an exponential function based on natural constantsTo present the inverse proportional relationship and normalization process.
What needs to be described is: the rapid load change rate index of the trend data sequence segment shows the trend change degree, so that the trend change difference in the trend item can be obtained according to the difference between the rapid load change rate indexes of the trend data sequence segment in the target cluster and the reference cluster.
Preferably, in one embodiment of the present invention, the method for acquiring the trend change difference in the trend item includes:
And (3) recording the average value of the load rapid change rate indexes of all the trend data sequence segments in the target cluster as a first average value. And (3) recording the average value of the load rapid change rate indexes of all the trend data sequence segments in the reference cluster as a second average value. And recording the absolute value of the difference between the first mean value and the second mean value as the trend change difference in the trend item corresponding to any trend window.
Step S004: dividing each season term into a plurality of season data sequence segments; the seasonal data in the seasonal data sequence segment is a fixed trend change; and obtaining the seasonal periodic index in the seasonal item corresponding to each trend window according to the data quantity in each seasonal data sequence segment and the difference between the seasonal data.
What needs to be described is: the seasonal window size is known as the number of one-week equipment load data acquired by the data center, but the STL is affected by load change during decomposition, and the case where the load change is fast may be mistaken for a case of one cycle change, so there is a case of shortening the cycle. And in the seasonal term, the difference between the obtained change magnitudes is not large according to the periodicity of the load change. If the load rapid change interval is misjudged to be a periodic change, the change amplitude suddenly increases in the period. It is therefore necessary to segment the seasonal items first according to periodicity.
In one embodiment of the invention, for a season term corresponding to any trend window, a first derivative method is used for deriving the season term to obtain a plurality of local extremum points. And dividing the seasonal item into a plurality of seasonal data sequence segments by taking the local extreme point as a dividing point. What needs to be described is: the data in the segments between adjacent local extreme points are in an incremental change or a decremental change, i.e. the data in the segments is a fixed trend change.
What needs to be described is: the difference between two adjacent local extreme points is the magnitude of one incremental or decremental change, which represents the magnitude of a segment change.
Preferably, in one embodiment of the present invention, the method for acquiring the variation amplitude of the seasonal data sequence segment includes:
And recording the absolute value of the difference value of the first and the last seasonal data in each seasonal data sequence segment as the change amplitude value of each seasonal data sequence segment.
What needs to be described is: the change amplitude in the seasonal items is changed rapidly to be an abnormal period, and the fewer the abnormal period is, the more accurate and reliable the decomposed seasonal items are described, namely the more accurate and reliable the corresponding trend items are. The smaller the difference between the length of each seasonal data sequence segment and the standard period length is, the more regular the period in the seasonal term is, the more accurate and reliable the seasonal term is, so that the larger the value of the seasonal periodical index is, the more accurate and reliable the decomposed seasonal term is.
Preferably, in one embodiment of the present invention, the method for acquiring a seasonal index in a seasonal item includes:
and recording the number of the seasonal data sequence segments with the change amplitude larger than a preset judgment threshold as the target period number. Counting the data quantity in each seasonal data sequence segment, taking the average value of the data quantity in all the seasonal data sequence segments as the standard period duration, calculating the absolute value of the difference value between the data quantity in each seasonal data sequence segment and the standard period duration, and recording the average value of the absolute value of the difference value between the data quantity in all the seasonal data sequence segments and the standard period duration as the period duration difference of the seasonal item programs corresponding to the trend window. And taking an inverse proportion mapping value of the product of the period duration difference and the target period number as a seasonal periodicity index in a seasonal item corresponding to the trend window.
What needs to be described is: in this embodiment, each trend window corresponds to a preset judgment threshold, and the process of obtaining the preset judgment threshold corresponding to each trend window is as follows: in all the seasonal data sequence segments divided by the seasonal items corresponding to any one of the trend windows, calculating the average value of the variation amplitude values of all the seasonal data sequence segments toThe mean value of the times is a judgment threshold value, wherein,For the preset judgment coefficient, in this embodimentThis is 1.5 and is described as an example. A preferred embodiment of the inverse proportional mapping value of the product is: taking the inverse of the product to perform inverse proportion mapping, another embodiment is: inputting the product toInverse proportion mapping in whichThe present embodiment uses an exponential function based on natural constantsTo present the inverse proportional relationship and normalization process.
Step S005: according to the trend rapid change index and the seasonal periodical index corresponding to each trend window, the optimal trend window is screened out from all trend windows; and optimizing the energy efficiency of the data center by taking a trend item corresponding to the optimal trend window as a regulation index.
What needs to be described is: the larger the trend rapid change index in the trend item is, the more accurate and reliable the decomposed trend item is, the larger the seasonal periodic index in the seasonal item is, the more accurate and reliable the decomposed seasonal item is, therefore, the product of the trend rapid change index and the seasonal periodic index is used as a load change capturing index corresponding to the trend window, the larger the value is, the better the decomposition effect is, the more accurate the capturing of the load change is, and therefore accurate and reliable data support is provided for energy efficiency optimization.
In one embodiment of the present invention, for a trend term and a season term corresponding to any one trend window, a product of a trend rapid change index in the trend term and a season periodicity index in the season term is recorded as a load change capturing index corresponding to the any one trend window.
Among the load change capturing indexes corresponding to all the trend windows, the trend window corresponding to the maximum load change capturing index is recorded as an optimal trend window, the trend item corresponding to the optimal trend window is input into the PID controller, and a regulating instruction is output, which is a known technology, and a specific method is not described herein. The flow of obtaining the regulation command is shown in fig. 2.
What needs to be described is: PID controllers are a very common and well known type of controller used to control industrial processes, mechanical systems, and other various systems. PID stands for Proportional (pro), integral (Integral) and Derivative (Derivative), representing three main parts of the controller, respectively. The trend item corresponding to the optimal trend window is an optimal trend item of equipment load data sequence decomposition, so that the temperature of the data center is regulated and controlled according to the optimal trend item of equipment load data sequence decomposition serving as a regulation and control index, for example: when the trend in the current period of time in the optimal trend item is stable, the regulating and controlling instruction output by the PID controller is constant in temperature, when the trend in the current period of time in the optimal trend item is large in incremental change, the regulating and controlling instruction output by the PID controller is low in temperature, and when the trend in the current period of time in the optimal trend item is large in decremental change, the regulating and controlling instruction output by the PID controller is high in temperature, so that the temperature is reasonably controlled, the cooling cost is reduced, the energy consumption is reduced, and the normal operation of hardware equipment of a data center is ensured.
The present invention has been completed.
In summary, in the embodiment of the present invention, a device load data sequence of a data center is obtained; decomposing the equipment load data sequence by using a plurality of trend windows with different sizes respectively to obtain trend items and season items corresponding to each trend window; dividing each trend item into a plurality of trend data sequence segments; acquiring a load rapid change rate index of each trend data sequence segment, thereby acquiring a trend rapid change index in a trend item corresponding to each trend window; dividing each season term into a plurality of season data sequence segments; acquiring seasonal periodic indexes in seasonal items corresponding to each trend window, and screening out an optimal trend window from all trend windows; and optimizing the energy efficiency of the data center by taking a trend item corresponding to the optimal trend window as a regulation index. According to the method, the optimal trend window is selected through screening, the optimal trend item is obtained, and the optimal trend item is used as a regulation and control index, so that the energy efficiency optimization effect of the data center is improved.
The invention also provides an energy efficiency optimization system for the data center, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program stored in the memory so as to realize the steps of the energy efficiency optimization method for the data center.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the invention, but any modifications, equivalent substitutions, improvements, etc. within the principles of the present invention should be included in the scope of the present invention.
Claims (7)
1. An energy efficiency optimization method for a data center, the method comprising the steps of:
Acquiring a device load data sequence of a data center; decomposing the equipment load data sequence by using a plurality of trend windows with different sizes respectively to obtain trend items and season items corresponding to each trend window;
Dividing each trend item into a plurality of trend data sequence segments; according to the trend data change in each trend data sequence segment, obtaining a load rapid change rate index of each trend data sequence segment;
Clustering all trend data sequence segments divided by each trend item according to the load rapid change rate index of each trend data sequence segment to obtain a plurality of clustering clusters; according to the difference between the quantity distribution of the trend data sequence segments among the clusters and the load rapid change rate index, obtaining a trend rapid change index in a trend item corresponding to each trend window;
Dividing each season term into a plurality of season data sequence segments; the seasonal data in the seasonal data sequence segment is a fixed trend change; according to the data quantity in each seasonal data sequence segment and the difference between the seasonal data, obtaining a seasonal periodic index in a seasonal item corresponding to each trend window;
according to the trend rapid change index and the seasonal periodical index corresponding to each trend window, the optimal trend window is screened out from all trend windows; the energy efficiency of the data center is optimized by taking a trend item corresponding to the optimal trend window as a regulation index;
The trend rapid change index in the trend item corresponding to each trend window comprises the following specific steps:
The cluster with the least number of the sequence segments containing the trend data is marked as a target cluster; the cluster with the largest number of the sequence segments containing the trend data is marked as a reference cluster;
according to the difference between the load rapid change rate indexes of the trend data sequence segments in the target cluster and the reference cluster, obtaining trend change difference in trend items corresponding to each trend window;
According to the number of the trend data sequence segments in the target cluster and the reference cluster and the trend change difference in the trend item corresponding to each trend window, obtaining a trend rapid change index in the trend item corresponding to each trend window;
the seasonal periodic index in the seasonal item corresponding to each trend window comprises the following specific steps:
obtaining the change amplitude of each seasonal data sequence segment according to the difference between the seasonal data in each seasonal data sequence segment;
the number of the seasonal data sequence segments with the change amplitude larger than a preset judging threshold value is recorded as the number of target periods; counting the data quantity in each seasonal data sequence segment, taking the average value of the data quantity in all the seasonal data sequence segments as a standard period duration, and recording the average value of the difference between the data quantity in all the seasonal data sequence segments and the standard period duration as the period duration difference of the seasonal item mark corresponding to each trend window; inversely proportional mapping is carried out on the product of the period duration difference and the target period quantity, and seasonal periodic indexes in seasonal items corresponding to each trend window are determined;
the method for screening the optimal trend window from all trend windows comprises the following specific steps:
According to the trend rapid change index and the seasonal periodical index corresponding to each trend window, determining a load change capturing index corresponding to each trend window; the trend rapid change index and the seasonal periodical index are positively correlated;
And (3) in the load change capturing indexes corresponding to all the trend windows, marking the trend window corresponding to the maximum load change capturing index as an optimal trend window.
2. The method for optimizing energy efficiency for a data center according to claim 1, wherein the load rapid rate of change indicator of each trend data sequence segment comprises the specific steps of:
In each trend data sequence segment divided by the corresponding trend item of each trend window, the average value of the differences of all adjacent trend data is recorded as the adjacent difference of each trend data sequence segment;
Determining the overall difference of each trend data sequence segment according to the maximum value and the minimum value in each trend data sequence segment;
Obtaining a load rapid change rate index of each trend data sequence segment according to the adjacent difference and the overall difference of each trend data sequence segment; the adjacent differences are positively correlated to the overall differences.
3. The method for optimizing energy efficiency for a data center of claim 2, wherein said overall difference comprises the specific steps of:
And calculating the difference value of the maximum value and the minimum value in each trend data sequence segment, and recording the ratio of the difference value to the minimum value as the integral difference of each trend data sequence segment.
4. The method for optimizing energy efficiency for a data center according to claim 1, wherein the trend change difference in the trend term corresponding to each trend window comprises the following specific steps:
The average value of the load rapid change rate indexes of all trend data sequence segments in the target cluster is recorded as a first average value;
The average value of the load rapid change rate indexes of all the trend data sequence segments in the reference cluster is recorded as a second average value;
And recording the difference between the first mean value and the second mean value as the trend change difference in the trend item corresponding to each trend window.
5. The method for optimizing energy efficiency for a data center according to claim 1, wherein the trend rapid change index in the trend term corresponding to each trend window comprises the following specific steps:
Acquiring an inverse proportion mapping value of the ratio of the number of the trend data sequence segments in the target cluster to the number of the trend data sequence segments in the reference cluster, and determining a trend rapid change index in a trend item corresponding to each trend window according to the trend change difference of the inverse proportion mapping value and the trend item corresponding to each trend window; and the inverse proportion mapping value is positively correlated with the trend change difference in the trend item corresponding to each trend window.
6. The method for optimizing energy efficiency for a data center according to claim 1, wherein the changing amplitude of each seasonal data sequence segment comprises the steps of:
the difference between the first and last season data in each of the sequence of season data segments is noted as the magnitude of change in each of the sequence of season data segments.
7. An energy efficiency optimization system for a data center comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the computer program when executed by the processor performs the steps of an energy efficiency optimization method for a data center as claimed in any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410707432.XA CN118297285B (en) | 2024-06-03 | 2024-06-03 | Energy efficiency optimization method and system for data center |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410707432.XA CN118297285B (en) | 2024-06-03 | 2024-06-03 | Energy efficiency optimization method and system for data center |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118297285A CN118297285A (en) | 2024-07-05 |
| CN118297285B true CN118297285B (en) | 2024-08-16 |
Family
ID=91678181
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410707432.XA Active CN118297285B (en) | 2024-06-03 | 2024-06-03 | Energy efficiency optimization method and system for data center |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118297285B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117539726A (en) * | 2024-01-09 | 2024-02-09 | 广东奥飞数据科技股份有限公司 | Energy efficiency optimization method and system for green intelligent computing center |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107005935B (en) * | 2014-12-22 | 2020-05-26 | 北京花旺在线商贸有限公司 | Network energy efficiency optimization method and device |
| CA3159151C (en) * | 2019-11-27 | 2023-02-07 | Robert F. Cruickshank Iii | An optimized load shaping system, method & apparatus for optimizing production and consumption of energy |
| CN114970358A (en) * | 2022-06-06 | 2022-08-30 | 深圳市证通电子股份有限公司 | Data center energy efficiency optimization method and system based on reinforcement learning |
| CN117271143B (en) * | 2023-11-22 | 2024-01-26 | 无锡尚航数据有限公司 | Data center optimization energy-saving scheduling method and system |
-
2024
- 2024-06-03 CN CN202410707432.XA patent/CN118297285B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117539726A (en) * | 2024-01-09 | 2024-02-09 | 广东奥飞数据科技股份有限公司 | Energy efficiency optimization method and system for green intelligent computing center |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118297285A (en) | 2024-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116089846A (en) | New energy settlement data anomaly detection and early warning method based on data clustering | |
| CN112215442B (en) | Method, system, device and medium for predicting short-term load of power system | |
| CN109376772B (en) | Power load combination prediction method based on neural network model | |
| CN111079989B (en) | DWT-PCA-LSTM-based water supply amount prediction device for water supply company | |
| CN113037531A (en) | Flow prediction method, device and storage medium | |
| CN102426674A (en) | Power system load prediction method based on Markov chain | |
| CN112801388A (en) | Power load prediction method and system based on nonlinear time series algorithm | |
| CN111754029A (en) | Community load prediction system | |
| CN115933787A (en) | Indoor multi-terminal intelligent control system based on indoor environment monitoring | |
| CN115962551B (en) | Intelligent air conditioner control system and method for building automatic control | |
| CN113807568B (en) | Power load prediction method and device and terminal equipment | |
| CN116073436B (en) | Capacity optimization control method for photovoltaic new energy power system | |
| CN109344990A (en) | A kind of short-term load forecasting method and system based on DFS and SVM feature selecting | |
| CN108334988A (en) | A kind of short-term Load Forecasting based on SVM | |
| CN115169746A (en) | Power load short-term prediction method and device based on fusion model and related medium | |
| CN118170004B (en) | Control method and system based on Internet of things | |
| CN115115107A (en) | Photovoltaic power prediction method and device and computer equipment | |
| CN113868953A (en) | Multi-unit operation optimization method, device and system in industrial system and storage medium | |
| CN111915211A (en) | Power resource scheduling method and device and electronic equipment | |
| CN116757337B (en) | House construction progress prediction system based on artificial intelligence | |
| CN118297285B (en) | Energy efficiency optimization method and system for data center | |
| CN117748496B (en) | Distributed photovoltaic energy storage regulation and control system and method based on intelligent prediction | |
| CN116963482A (en) | Intelligent energy-saving method and related equipment based on data center heating and ventilation system | |
| CN112066264A (en) | Pipeline hydraulic pressure regulation and control method, device and system | |
| US20200019858A1 (en) | Optimization of cyber-physical systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |