CN111210060A - Method for predicting temperature of machine room during working day - Google Patents
Method for predicting temperature of machine room during working day Download PDFInfo
- Publication number
- CN111210060A CN111210060A CN201911395569.1A CN201911395569A CN111210060A CN 111210060 A CN111210060 A CN 111210060A CN 201911395569 A CN201911395569 A CN 201911395569A CN 111210060 A CN111210060 A CN 111210060A
- Authority
- CN
- China
- Prior art keywords
- data
- temperature
- working day
- machine room
- algorithm
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
The invention relates to a method for predicting the temperature of a machine room during a working day, which only considers the temperature data of the machine room acquired during the working day, fits the head and tail data by using a data fitting algorithm between two adjacent working days, replaces part of data at a splicing part with the fitted data so as to reduce the influence caused by the sudden change of the temperature data at the splicing part, and predicts the temperature value of the machine room during the working day by using new temperature time series data. According to the invention, only the temperature data of the machine room during the working day is considered, the interference of the temperature data during the holiday period on the temperature prediction during the working day is eliminated, the data quantity required by the prediction algorithm is less, and the prediction precision is higher.
Description
Technical Field
The invention belongs to the technical field, and particularly relates to a method for predicting the temperature of a machine room during a working day.
Background
With the rapid development of big data and cloud computing technology, more and more data centers are built at home and abroad. In order to ensure the stable operation of various IT devices in a data center machine room, the temperature and humidity environmental conditions of the machine room need to be strictly controlled. When the data center machine room works, due to thermal coupling generated by high density of electronic components, the temperature of the machine room can be increased, and the working efficiency and the service life of the components can be influenced by overhigh temperature, so that the data center usually adopts all-day high-strength air conditioning refrigeration to ensure that the machine room is in a proper temperature state.
The control strategy of most current machine room air conditioners is only to simply set the temperature of the air conditioner, and the control strategy of the air conditioner is too extensive. In the actual operation process, the heat dissipation capacity of IT equipment in a machine room is time-varying, and in addition, a machine room air conditioner control system has great hysteresis, so that the air conditioner refrigeration efficiency is not high, energy waste is caused, and the operation cost of a data center is increased. Therefore, it is necessary to predict the load or temperature of the machine room to reasonably adjust the operating parameters of the air conditioner, so as to achieve a better energy-saving effect.
The patent publication No. 109189190a provides a data center thermal management method based on temperature prediction, which trains an artificial neural network prediction model by using server node real-time operation data, wireless sensor data and CFD simulation data to predict the inlet temperature of a data center server, thereby realizing a data center temperature prediction algorithm based on a neural network.
The patent with publication number 109375994A provides a data center task temperature prediction and scheduling method based on an RBF neural network, and a task temperature prediction model is established by adopting the RBF neural network; the model inputs the temperature, the operation parameters and the environment temperature of all the servers in the collected area through real-time monitoring, and outputs the temperature of the air inlet of the server after operation and the predicted time step length.
Publication 102436296A provides a system and method for predicting temperature values in a data center by determining inlet and outlet temperatures of each of the racks and at least one cooling provider including establishing a set of S-coupling equations, S being equal to the number of temperature values to be determined, and solving the S-coupling equations.
In the existing special data centers for most enterprises, the access amount of IT equipment during working days is large, the fluctuation of machine room load is large, the change of machine room temperature during working days is large, and the temperature prediction is difficult; the access amount of the IT equipment during holidays is small, the load is stable, the temperature change of the machine room is small, and the temperature prediction is simple. Therefore, the importance of the temperature prediction of the machine room during the working day is generally greater than that during the holiday, and technical means are needed to improve the accuracy of the temperature prediction of the machine room during the working day.
According to the prior art, the difference of the load and temperature change of a machine room between a working day and a holiday is ignored, the temperature data of the working day and the holiday are not distinguished, and the collected temperature time sequence data are directly utilized for temperature prediction. In this case, the temperature data during holidays rather becomes a disturbance, which affects the accuracy of temperature prediction during workdays. Although the data during holidays can be deleted from the collected temperature time-series data, the head-to-tail data of the temperature time-series data during a plurality of working days cannot be smoothly connected, and the accuracy of temperature prediction is greatly influenced.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a method for predicting the temperature of a machine room during a working day, which removes the temperature data of the machine room collected during a holiday, only considers the temperature data of the machine room collected during the working day, thereby eliminating the interference of the temperature data during the holiday on the temperature prediction during the working day, and realizes the smooth connection of time series data collected in different working day time periods through a head-tail data fitting technology during the working day, so that the temperature prediction precision of the machine room during the working day is higher.
In order to achieve the purpose, the invention adopts the technical scheme that:
a method for predicting the temperature of a machine room during a working day comprises the following steps:
(a) temperature time series data T of collection computer lab
T＝{TW1,TE1,TW2,TE2…,TWi,TEi…,TWs,TEs|s∈N+}
Wherein TWiAnd TEiTemperature time-series data respectively representing the working day and non-working day of the ith week, s represents the number of weeks, N+Represents a positive integer, i is more than or equal to 1 and less than or equal to s;
(b) deleting the temperature time-series data of the holidays from T, and remaining the temperatureThe time-series data constitute a set TN, { TN ═ TW1,TW2,…,TWi,…,TWs|s∈N+}
Wherein TWi＝{Twi,1,…,Twi,j,…,Twi,k|j,k∈N+}，Twi,jTemperature time-series data indicating the jth working day of the ith week, wherein k is the serial number of the last working day of the ith week;
Twi,j＝{twi,j,1,twi,j,2,…,twi,j,m,…,twi,j,n|m,n∈N+}，twi,j,mthe mth temperature data of the jth working day of the ith week is shown, and n is the serial number of the last temperature data of the jth working day of the ith week;
(c) adjacent TW in a pair-by-pair TNiAnd TWi+1The following operations are completed:
(1) from TWiGet Twi,kH last data of TW from TWi+1Get Tw(i+1),1The first h data of (2) constitute temperature time-series data of the total data;
(2) fitting the 2h data by adopting a data fitting algorithm to obtain 2h new time sequence data; the data fitting algorithm can select a polynomial fitting algorithm or a nonlinear least square fitting algorithm or a neural network algorithm;
(3) selecting s data from the 2h new time sequence data to replace Twi,kAnd Tw(i+1),1Data of the corresponding position;
(d) and predicting the temperature value of the machine room by adopting a time series prediction algorithm based on the temperature time series data in the new TN set, wherein the time series prediction algorithm can select a moving average algorithm or an exponential smoothing algorithm or a difference integration moving average autoregressive algorithm or a long-term memory network algorithm.
Compared with the prior art, in the aspect of machine room temperature prediction during working days, the invention only needs to consider the machine room temperature time series data collected during the working days, and can eliminate the interference of holiday data on the working day temperature prediction. When the temperature is predicted by using the common LSTM (long-short time memory network) algorithm or MA (moving average) algorithm, the required data volume is greatly reduced, and the prediction precision is higher
Drawings
Fig. 1 is a schematic diagram of fitting head and tail data between two adjacent working days.
FIG. 2 is an algorithm flow of a differential integrated moving average autoregressive model (ARIMA).
Detailed Description
The embodiments of the present invention will be described in detail below with reference to the drawings and examples.
Example 1
A method for predicting the temperature of a machine room during a working day comprises the following steps:
step 1), collecting temperature time sequence data T of a machine room, wherein the temperature time sequence data T is as follows:
T＝{TW1,TE1,TW2,TE2…,TWi,TEi…,TWs,TEs|s∈N+}
wherein TWiAnd TEiAnd (c) temperature time-series data respectively representing the working day and the non-working day of the ith week, wherein i is more than or equal to 1 and less than or equal to s.
And 2), deleting the temperature time-series data of the holidays from the T, wherein the rest temperature time-series data form a set TN in the form of:
TN＝{TW1,TW2,…,TWi,…,TWs|s∈N+}
wherein TWi＝{Twi,1,…,Twi,j,…,Twi,k|j,k∈N+}，Twi,jTemperature time-series data indicating the jth working day of the ith week, wherein k is the serial number of the last working day of the ith week;
Twi,j＝{twi,j,1,twi,j,2,…,twi,j,m,…,twi,j,n|m,n∈N+}，twi,j,mthe "n" is the number of the last temperature data of the jth working day on the ith week.
Examples are as follows:
temperature time series data in degrees celsius for weekdays during a location 2018/5/21-2018/5/28 of a data center facility are collected at 5 minute intervals, with the first row of numbers 1-288 representing 288 data collected during the day. 2018/5/21 on the ith week and 2018/5/28 on the ith +1 week, the temperature time series data of holidays were deleted and the data obtained are shown in Table 1-1:
TABLE 1-1 temperature time series data during working day at certain location of machine room
The temperature data of a certain position of the data center machine room can be the temperature value of a temperature sensor at the air inlet of the machine cabinet or the average value of the temperature values of a plurality of temperature sensors in the channel of the machine cabinet. The data in table 1-1 is an average value of actually acquired temperature values of a plurality of temperature sensors in a certain cabinet channel.
Step 3), adjacent TWs in a pair of TN one by oneiAnd TWi+1The following operations are completed:
(1) from TWiGet Twi,kH last data of TW from TWi+1Get Tw(i+1),1The first h data of (2) data constitute temperature time-series data of the total.
(2) And fitting the 2h data by adopting a polynomial fitting algorithm to obtain 2h new time sequence data.
Taking the data in table 1-1 as an example, take h as 6, Twi,5The last h data of (1) is { twi,5,283,twi,5,284,twi,5,285,twi,5,286,twi,5,287,twi,5,288{25.77,25.74,25.74,25.71,25.74,25.74 }; tw(i+1),1The first h data of (1) are { twi+1,1,1,twi+1,1,2,twi+1,1,3,twi+1,1,4,twi+1,1,5,twi+1,1,6I.e., {24.67,24.69,24.80,24.86,24.94,25.02 }.
For time series data:
{twi,5,283,twi,5,284,twi,5,285,twi,5,286,twi,5,287,twi,5,288,twi+1,1,1,twi+1,1,2,twi+1,1,3,twi+1,1,4,twi+1,1,5,twi+1,1,6}
polynomial fitting was performed to obtain fitting values shown in tables 1-2.
Table 1-2 polynomial fitting results obtained when h is 6
| Serial number | 283 | 284 | 285 | 286 | 287 | 288 | 1 | 2 | 3 | 4 | 5 | 6 |
| Actual value | 25.77 | 25.74 | 25.74 | 25.71 | 25.74 | 25.74 | 24.67 | 24.69 | 24.80 | 24.86 | 24.94 | 25.02 |
| Fitting value | 25.76 | 25.78 | 25.65 | 25.80 | 25.82 | 25.46 | 24.93 | 24.63 | 24.71 | 24.95 | 24.91 | 25.02 |
Fig. 1 shows the result of polynomial fitting on the actual values. Wherein, the points of the triangle are actual values, and the points of the cross star are fitting values; the first 6 points are Twi,5The last 6 data, the last 6 points are Tw(i+1),1The first 6 data.
(3) Selecting s data from the 2h new time sequence data to replace Twi,kAnd Tw(i+1),1Of the corresponding location.
When the fitting value is used for replacing the actual value, only the middle part of data is replaced, and the purpose is to enable the temperature time sequence data to keep the original variation trend as much as possible and avoid the great change of the original temperature data variation trend caused by data replacement.
Taking the data in table 1-2 as an example, taking s as 6 can obtain the partially replaced data shown in table 1-3:
tables 1-3 results of partial substitution when s is 6
| 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | |
| Actual value | 25.77 | 25.74 | 25.74 | 25.71 | 25.74 | 25.74 | 24.67 | 24.69 | 24.80 | 24.86 | 24.94 | 25.02 |
| Replacement value | 25.77 | 25.74 | 25.74 | 25.80 | 25.82 | 25.46 | 24.93 | 24.63 | 24.71 | 24.86 | 24.94 | 25.02 |
And 4) predicting the temperature value of the machine room by adopting a time series prediction algorithm based on the temperature time series data in the new TN set.
In this embodiment, a difference integrated moving average autoregressive model (ARIMA) is used to predict the temperature value of the machine room. The algorithm flow of the difference integration moving average autoregressive model is shown in fig. 2. The temperature time sequence data of the first 4 working days can be selected as a training set, and the temperature time sequence data of the last 2 working days can be used as a testing set for prediction and testing the prediction precision.
In summary, the invention relates to a method for predicting the temperature of a machine room during a working day, which only considers the temperature data of the machine room collected during the working day, fits the head and tail data by using a data fitting algorithm between two adjacent working days, replaces part of data at a splicing part with the fitted data so as to reduce the influence caused by the sudden change of the temperature data at the splicing part, and predicts the temperature value of the machine room during the working day by using new temperature time series data. According to the invention, only the temperature data of the machine room during the working day is considered, the interference of the temperature data during the holiday period on the temperature prediction during the working day is eliminated, the data quantity required by the prediction algorithm is less, and the prediction precision is higher.
Claims (3)
1. A method for predicting the temperature of a machine room during a working day is characterized by comprising the following steps:
(a) temperature time series data T of collection computer lab
T＝{TW1,TE1,TW2,TE2…,TWi,TEi…,TWs,TEs|s∈N+}
Wherein TWiAnd TEiTemperature time-series data respectively representing the working day and non-working day of the ith week, s represents the number of weeks, N+Represents a positive integer, i is more than or equal to 1 and less than or equal to s;
(b) the temperature time-series data of holidays are deleted from T, and the remaining temperature time-series data form a set TN, TN ═ TW1,TW2,…,TWi,…,TWs|s∈N+}
Wherein TWi＝{Twi,1,…,Twi,j,…,Twi,k|j,k∈N+}，Twi,jIndicates the temperature of the jth working day of the ith weekTime series data, k is the serial number of the last working day of the ith week;
Twi,j＝{twi,j,1,twi,j,2,…,twi,j,m,…,twi,j,n|m,n∈N+}，twi,j,mthe mth temperature data of the jth working day of the ith week is shown, and n is the serial number of the last temperature data of the jth working day of the ith week;
(c) adjacent TW in a pair-by-pair TNiAnd TWi+1The following operations are completed:
(1) from TWiGet Twi,kH last data of TW from TWi+1Get Tw(i+1),1The first h data of (2) constitute temperature time-series data of the total data;
(2) fitting the 2h data by adopting a data fitting algorithm to obtain 2h new time sequence data;
(3) selecting s data from the 2h new time sequence data to replace Twi,kAnd Tw(i+1),1Data of the corresponding position;
(d) and predicting the temperature value of the machine room by adopting a time series prediction algorithm based on the temperature time series data in the new TN set.
2. The method for predicting the temperature of the machine room during the working day according to claim 1, wherein the data fitting algorithm in the step (2) is a polynomial fitting algorithm or a nonlinear least squares fitting algorithm or a neural network algorithm.
3. The method for predicting the temperature of the machine room during the working day according to claim 1, wherein the time series prediction algorithm in the step (d) is a moving average algorithm, an exponential smoothing algorithm, a difference integration moving average autoregressive algorithm or a long-term memory network algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911395569.1A CN111210060B (en) | 2019-12-30 | 2019-12-30 | Method for predicting temperature of machine room during working days |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911395569.1A CN111210060B (en) | 2019-12-30 | 2019-12-30 | Method for predicting temperature of machine room during working days |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111210060A true CN111210060A (en) | 2020-05-29 |
| CN111210060B CN111210060B (en) | 2023-04-28 |
Family
ID=70789375
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911395569.1A Active CN111210060B (en) | 2019-12-30 | 2019-12-30 | Method for predicting temperature of machine room during working days |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111210060B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112612305A (en) * | 2020-12-04 | 2021-04-06 | 格力电器（武汉）有限公司 | Temperature adjusting method, device, equipment, storage medium and air conditioning system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030021473A1 (en) * | 2001-07-26 | 2003-01-30 | Sean Chiu | Fast gamma correction method for image reading apparatus |
| KR20140116619A (en) * | 2013-03-25 | 2014-10-06 | 삼성에스디에스 주식회사 | System and method for predicting energy consumption |
| CN104156783A (en) * | 2014-07-29 | 2014-11-19 | 广西电网有限责任公司 | Maximum daily load prediction system and method of electric system considering meteorological accumulative effect |
| CN106874676A (en) * | 2017-02-20 | 2017-06-20 | 广东工业大学 | A kind of electric power meter state evaluating method |
| CN107463633A (en) * | 2017-07-17 | 2017-12-12 | 中国航天系统科学与工程研究院 | A kind of real time data rejecting outliers method based on EEMD neutral nets |
| CN108518804A (en) * | 2018-03-21 | 2018-09-11 | 武汉物联远科技有限公司 | A kind of machine room humiture environmental forecasting method and system |
| CN109117996A (en) * | 2018-08-01 | 2019-01-01 | 淮安市农业信息中心 | The method for constructing greenhouse winter temperature prediction model |
-
2019
- 2019-12-30 CN CN201911395569.1A patent/CN111210060B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030021473A1 (en) * | 2001-07-26 | 2003-01-30 | Sean Chiu | Fast gamma correction method for image reading apparatus |
| KR20140116619A (en) * | 2013-03-25 | 2014-10-06 | 삼성에스디에스 주식회사 | System and method for predicting energy consumption |
| CN104156783A (en) * | 2014-07-29 | 2014-11-19 | 广西电网有限责任公司 | Maximum daily load prediction system and method of electric system considering meteorological accumulative effect |
| CN106874676A (en) * | 2017-02-20 | 2017-06-20 | 广东工业大学 | A kind of electric power meter state evaluating method |
| CN107463633A (en) * | 2017-07-17 | 2017-12-12 | 中国航天系统科学与工程研究院 | A kind of real time data rejecting outliers method based on EEMD neutral nets |
| CN108518804A (en) * | 2018-03-21 | 2018-09-11 | 武汉物联远科技有限公司 | A kind of machine room humiture environmental forecasting method and system |
| CN109117996A (en) * | 2018-08-01 | 2019-01-01 | 淮安市农业信息中心 | The method for constructing greenhouse winter temperature prediction model |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112612305A (en) * | 2020-12-04 | 2021-04-06 | 格力电器（武汉）有限公司 | Temperature adjusting method, device, equipment, storage medium and air conditioning system |
| CN112612305B (en) * | 2020-12-04 | 2022-04-08 | 格力电器（武汉）有限公司 | Temperature adjusting method, device, equipment, storage medium and air conditioning system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111210060B (en) | 2023-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11409274B2 (en) | Model predictive maintenance system for performing maintenance as soon as economically viable | |
| CN109800066B (en) | Energy-saving scheduling method and system for data center | |
| US11487277B2 (en) | Model predictive maintenance system for building equipment | |
| US20200356087A1 (en) | Model predictive maintenance system with event or condition based performance | |
| Li et al. | A multi-agent based distributed approach for optimal control of multi-zone ventilation systems considering indoor air quality and energy use | |
| KR101973652B1 (en) | Variable air volume modeling for HVAC systems | |
| JP5511698B2 (en) | Air conditioner linkage control system, air conditioner linkage control method, and air conditioner linkage control program | |
| JP5801732B2 (en) | Operation management method of information processing system | |
| US11497144B2 (en) | Optimized thermal control of data center | |
| US8155793B2 (en) | System and method for controlling air conditioning facilities, and system and method for power management of computer room | |
| CN112616292B (en) | Data center energy efficiency optimization control method based on neural network model | |
| WO2021108224A1 (en) | Building control system with smart edge devices having embedded model predictive control | |
| US11409250B2 (en) | Adaptively learning surrogate model for predicting building system dynamics from system identification model | |
| Wang et al. | Event-driven optimal control of central air-conditioning systems: Event-space establishment | |
| CN103513632A (en) | Energy management system | |
| Bhattacharya et al. | Control co-design of commercial building chiller plant using Bayesian optimization | |
| Hussain et al. | Adaptive regression model-based real-time optimal control of central air-conditioning systems | |
| CN114970358A (en) | Data center energy efficiency optimization method and system based on reinforcement learning | |
| Hou et al. | Real-time optimal control of HVAC systems: Model accuracy and optimization reward | |
| CN111210060B (en) | Method for predicting temperature of machine room during working days | |
| CN113225994B (en) | Intelligent air conditioner control method facing data center | |
| JP6455937B2 (en) | Simulation apparatus, simulation method, and program | |
| CN111083201B (en) | Energy-saving resource allocation method for data-driven manufacturing service in industrial Internet of things | |
| Tashiro et al. | A network model for prediction of temperature distribution in data centers | |
| Kumar et al. | Power usage efficiency (PUE) optimization with counterpointing machine learning techniques for data center temperatures |
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 |