Data center precision air conditioner group control method

Abstract

The invention provides a precise air conditioner group control method for a data center, and the basic thought of the scheme is that temperature setting of a certain step length of each air conditioner of the data center is changed in turn, then temperature changes of all sensors in a machine room in corresponding time periods are obtained, and after temperature change data are vectorized, the temperature change data are stored in a database; a two-dimensional array of temperature changes and set temperatures of the air conditioning units is constructed, sensor temperature change similarities are obtained through calculation and sorted and graded, comprehensive membership vectors of the temperature change similarities of all levels are obtained, and comprehensive association evaluation indexes of the temperature sensors and the different air conditioning units are constructed based on the comprehensive membership vectors. And the influence correlation degree between the air conditioning unit and each sensor is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of data center air conditioning control, and particularly relates to a method for group control of precision air conditioners. Background Art

[0002] Currently, the energy-saving measures applied to computer room air conditioners include using outdoor natural cold sources and selecting highly efficient and energy-saving air

[0003] conditioning equipment, etc. These measures are all essential, but they are far from enough. The air conditioning refrigeration system is to ensure the temperature and humidity environment of the computer room. However, the cooling and humidity loads of the computer room are constantly changing with the operation of the equipment, the outdoor weather conditions, etc. If there is no intelligent control system to control the air conditioning system to follow these changes, it is impossible to ensure the constant temperature and humidity of the computer room, and at the same time, air conditioning energy saving cannot be guaranteed.

[0004] The existing technical solutions for air conditioner group control are basically pure theoretical models, that is, the air conditioner group control method, which is difficult to solve the following technical problems. First, calculate the cooling capacity demand in real time. Secondly, it is necessary to realize automatic cooling capacity supply according to the demand, and the environment needs to be within a safe and reasonable range, resulting in the problem that the air conditioner group control method cannot be specifically implemented. For this reason, we propose an air conditioner group control system applicable to the data center under maintenance. Summary of the Invention

[0005] Object of the Invention: Aiming at the problems existing in the prior art, the present invention provides a method for group control of precision air conditioners in a data center.

[0006] Summary of the Invention: To solve the above technical problems, the present invention adopts the following technical solutions: A method for group control of precision air conditioners in a data center includes the following steps: S1, adjust the return air temperature or supply air temperature of each air conditioner one by one in real time, and collect, record and statistically obtain the average temperature within 30 minutes before adjustment and the temperature within 30 minutes after adjustment , and use the temperature change values of the sensors before and after adjustment to construct a group control sample data set S with the adjusted temperature amplitude t respectively; S2, calculate the temperature change similarity d i of the sensor affected by the adjustment of the return air temperature or supply air temperature, and sort and grade according to the size of the temperature change similarity; S3, based on the temperature change similarity d i , establish an affected membership matrix μ of all temperature sensor temperature change indexes, and calculate the comprehensive membership vector L of each temperature sensor based on the entropy value method; S4, based on the comprehensive membership vector L of the temperature sensor, construct a comprehensive correlation evaluation index between each temperature sensor and different air conditioner units to complete the cluster control of the air conditioner units.

[0007] Furthermore, the construction steps of the group control sample data set S described in step S1 include: (1) For the i-th air conditioner unit, by adjusting its return air temperature and supply air temperature respectively, the temperature adjustment amplitude of the return air temperature and the supply air temperature is t, and record the average temperature of each temperature sensor in the data center within 30 minutes before the adjustment and the temperature within 30 minutes after the adjustment , so as to obtain the temperature change before and after the adjustment , ; (2) Construct a two-dimensional array with the temperature change value of the sensor before and after the adjustment and the temperature adjustment amplitude t, and establish the real-time sample data S i , of the i-th air conditioner unit, where the subscript i represents the number of the air conditioner unit, K represents the total number of temperature sensors in the data room, and the subscript j represents the number of the temperature sensor; (3) Then, complete the construction of the real-time group control sample data S of all air conditioner units one by one, , where N represents the total number of air conditioner units in the data room.

[0008] Furthermore, the clustering algorithm described in step S2 includes the following steps: (1) Considering the uncertainty of the temperature change in the computer room, taking the similarity d (instead of the conventional absolute value of the change) of the temperature change value ij collected within 30 minutes after the temperature adjustment as the evaluation basis, calculate the temperature change similarity d of the air conditioner unit under different temperature adjustment amplitudes t through the following formula ij calculated through the following formula, , In the formula, d ij is the temperature change similarity between two points; Sim ij is the cosine value of the included angle between two vectors A and B, and its value range is between -1 and 1; A ij and B ij are the respective dimensional components of vectors A and B, corresponding to the average temperature change in the sample data; K and N are the number of dimensional components owned by vectors A and B, that is, the number of temperature sensors and air conditioner units in the data center; (2) Statistically analyze the temperature change similarity d i index values of each sensor of the air conditioner unit under different temperature adjustment amplitudes t and sort them according to the size, and classify them into ranks R with percentile ranks of (0, 30%), [30%, 50%), [50%, 70%), [70%, 90%), (90%, 100%), and record them as the temperature change level R, 。

[0009] Furthermore, the establishment of the membership matrix μ in step S3 includes the following steps: (1) Respectively select the temperature change similarity of the sensor at the median of the temperature change level R as a reference to obtain a 5-level membership reference vector U. , (2) For the membership value matrix μ of the temperature change similarity of each sensor under different temperature adjustment ranges t of the air conditioner unit ij , it is obtained through the following formula. When the temperature change level r = 5, , When the temperature change level r = 2, 3, or 4, , When the temperature change level r = 1, , In the formula, μ ij represents the membership of the temperature change similarity of the temperature sensor numbered j when the air conditioner unit numbered i adjusts the temperature, and the value range is 0~1. (3) Consider the influence of the return air temperature and the supply air temperature, and based on the entropy method, calculate the weight vectors W ij,pre and W ij,s , and then calculate the comprehensive membership l ij of the sensor i affected by the air conditioner unit j through fuzzy transformation. , In the formula, μ ij,re and μ ij,s respectively represent the membership of the temperature change similarity under the adjustment of the return air temperature and the supply air temperature, and W 1ij and W 2ij are the weight coefficients of the return air temperature and the supply air temperature respectively. Continue to complete the calculation of the comprehensive membership of all other sensors affected by the air conditioner unit j, and classify the comprehensive membership l ij according to the temperature change level r to obtain the classified comprehensive membership vector L rij . .

[0010] Furthermore, the comprehensive correlation evaluation index model of each temperature sensor and the air conditioner unit j is as follows: Based on the comprehensive membership vector L rj of the temperature sensor, construct the comprehensive correlation evaluation index Rate j of each temperature sensor and different air conditioner units to complete the cluster control of the air conditioner unit. , In the formula, Rate j represents the associated evaluation index of the temperature sensor affected by the air-conditioning unit j, r represents the temperature change level, p represents the percentile rank under the temperature change level where the sensor numbered i is located, and the value range is (0.1, 0.3, 0.5, 0.7, 0.9), and L rij represents the comprehensive membership degree vector after classification.

[0011] Furthermore, the value of the adjusted temperature amplitude t described in step S1 is 1°C or 3°C.

[0012] Beneficial effects: Compared with the prior art, the present invention constructs sample data of temperature changes in the data center based on the adjustment of the temperature amplitude, and establishes the relationship of the affected membership degree matrix of the regional temperature change based on the sorting and grading of the similarity of the regional temperature changes, completes the evaluation of the non-linear complex correlation between the regional temperature change and the air-conditioning unit, and realizes the cluster control of the air-conditioning unit according to the real-time cooling demand. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic diagram of the logic flow of the precision air-conditioning group control method for the data center described in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0014] The present invention will be further clarified below with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. After reading the present invention, those skilled in the art will fall within the scope defined by the appended claims of this application for various equivalent modifications of the present invention.

[0015] As Figure 1 shown, the present invention provides a precision air-conditioning group control method for a data center. The basic idea of this solution is to alternately change the temperature setting of each air-conditioning unit in the data center by a certain step, and then obtain the temperature changes of all sensors in the computer room during the corresponding period. After vectorizing the temperature change data, a two-dimensional array of temperature changes and the set temperature of the air-conditioning unit is constructed. By calculating the similarity of the sensor temperature changes and sorting and grading, the comprehensive membership degree vector of the similarity of each level of temperature change is obtained, and the comprehensive associated evaluation index of each temperature sensor and different air-conditioning units is constructed based on the comprehensive membership degree vector to realize the influence correlation degree between the air-conditioning unit and each sensor. Its main logical process includes the following steps: S1. First, numbers are set for the sensors and air-conditioning units. The subscript i represents the number of the air-conditioning unit, and the number of air-conditioning units is K; the subscript j represents the number of the temperature sensor, and the number of sensors is N.

[0016] Adjust the return air temperature or supply air temperature of the air conditioner unit one by one in real time at intervals of t = 1°C, 2°C, and 3°C, and collect the average temperature of each sensor within 30 minutes before the large adjustment for statistics. At the same time, collect the real-time change of the sensor temperature within 30 minutes after the adjustment. The collection frequency is once every 3 minutes.

[0017] Then, calculate the temperature change before and after the adjustment. , , and use and t as the corresponding relationship to construct a two-dimensional array. Complete the establishment of the real-time sample data S of the i-th air conditioner unit. i ; ; Next, complete the construction of the real-time group control sample data S of all the remaining air conditioner units one by one. , where N represents the total number of air conditioner units in the data room.

[0018] Since the non-linear complex situation of the temperature series influence in the enclosed environmental space of the data center is considered, the method of the present invention also considers the influence of the return air temperature and supply air temperature settings, and completes the data collection and processing according to the above method.

[0019] S2. The present invention considers the randomness and uncertainty of temperature changes. Compared with the conventional method of directly using the temperature change value, the present invention uses the temperature change similarity d i , which solves the problem that the traditional measurement using the single-point distance difference cannot better reflect the similarity of temperature fluctuations over time, and thus has the advantage of higher calculation efficiency. Specifically as follows: Take the similarity d of the temperature change values collected within 30 minutes after temperature adjustment ij (instead of the conventional absolute change value) as the evaluation basis, and calculate the temperature change similarity d of the air conditioner unit under different temperature adjustment amplitudes t through the following formula. ij Calculated through the following formula: , In the formula, d ij is the temperature change similarity between two points; Sim ij is the cosine value of the included angle between two vectors A and B, and its value range is between -1 and 1; A ij and B ij are the respective dimensional components of vectors A and B, corresponding to the average temperature change in the sample data. ; K and N are the number of dimensional components of vectors A and B respectively, that is, the number of temperature sensors and air conditioner units in the data center; Next, the temperature change similarity d of each sensor under different temperature adjustment ranges t of the air conditioner unit is statistically analyzed respectively i The index values are sorted according to their magnitudes. In the present invention, the sorting is divided into five levels according to the digit percentage, that is, (0, 30%), [30%, 50%), [50%, 70%), [70%, 90%), (90%, 100%), and is denoted as the temperature change level R .

[0020] S3. After determining the temperature change similarity d i , an affected membership matrix μ of the temperature change indexes of all temperature sensors is established, and the comprehensive membership vector L of each temperature sensor is calculated based on the entropy value method First, the temperature change similarities of the sensors at the median of the temperature change level R are selected as references respectively to obtain a 5-level membership reference vector U , Then, for different temperature adjustment ranges t of the air conditioner unit, the membership value matrix μ of the temperature change similarity of each sensor ij is obtained through the following formula. When the temperature change level R = 5 , When the temperature change level R = 2, 3, or 4 , When the temperature change level R = 1 , In the formula, μ ij represents the membership of the temperature change similarity of the temperature sensor numbered j when the air conditioner unit numbered i undergoes temperature adjustment, and the value range is 0 to 1; i - K represents the number of the air conditioner unit; j - N represents the number of the temperature sensor

[0021] Next, for the membership of the temperature change similarity under the adjustment of the return air temperature and supply air temperature of the air conditioner unit, and based on the entropy value method, the corresponding weight vector W is calculated, and the comprehensive membership vector L under each temperature change level is obtained through fuzzy transformation operation. The fuzzy membership calculation method takes into account the weights under different setting conditions of the return air temperature and supply air temperature and the fuzziness of the actual situation, and has a more accurate performance and reflects the complexity of the specific membership relationship of the affected sensors, thereby improving the accuracy and reliability of the calculation

[0022] , In the formula, μ ij,re and μ ij,s respectively represent the membership of the temperature change similarity under the adjustment of the return air temperature and supply air temperature, and W 1ij and W2ij The weight coefficients of the return air temperature and the supply air temperature respectively; Continue to complete the calculation of the comprehensive membership degrees of all other sensors affected by the air conditioning unit j, and classify the comprehensive membership degree l according to the temperature change level r ij to obtain the comprehensive membership degree vector L after classification rij , .

[0023] Step S4, the purpose of this step is to construct the comprehensive correlation evaluation index Rate of each temperature sensor affected by the air conditioning unit j based on the comprehensive membership degree vector L of the temperature sensors rj . j

[0024] , In the formula, Rate j represents the correlation evaluation index of the temperature sensor affected by the air conditioning unit j, r represents the temperature change level, p represents the ranking percentile under the temperature change level where the sensor numbered i is located, and the value range is (0.1, 0.3, 0.5, 0.7, 0.9), and L rij represents the comprehensive membership degree vector after classification.

[0025] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.​

Claims

1. A method for group control of precision air conditioners in a data center, characterized in that, It includes the following steps: S1. Adjust the return air temperature or supply air temperature of each air conditioner one by one in real time, and collect, record and statistically obtain the average temperature within 30 minutes before adjustment. And the temperature within 30 minutes after adjustment , taking the temperature change values of the sensors before and after adjustment Respectively construct a group control sample data set S with the adjustment temperature amplitude t; S2, calculate the temperature change similarity d after the sensor is affected by the return air temperature or supply air temperature adjustment i , and sort and classify according to the magnitude of the temperature change similarity; S3. Based on the temperature change similarity d i , establish the affected membership matrix μ of the temperature change indicators of all temperature sensors, and calculate the comprehensive membership vector L of each temperature sensor based on the entropy value method; S4. Based on the comprehensive membership degree vector L of the temperature sensors, construct the comprehensive correlation evaluation indexes of each temperature sensor and different air-conditioning units, and complete the cluster control of the air-conditioning units.

2. The data center precision air conditioner group control method according to claim 1, characterized in that, The construction steps of the group control sample data set S described in step S1 include: (1)For the i-th air conditioner unit, by separately adjusting its return air and supply air temperatures, with the temperature adjustment amplitude of t for both the return air and supply air, and recording the average temperatures of each temperature sensor in the data center within 30 minutes before the adjustment and the temperatures within 30 minutes after the adjustment , so as to obtain the temperature changes before and after the adjustment , ; (2) Using the temperature change values of the sensors before and after adjustment and the temperature adjustment amplitude t to construct a two-dimensional array , establishing the real-time sample data S of the i-th air conditioning unit i , , where the subscript i represents the number of the air conditioning unit, K represents the total number of temperature sensors in the data room, and the subscript j represents the number of the temperature sensor; (3) Next, complete the construction of the real-time group control sample data S for all air-conditioning units one by one, , where N represents the total number of air-conditioning units in the data room.

3. The method for group control of precision air conditioners in a data center according to claim 1, wherein The clustering algorithm described in step S2 includes the following steps: (1) Considering the uncertainty of the temperature change in the computer room, use the similarity d of the temperature change values collected within 30 minutes after temperature adjustment ij (instead of the conventional absolute value of change) as the evaluation basis, and calculate the temperature change similarity d at different temperature adjustment amplitudes t of the air conditioner unit through the following formula ij It is calculated through the following formula ， where d ij is the temperature change similarity between two points; Sim ij is the cosine value of the angle between two vectors A and B, and its value range is between -1 and 1; A ij and B ij are the respective dimensional components of vectors A and B, corresponding to the average temperature change in the sample data ; K and N are the numbers of dimensional components of vectors A and B respectively, that is, the numbers of temperature sensors and air conditioning units in the data center; (2)Statistically analyze the temperature change similarity d of each sensor under different temperature adjustment ranges t of the air conditioner unit i for the index values, sort them by size, and classify them into ranking percentiles R of (0, 30%), [30%, 50%), [50%, 70%), [70%, 90%), (90%, 100%), and record them as the temperature change level R. .

4. The method for centralized control of precision air conditioners in a data center according to claim 3, wherein The establishment of the membership degree matrix μ described in step S3 includes the following steps: (1) Select the temperature change similarity of the sensor at the median of the temperature change level R as a reference respectively to obtain a 5-level membership reference vector U. , (2)Membership value matrix μ of the temperature change similarity of each sensor under different temperature adjustment ranges t of the air conditioning unit ij , obtained by the following formula. When the temperature change level r = 5, ， When the temperature change level r = 2, 3 or 4, ， When the temperature change level r = 1, ， where μ ij represents the membership degree of the temperature change similarity of the temperature sensor numbered j when the air conditioner unit numbered i adjusts its temperature, and the value range is 0 to 1; (3) Consider the influence of the return air temperature and the supply air temperature, and calculate the weight vectors W for adjusting the return air temperature and the supply air temperature based on the entropy method. ij,pre and W ij,s Then, through fuzzy transformation, calculate the comprehensive membership degree l of sensor i affected by air conditioning unit j. ij , ， where μ ij,re and μ ij,s represent the membership degrees of temperature change similarity under the regulation of return air temperature and supply air temperature respectively, and W 1ij and W 2ij are the weight coefficients of return air temperature and supply air temperature respectively; Continue to complete the calculation of the comprehensive membership degrees of all other sensors affected by the air-conditioning unit j, and classify the comprehensive membership degree l according to the temperature change level r ij to obtain the comprehensive membership degree vector L after classification rij , .

5. The method for group control of precision air conditioners in a data center according to claim 3, wherein The comprehensive correlation evaluation index model of each temperature sensor and air conditioner unit j in step S4 is as follows: Based on the comprehensive membership degree vector L of the temperature sensor rj , construct the comprehensive correlation evaluation index Rate of each temperature sensor and different air conditioner units j , and complete the cluster control of the air conditioner units ， where Rate j represents the associated evaluation index of the temperature sensor affected by the air conditioner unit j, r represents the temperature change level, p represents the percentile ranking of the sensor numbered i under the temperature change level, and the value range is (0.1, 0.3, 0.5, 0.7, 0.9), L rij represents the comprehensive membership degree vector after grading.

6. The method for group control of precision air conditioners in a data center according to claim 3, wherein The value of the adjusted temperature amplitude t described in step S1 is 1°C or 3°C.

Images