Multi-input multi-output data center energy efficiency model based on fuzzy network

Abstract

The invention discloses a multi-input multi-output data center energy efficiency model based on a fuzzy network. The method comprises the following steps: acquiring operation data of IT equipment and air conditioning equipment; processing the collected operation data into a data set which can be used by an energy efficiency model; building a multi-input multi-output data center energy efficiency model based on the fuzzy network; the energy efficiency model uses a fuzzy neural network to predict the air supply temperature and the air supply volume of control parameters, and further optimizes the control parameters by means of a multi-objective optimization algorithm; dividing the data set into a training set and a test set, training the designed energy efficiency model by using the training set, and verifying a training result through the test set; the optimal combination of the control parameters is obtained through the energy efficiency model, and finally the optimal control parameters are sent to the air conditioner equipment, so that optimal control over the air conditioner of the data center is achieved. According to the energy efficiency model provided by the invention, the internal relation among multiple inputs can be captured, an air conditioner control scheme is scientifically and reasonably formulated, and energy conservation of the air conditioner is realized while the temperature of a machine room is guaranteed.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence technology and multi-objective optimization, and in particular to a data center energy efficiency model based on fuzzy networks with multiple inputs and multiple outputs. Background Technology

[0002] As a critical infrastructure for next-generation information technologies such as artificial intelligence and cloud computing, data centers are continuously expanding in scale and their importance is becoming increasingly prominent. However, this growth in scale has also led to a sharp increase in energy consumption. Data centers not only need to ensure the stable operation of IT equipment, but also rely on cooling equipment to maintain a suitable ambient temperature. How to reduce energy consumption and improve energy efficiency has become a pressing problem to be solved in the data center field. Traditional data centers have exposed many limitations when facing complex and ever-changing operating environments. The operation of data centers involves numerous variables, such as the energy consumption of IT equipment and the operating parameters of air conditioning equipment. These variables interact with each other, exhibiting a high degree of nonlinearity and uncertainty. In such a complex system with multiple inputs and outputs, it is difficult to accurately capture the intrinsic relationships between the variables, resulting in poor assessment and optimization of energy efficiency.

[0003] Fuzzy neural networks, as a method that integrates the advantages of fuzzy systems and neural networks, offer a new perspective for solving such complex problems. They can handle fuzzy and uncertain information, constructing complex nonlinear mapping relationships by learning from large amounts of data. Simultaneously, multi-objective optimization algorithms can seek the optimal balance between energy saving and temperature control. Combining fuzzy neural networks with multi-objective optimization algorithms to construct a multi-input, multi-output data center energy efficiency model based on fuzzy networks holds promise for overcoming the limitations of traditional models. Summary of the Invention

[0004] 1. The technical problem to be solved:

[0005] To address the aforementioned technical problems, this invention provides a data center energy efficiency model based on fuzzy networks with multiple inputs and multiple outputs.

[0006] 2. Technical solution:

[0007] A data center energy efficiency model based on fuzzy networks with multiple inputs and multiple outputs, characterized by the following steps:

[0008] Step 1: Obtain the operating data of the data center IT equipment and air conditioning equipment and perform data preprocessing; the data center operating data includes cabinet power consumption, cold aisle temperature, hot aisle temperature, inlet air temperature, return air temperature, air volume, air supply temperature, total power of internal fans, total power of external fans, total power of compressors, and total power of air conditioning.

[0009] Step 2: Build a data center energy efficiency model based on fuzzy networks with multiple inputs and multiple outputs; the energy efficiency model uses fuzzy neural networks to predict the control parameters supply air temperature and supply air volume, and further optimizes the control parameters with multi-objective optimization algorithms;

[0010] Step 3: Divide the preprocessed overall sample set into training set and test set according to a certain ratio. Based on the completed fuzzy network architecture, adopt a multi-input multi-output design, use the training set data to iteratively train the model, and use the test set to verify the training results, so as to obtain a trained multi-input multi-output data center energy efficiency model based on fuzzy network.

[0011] Step 4: Send the optimal control parameters obtained from the energy efficiency model to the air conditioner to achieve timely and accurate control of the air conditioning equipment.

[0012] Furthermore, in step one, relevant operational data is collected and preprocessed, specifically including: identifying and removing outliers in historical data, filling in missing data using interpolation, and standardizing the data using normalization to improve data quality.

[0013] Furthermore, the fuzzy neural network described in step two combines fuzzy logic and neural networks, enabling it to process fuzzy information while also possessing the learning capabilities of a neural network. It is an algorithm capable of multi-parameter input and multi-parameter output, and specifically includes the following steps:

[0014] S21: Input parameter X = [x1, x2, ..., x n The input to the fuzzy neural network predicts the control parameters Y = [V,T]. s The input parameters are used as input to the fuzzy layer and transformed into fuzzy values. Each parameter is mapped to multiple fuzzy sets. The membership degree μ of the input belonging to different fuzzy sets is calculated using a Gaussian membership function. ij (x i The details are as follows:

[0015]

[0016] Among them, c ij σ represents the center of the j-th fuzzy set corresponding to the i-th parameter. ij This represents the width of the j-th fuzzy set corresponding to the i-th parameter;

[0017] S22: The membership degrees obtained in S21 are used as input to the rule layer, which is responsible for matching and reasoning fuzzy rules. The premise of the rule is composed of the outputs of the fuzzy layer, and the conclusion of the rule is a fuzzy set. The output of this layer is the excitation intensity of the rule. Specifically as follows:

[0018]

[0019] Where l = [1,2,...,L] represents the number of rules;

[0020] S23: Normalize the output of the rule layer in S22, and output the control parameter Y through the output layer, as follows:

[0021]

[0022] in, w represents the normalization strength of the l-th rule. l Indicates the weighting coefficient;

[0023] Furthermore, step two, based on the optimal temperature of the data center IT equipment in the server room to prevent the return air temperature from exceeding the safe temperature and to maximize the energy-saving effect of the air conditioning equipment, specifically includes the following steps:

[0024] S24: Based on the cooling capacity Q of the air conditioning equipment at different times t C,t and the heat dissipation Q of IT equipment IT,t Based on the mathematical relationship between them, a temperature control objective function f1(Y) is established to ensure that the return air temperature in the computer room is below the safe temperature, as follows:

[0025]

[0026] Q C =C P ρV(T r -T s ) (6)

[0027]

[0028] Where ω is the coefficient, C P Let T be the heat capacity of air, ρ be the density of air, and T be the temperature of air. r T represents the return air temperature. safe For safe temperature;

[0029] S25: Based on the total power P of the internal fan fan Total power of external fan P efan and the total power P of the compressor com To establish the relationship between the total power of the air conditioner and the total energy consumption of the air conditioning unit, a target function f2(Y) is established to minimize the total energy consumption of the air conditioning equipment, as follows:

[0030]

[0031] Among them, P ifan For rated input power, V ifanWhere is the rated air volume, and g is a nonlinear function of the total compressor power and the air supply temperature;

[0032] Furthermore, the multi-objective problem defined in step two involves optimizing the control parameters of the air conditioning equipment using the multi-objective algorithm NSGA-II, specifically including the following steps:

[0033] S26: In the operation system of a data center server room, the air supply volume V and the air supply temperature T s As a key control parameter, Y directly affects the cooling efficiency and energy consumption level of the air conditioning system. s Optimization also requires consideration of the actual conditions of the computer room, ensuring that the control parameters are within the constraints and establishing corresponding constraints, as follows:

[0034]

[0035] Where f1(Y) represents the objective function for achieving a suitable temperature, f2(Y) represents the objective function for energy saving in air conditioning, and V min V max This indicates the constraint range for the air supply volume. Indicates the constraint range of supply air temperature;

[0036] S27: Using the NSGA-II algorithm, multiple feasible solutions for supply air temperature and supply air volume can be obtained. Based on the safe range of return air temperature in the computer room and the energy-saving effect of data center cooling, these feasible solutions are comprehensively ranked. Combining the real-time operating status of IT equipment and air conditioning equipment, the optimal control parameters suitable for the air conditioning equipment are accurately selected from the ranked solution set, thereby achieving an efficient balance between data center temperature control and energy saving.

[0037] 3. Beneficial effects:

[0038] The multi-input multi-output (MIMO) data center energy efficiency model based on fuzzy networks proposed in this invention has significant beneficial effects: Addressing the complex characteristics of multiple input variables such as fluctuating IT equipment load and changing air conditioning operating parameters in data centers, as well as multiple output control parameters, the energy efficiency model of this invention can capture the dynamic correlation between various variables through the powerful nonlinear mapping capability of fuzzy neural networks. The model employs a multi-objective optimization algorithm to optimize the control parameters, overcoming the bottleneck of traditional methods that struggle to balance energy saving and temperature control. It seeks the optimal compromise among multiple feasible solutions, ensuring stable data center temperature to meet the safe operation requirements of IT equipment while maximizing air conditioning energy savings. This provides a systematic and intelligent solution for the green and efficient operation of data centers and has significant practical value for promoting energy conservation and emission reduction in the industry. Attached Figure Description

[0039] Figure 1 This is a diagram illustrating the overall framework of this method;

[0040] Figure 2 This is a flowchart of the data center operation process of the present invention;

[0041] Figure 3 This is a diagram of the fuzzy neural network structure of the present invention. Detailed Implementation

[0042] The present invention will now be described in detail with reference to the accompanying drawings.

[0043] As attached Figure 1 As shown, a data center energy efficiency model based on fuzzy networks with multiple inputs and multiple outputs is characterized by the following steps:

[0044] Step 1: Obtain the operating data of the data center IT equipment and air conditioning equipment and perform data preprocessing; the data center operating data includes cabinet power consumption, cold aisle temperature, hot aisle temperature, inlet air temperature, return air temperature, air volume, air supply temperature, total power of internal fans, total power of external fans, total power of compressors, and total power of air conditioning.

[0045] Step 2: Build a data center energy efficiency model based on fuzzy networks with multiple inputs and multiple outputs; the energy efficiency model uses fuzzy neural networks to predict the control parameters supply air temperature and supply air volume, and further optimizes the control parameters with multi-objective optimization algorithms;

[0046] Step 3: Divide the preprocessed overall sample set into training set and test set according to a certain ratio. Based on the completed fuzzy network architecture, adopt a multi-input multi-output design, use the training set data to iteratively train the model, and use the test set to verify the training results, so as to obtain a trained multi-input multi-output data center energy efficiency model based on fuzzy network.

[0047] Step 4: Send the optimal control parameters obtained from the energy efficiency model to the air conditioner to achieve timely and accurate control of the air conditioning equipment.

[0048] Furthermore, in step one, relevant operational data is collected and preprocessed, specifically including: identifying and removing outliers in historical data, filling in missing data using interpolation, and standardizing the data using normalization to improve data quality.

[0049] Furthermore, the fuzzy neural network described in step two combines fuzzy logic and neural networks, enabling it to process fuzzy information while also possessing the learning capabilities of a neural network. It is an algorithm capable of multi-parameter input and multi-parameter output, and specifically includes the following steps:

[0050] S21: Input parameter X = [x1, x2, ..., x nThe input to the fuzzy neural network predicts the control parameters Y = [V,T]. s The input parameters are used as input to the fuzzy layer and transformed into fuzzy values. Each parameter is mapped to multiple fuzzy sets. The membership degree μ of the input belonging to different fuzzy sets is calculated using a Gaussian membership function. ij (x i The details are as follows:

[0051]

[0052] Among them, c ij σ represents the center of the j-th fuzzy set corresponding to the i-th parameter. ij This represents the width of the j-th fuzzy set corresponding to the i-th parameter;

[0053] S22: The membership degrees obtained in S21 are used as input to the rule layer, which is responsible for matching and reasoning fuzzy rules. The premise of the rule is composed of the outputs of the fuzzy layer, and the conclusion of the rule is a fuzzy set. The output of this layer is the excitation intensity of the rule. Specifically as follows:

[0054]

[0055] Where l = [1,2,...,L] represents the number of rules;

[0056] S23: Normalize the output of the rule layer in S22, and output the control parameter Y through the output layer, as follows:

[0057]

[0058] in, w represents the normalization strength of the l-th rule. l Indicates the weighting coefficient;

[0059] Furthermore, step two, based on the optimal temperature of the data center IT equipment in the server room to prevent the return air temperature from exceeding the safe temperature and to maximize the energy-saving effect of the air conditioning equipment, specifically includes the following steps:

[0060] S24: Based on the cooling capacity Q of the air conditioning equipment at different times t C,t and the heat dissipation Q of IT equipment IT,t Based on the mathematical relationship between them, a temperature control objective function f1(Y) is established to ensure that the return air temperature in the computer room is below the safe temperature, as follows:

[0061]

[0062] Q C =C P ρV(T r -Ts ) (6)

[0063]

[0064] Where ω is the coefficient, C P Let T be the heat capacity of air, ρ be the density of air, and T be the temperature of air. r T represents the return air temperature. safe For safe temperature;

[0065] S25: Based on the total power P of the internal fan fan Total power of external fan P efan and the total power P of the compressor com To establish the relationship between the total power of the air conditioner and the total energy consumption of the air conditioning unit, a target function f2(Y) is established to minimize the total energy consumption of the air conditioning equipment, as follows:

[0066]

[0067] Among them, P ifan For rated input power, V ifan Where is the rated air volume, and g is a nonlinear function of the total compressor power and the air supply temperature;

[0068] Furthermore, the multi-objective problem defined in step two involves optimizing the control parameters of the air conditioning equipment using the multi-objective algorithm NSGA-II, specifically including the following steps:

[0069] S26: In the operation system of a data center server room, the air supply volume V and the air supply temperature T s As a key control parameter, Y directly affects the cooling efficiency and energy consumption level of the air conditioning system. s Optimization also requires consideration of the actual conditions of the computer room, ensuring that the control parameters are within the constraints and establishing corresponding constraints, as follows:

[0070]

[0071] Where f1(Y) represents the objective function for achieving a suitable temperature, f2(Y) represents the objective function for energy saving in air conditioning, and V min V max This indicates the constraint range for the air supply volume. Indicates the constraint range of supply air temperature;

[0072] S27: Using the NSGA-II algorithm, multiple feasible solutions for supply air temperature and supply air volume can be obtained. Based on the safe range of return air temperature in the computer room and the energy-saving effect of data center cooling, these feasible solutions are comprehensively ranked. Combining the real-time operating status of IT equipment and air conditioning equipment, the optimal control parameters suitable for the air conditioning equipment are accurately selected from the ranked solution set, thereby achieving an efficient balance between data center temperature control and energy saving.

[0073] Verification example:

[0074] This verification example uses actual operational data collected in a data center server room. Figure 1 This is the overall framework diagram of the multi-input multi-output data center energy efficiency model based on fuzzy networks of this invention. Figure 2 This involves analyzing the operational logic between the IT equipment and cooling system in the data center, and determining the operational data that needs to be collected for use in the energy efficiency model. Figure 3 This is a diagram of a fuzzy neural network structure. This method can effectively solve the problem of multiple inputs and multiple outputs.

Claims

1. A multi-input multi-output data center energy efficiency model based on fuzzy networks, characterized in that, Includes the following steps: Step 1: Obtain the operating data of the IT equipment and air conditioning equipment in the cooling system of the data center and perform data preprocessing; the data center operating data includes cabinet power, cold aisle temperature, hot aisle temperature, inlet air temperature, return air temperature, air volume, air supply temperature, total power of internal fans, total power of external fans, total power of compressors, and total power of air conditioners; Step 2: Build a data center energy efficiency model based on fuzzy networks with multiple inputs and multiple outputs; the energy efficiency model uses fuzzy neural networks to predict the control parameters supply air temperature and supply air volume, and further optimizes the control parameters with multi-objective optimization algorithms; Step 3: Divide the preprocessed overall sample set into training set and test set according to a certain ratio. Based on the completed fuzzy network architecture, adopt a multi-input multi-output design, use the training set data to iteratively train the model, and use the test set to verify the training results, so as to obtain a trained multi-input multi-output data center energy efficiency model based on fuzzy network. Step 4: Send the optimal control parameters obtained from the energy efficiency model to the air conditioner to achieve timely and accurate control of the air conditioning equipment.

2. The data center energy efficiency model based on fuzzy networks according to claim 1, characterized in that, In step one, relevant operational data is collected and preprocessed. Specifically, this includes: identifying and removing outliers from historical data, filling in missing data using interpolation, and standardizing the data using normalization to improve data quality.

3. The data center energy efficiency model based on fuzzy networks according to claim 2, characterized in that, In step two, the fuzzy neural network combines fuzzy logic and neural networks, enabling it to process fuzzy information while also possessing the learning capabilities of a neural network. It is an algorithm capable of multi-parameter input and multi-parameter output, and specifically includes the following steps: S21: Input parameter X = [x1, x2, ..., x n The input to the fuzzy neural network predicts the control parameters Y = [V,T]. s The input parameters are used as input to the fuzzy layer and transformed into fuzzy values. Each parameter is mapped to multiple fuzzy sets. The membership degree μ of the input belonging to different fuzzy sets is calculated using a Gaussian membership function. ij (x i The details are as follows: Among them, c ij σ represents the center of the j-th fuzzy set corresponding to the i-th parameter. ij This represents the width of the j-th fuzzy set corresponding to the i-th parameter; S22: The membership degrees obtained in S21 are used as input to the rule layer, which is responsible for matching and reasoning fuzzy rules. The premise of the rule is composed of the outputs of the fuzzy layer, and the conclusion of the rule is a fuzzy set. The output of this layer is the excitation intensity of the rule. Specifically as follows: Where l = [1,2,...,L] represents the number of rules; S23: Normalize the output of the rule layer in S22, and output the control parameter Y through the output layer, as follows: in, w represents the normalization strength of the l-th rule. l This represents the weighting coefficient.

4. The data center energy efficiency model based on fuzzy networks according to claim 3, characterized in that, Step two, based on the principle of keeping the data center IT equipment temperature at a suitable temperature in the server room to avoid the return air temperature exceeding the safe temperature and the energy-saving effect of the air conditioning equipment, specifically includes the following steps: S24: Based on the cooling capacity Q of the air conditioning equipment at different times t C,t and the heat dissipation Q of IT equipment IT,t Based on the mathematical relationship between them, a temperature control objective function f1(Y) is established to ensure that the return air temperature in the computer room is below the safe temperature, as follows: Q C =C P ρV(T r -T s ) (6) Where ω is the coefficient, C P Let T be the heat capacity of air, ρ be the density of air, and T be the temperature of air. r T represents the return air temperature. safe For safe temperature; S25: Based on the total power P of the internal fan fan Total power of external fan P efan and the total power P of the compressor com To establish the relationship between the total power of the air conditioner and the total energy consumption of the air conditioning unit, a target function f2(Y) is established to minimize the total energy consumption of the air conditioning equipment, as follows: Among them, P ifan For rated input power, V ifan denoted as the rated air volume, and g is a nonlinear function of the compressor's total power and air supply temperature.

5. A data center energy efficiency model based on fuzzy networks according to claim 4, characterized in that, In step two, the multi-objective problem is defined, and the control parameters of the air conditioning equipment are optimized using the multi-objective algorithm NSGA-II. Specifically, this includes the following steps: S26: In the operation system of a data center server room, the air supply volume V and the air supply temperature T s As a key control parameter, Y directly affects the cooling efficiency and energy consumption level of the air conditioning system. s Optimization also requires consideration of the actual conditions of the computer room, ensuring that the control parameters are within the constraints and establishing corresponding constraints, as follows: Where f1(Y) represents the objective function for achieving a suitable temperature, f2(Y) represents the objective function for energy saving in air conditioning, and V min V max This indicates the constraint range for the air supply volume. Indicates the constraint range of supply air temperature; S27: Using the NSGA-II algorithm, multiple feasible solutions for supply air temperature and supply air volume can be obtained. Based on the safe range of return air temperature in the computer room and the energy-saving effect of data center cooling, these feasible solutions are comprehensively ranked. By combining the real-time operating status of IT equipment and air conditioning equipment, the optimal control parameters for air conditioning equipment are accurately selected from the sorted solution set, thereby achieving an efficient balance between temperature control and energy saving in the data center.

Images