Data center energy saving method, device, system and storage medium

Abstract

The application discloses a data center energy-saving method and device and a storage medium. The method comprises the following steps: acquiring environment state data, IT total load and at least one data set; for each data set in the at least one data set, environment state data and a first sub-data set in each data set are input into a preset refrigeration station energy consumption prediction model, and first hidden layer data of the refrigeration station energy consumption prediction model is obtained; environment state data, the IT total load and a second sub-data set in the data set are input into a preset terminal air conditioner energy consumption prediction model, and second hidden layer data of the terminal air conditioner energy consumption prediction model is obtained; the first hidden layer data, the second hidden layer data, the environment state data and the IT total load corresponding to each data set are input into a preset machine room total energy consumption model, and the machine room total energy consumption corresponding to each data set is obtained; a target data set corresponding to the machine room total energy consumption satisfying a preset condition is determined, and an energy-saving scheme is determined according to the target data set.

Description

Technical Field

[0001] This invention relates to the field of network information security, and in particular to a data center energy-saving method, apparatus, system, and storage medium. Background Technology

[0002] With the large-scale commercialization of 5G (5th Generation Mobile Communication Technology) networks, data centers are gradually developing towards larger sizes and higher densities, with cumulative installed capacity potentially reaching millions of racks. At the same time, the high energy consumption of data centers is becoming increasingly prominent. Existing energy-saving methods for data centers fall into two categories: physical modeling of individual devices or traditional simple tuning methods based on human experience. However, both methods have their shortcomings.

[0003] Physical device modeling describes the cooling system using mathematical formulas, but the numerous, complex, and highly coupled components of the cooling system make it difficult to accurately model the entire system using data formulas. Traditional manual control relies on the industry experience of professional air conditioning maintenance personnel, which is time-consuming, labor-intensive, requires accumulated industry experience, and cannot achieve real-time, precise control, resulting in limited energy savings and failing to meet the requirements for further energy consumption reduction. Furthermore, data center building structure optimization methods require the building structure to be designed from the outset. Since each data center is not identical, the manual control experience accumulated in one data center cannot be widely extended to other data centers. Therefore, neither of the above two energy-saving solutions can be applied on a large scale to multiple data centers.

[0004] With the development and application of artificial intelligence (AI) technology, using AI and other technologies to model and optimize the control of data center cooling systems to achieve optimal overall energy efficiency has become a development direction in the field of data center energy conservation. However, while reinforcement learning, a method for finding optimal parameters based on data feedback, is theoretically suitable for data center energy conservation, its limited historical data parameter space, difficulty in data accumulation, and high complexity during model training make it difficult to apply in actual production. Summary of the Invention

[0005] In view of this, the main objective of the present invention is to provide a data center energy-saving method, apparatus, system, and storage medium.

[0006] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0007] This invention provides a data center energy-saving method, the method comprising:

[0008] Acquire environmental status data, total Internet device (IT) load, and at least one dataset; each of the at least one dataset includes: a first sub-dataset and a second sub-dataset; the first sub-dataset includes: data for at least one first control parameter and data for at least one second parameter; the second sub-dataset includes: data for at least one second control parameter and data for at least one third parameter;

[0009] For each of the at least one datasets, the environmental status data and the first subset of each dataset are input into a preset cooling station energy consumption prediction model to obtain the first hidden layer data of the cooling station energy consumption prediction model; the environmental status data, total IT load and the second subset of each dataset are input into a preset terminal air conditioning energy consumption prediction model to obtain the second hidden layer data of the terminal air conditioning energy consumption prediction model.

[0010] Input the first hidden layer data, the second hidden layer data, the environmental status data, and the total IT load corresponding to each dataset into a preset total energy consumption model for the computer room to obtain the total energy consumption of the computer room corresponding to each dataset.

[0011] A target dataset corresponding to the total energy consumption of the building that meets preset conditions is determined, and an energy-saving scheme is determined based on the target dataset; the energy-saving scheme is used to indicate the method of adjusting the at least one first control parameter and / or the at least one second control parameter.

[0012] In the above scheme, generating the preset energy consumption prediction model for the refrigeration station includes:

[0013] Obtain a first training sample set; the first training sample set includes: at least one first training set and a first label corresponding to each first training set in the at least one first training set; the first label represents the historical total energy consumption of the cooling station corresponding to the first training set;

[0014] The first training sample set is trained using the first neural network to obtain the trained first neural network, which serves as the energy consumption prediction model for the refrigeration station.

[0015] The first training set includes: historical environmental state data, historical data of at least one first control parameter, and historical data of at least one second parameter.

[0016] In the above scheme, generating the preset terminal air conditioning energy consumption prediction model includes:

[0017] Obtain a second training sample set; the first training sample set includes: at least one second training set and a second label corresponding to each second training set in the at least one second training set; the second label represents the historical total energy consumption of the terminal air conditioner corresponding to the second training set;

[0018] The second training sample set is trained using the second neural network to obtain the trained second neural network, which serves as the terminal air conditioner energy consumption prediction model.

[0019] The second training set includes: historical environmental status data, historical total IT load, historical data of at least one second control parameter, and historical data of at least one third parameter.

[0020] In the above scheme, generating the preset total energy consumption model of the terminal building includes:

[0021] During the training of the energy consumption prediction model of the refrigeration station and the energy consumption prediction model of the terminal air conditioner, the historical first hidden layer data of the energy consumption prediction model of the refrigeration station and the historical second hidden layer data of the energy consumption prediction model of the terminal air conditioner are extracted.

[0022] Determine the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the computer building; determine a first ratio based on the historical energy consumption of the refrigeration station and the historical total energy consumption of the computer building; and determine a second ratio based on the historical total energy consumption of the terminal air conditioning and the historical total energy consumption of the computer building.

[0023] Using the first ratio as the initial weight of the historical first hidden layer data and the second ratio as the initial weight of the historical second hidden layer data, fully connected layers are constructed on the historical first hidden layer data and the historical second hidden layer data, respectively. The constructed fully connected layers are then connected to the neural network to obtain the target training model.

[0024] The target training model is trained by taking the historical first hidden layer data, the historical second hidden layer data, the historical environmental status data, and the historical total IT load as inputs and the historical total energy consumption of the computer room as outputs. The trained target training model is then used as the total energy consumption model of the computer room.

[0025] In the above scheme, determining the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the server room, determining a first ratio based on the historical refrigeration station energy consumption and the historical server room total energy consumption, and determining a second ratio based on the historical terminal air conditioning total energy consumption and the historical server room total energy consumption, includes:

[0026] The average values ​​of the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the computer building are obtained by averaging the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the computer building.

[0027] The first ratio is obtained by dividing the average total energy consumption of the historical refrigeration station by the average total energy consumption of the historical computer building.

[0028] The second ratio is obtained by dividing the historical average total air conditioning energy consumption of the terminal by the historical average total energy consumption of the building.

[0029] The method in the above scheme further includes:

[0030] An adjustable parameter is determined as a control parameter; the control parameter includes: a first control parameter and a second control parameter;

[0031] Query the historical data of the control parameter to determine the historical maximum and minimum values ​​of the control parameter;

[0032] Determine a first multiple of the historical maximum value as the first limit;

[0033] Determine the second multiple of the historical minimum value as the second limit value;

[0034] Accordingly, obtaining at least one dataset includes:

[0035] Using a genetic algorithm, the combination of control parameters for each dataset is determined based on the first and second limits of each control parameter;

[0036] The combination of control parameters includes: at least one first control parameter and at least one second control parameter.

[0037] In the above scheme, the energy-saving scheme includes: at least one device to be adjusted, and an adjustment value of at least one control parameter for each of the at least one device to be adjusted;

[0038] The method further includes:

[0039] The energy-saving scheme is sent to the group control system; the group control system is used to adjust the device to be adjusted according to the energy-saving scheme so that the control parameters of the device to be adjusted reach the corresponding adjustment values.

[0040] This invention provides a data center energy-saving device, comprising: an acquisition module, a first processing module, a second processing module, and a third processing module; wherein,

[0041] The acquisition module is used to acquire environmental status data, total IT load, and at least one dataset; each of the at least one dataset includes: a first sub-dataset and a second sub-dataset; the first sub-dataset includes: data of at least one first control parameter and data of at least one second parameter; the second sub-dataset includes: data of at least one second control parameter and data of at least one third parameter.

[0042] The first processing module is configured to, for each of the at least one datasets, input the environmental status data and the first sub-dataset of each dataset into a preset cooling station energy consumption prediction model to obtain the first hidden layer data of the cooling station energy consumption prediction model; and input the environmental status data, total IT load and the second sub-dataset of each dataset into a preset terminal air conditioning energy consumption prediction model to obtain the second hidden layer data of the terminal air conditioning energy consumption prediction model.

[0043] The second processing module is used to input the first hidden layer data, the second hidden layer data, the environmental status data, and the total IT load corresponding to each dataset into a preset total energy consumption model of the computer room, so as to obtain the total energy consumption of the computer room corresponding to each dataset.

[0044] The third processing module is used to determine the target dataset corresponding to the total energy consumption of the building that meets the preset conditions, and to determine an energy-saving scheme based on the target dataset; the energy-saving scheme is used to indicate the method of adjusting the at least one first control parameter and / or the at least one second control parameter.

[0045] In the above scheme, the device further includes: a preprocessing module;

[0046] The preprocessing module is used to obtain a first training sample set; the first training sample set includes: at least one first training set and a first label corresponding to each first training set in the at least one first training set; the first label represents the total energy consumption of the historical cooling station corresponding to the first training set;

[0047] The first training sample set is trained using the first neural network to obtain the trained first neural network, which serves as the energy consumption prediction model for the refrigeration station.

[0048] The first training set includes: historical environmental state data, historical data of at least one first control parameter, and historical data of at least one second parameter.

[0049] In the above scheme, the preprocessing module is further used to obtain a second training sample set; the first training sample set includes: at least one second training set and a second label corresponding to each second training set in the at least one second training set; the second label represents the historical total energy consumption of the terminal air conditioner corresponding to the second training set;

[0050] The second training sample set is trained using the second neural network to obtain the trained second neural network, which serves as the terminal air conditioner energy consumption prediction model.

[0051] The second training set includes: historical environmental status data, historical total IT load, historical data of at least one second control parameter, and historical data of at least one third parameter.

[0052] In the above scheme, the preprocessing module is also used to extract the historical first hidden layer data of the refrigeration station energy consumption prediction model and the historical second hidden layer data of the terminal air conditioner energy consumption prediction model during the training of the refrigeration station energy consumption prediction model and the terminal air conditioner energy consumption prediction model.

[0053] Determine the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the computer building; determine a first ratio based on the historical energy consumption of the refrigeration station and the historical total energy consumption of the computer building; and determine a second ratio based on the historical total energy consumption of the terminal air conditioning and the historical total energy consumption of the computer building.

[0054] Using the first ratio as the initial weight of the historical first hidden layer data and the second ratio as the initial weight of the historical second hidden layer data, fully connected layers are constructed on the historical first hidden layer data and the historical second hidden layer data, respectively. The constructed fully connected layers are then connected to the neural network to obtain the target training model.

[0055] The target training model is trained by taking the historical first hidden layer data, the historical second hidden layer data, the historical environmental status data, and the historical total IT load as inputs and the historical total energy consumption of the computer room as outputs. The trained target training model is then used as the total energy consumption model of the computer room.

[0056] In the above scheme, the preprocessing module is used to calculate the average of the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the computer building determined within the collection time, so as to obtain the average historical total energy consumption of the refrigeration station, the average historical total energy consumption of the terminal air conditioning, and the average historical total energy consumption of the computer building.

[0057] The first ratio is obtained by dividing the average total energy consumption of the historical refrigeration station by the average total energy consumption of the historical computer building.

[0058] The second ratio is obtained by dividing the historical average total air conditioning energy consumption of the terminal by the historical average total energy consumption of the building.

[0059] In the above scheme, the acquisition module is further used to determine adjustable parameters as control parameters; the control parameters include: a first control parameter and a second control parameter;

[0060] Query the historical data of the control parameter to determine the historical maximum and minimum values ​​of the control parameter;

[0061] Determine a first multiple of the historical maximum value as the first limit;

[0062] Determine the second multiple of the historical minimum value as the second limit value;

[0063] Accordingly, the acquisition module is used to use a genetic algorithm to determine the combination of control parameters for each dataset based on the first and second limits of each control parameter;

[0064] The combination of control parameters includes: at least one first control parameter and at least one second control parameter.

[0065] In the above scheme, the energy-saving scheme includes: at least one device to be adjusted, and an adjustment value of at least one control parameter for each of the at least one device to be adjusted;

[0066] The device further includes: a communication module; the communication module is used to send the energy-saving scheme to the group control system; the group control system is used to adjust the device to be adjusted according to the energy-saving scheme so that the control parameters of the device to be adjusted reach the adjustment values ​​of the corresponding control parameters.

[0067] This invention provides a data center energy-saving device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of any of the methods described on the server side.

[0068] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of any of the methods described on the server side.

[0069] This invention provides a data center energy-saving method, apparatus, and storage medium. The method includes: acquiring environmental status data, total IT load, and at least one dataset; each dataset includes a first subset and a second subset; the first subset includes data for at least one first control parameter and data for at least one second parameter; the second subset includes data for at least one second control parameter and data for at least one third parameter; for each dataset, the environmental status data and the first subset are input into a preset cooling station energy consumption prediction model to obtain the first hidden layer data of the cooling station energy consumption prediction model; the environmental status data, total IT load, and second subset are input into a preset terminal air conditioning energy consumption prediction model to obtain the terminal... The second hidden layer data of the air conditioning energy consumption prediction model; inputting the first hidden layer data, the second hidden layer data, environmental status data, and total IT load corresponding to each dataset into a preset total building energy consumption model to obtain the total building energy consumption corresponding to each dataset; determining the target dataset corresponding to the total building energy consumption that meets preset conditions, and determining an energy-saving scheme based on the target dataset; the energy-saving scheme is used to indicate the method of adjusting at least one first control parameter and / or at least one second control parameter; thus, by predicting the total building power consumption of multiple schemes based on the real-time acquired current operating status data of each device, the optimal scheme is determined as the energy-saving scheme, which can effectively utilize real-time monitoring data to make inferences and solve the problems of insufficient adjustment timeliness and only local optimization; in addition, the above method uses artificial intelligence to find the optimal control parameters, reducing the reliance on human experience. Attached Figure Description

[0070] Figure 1 A schematic flowchart of a data center energy-saving method provided in an embodiment of the present invention;

[0071] Figure 2 A schematic diagram illustrating a model building method provided in an application embodiment of the present invention;

[0072] Figure 3 A schematic flowchart of a model training method provided for an application embodiment of the present invention;

[0073] Figure 4 A flowchart illustrating a data center energy-saving method provided as an application embodiment of the present invention;

[0074] Figure 5 This is a schematic diagram of the structure of a data center energy-saving device provided in an embodiment of the present invention;

[0075] Figure 6This is a schematic diagram of another data center energy-saving device provided in an embodiment of the present invention. Detailed Implementation

[0076] The present invention will be further described in detail below with reference to the embodiments.

[0077] Figure 1 This is a flowchart illustrating a data center energy-saving method according to an embodiment of the present invention; as shown below. Figure 1 As shown, the method includes:

[0078] Step 101: Obtain environmental status data, total load of Internet devices (IT, Internet Technology), and at least one dataset; each of the at least one dataset includes: a first sub-dataset and a second sub-dataset; the first sub-dataset includes: data of at least one first control parameter and data of at least one second parameter; the second sub-dataset includes: data of at least one second control parameter and data of at least one third parameter;

[0079] Step 102: For each dataset in the at least one dataset, input the environmental status data and the first subset of each dataset into a preset cooling station energy consumption prediction model to obtain the first hidden layer data of the cooling station energy consumption prediction model; input the environmental status data, total IT load and the second subset of each dataset into a preset terminal air conditioning energy consumption prediction model to obtain the second hidden layer data of the terminal air conditioning energy consumption prediction model.

[0080] Step 103: Input the first hidden layer data, the second hidden layer data, the environmental status data, and the total IT load corresponding to each dataset into the preset total energy consumption model of the computer room to obtain the total energy consumption of the computer room corresponding to each dataset;

[0081] Step 104: Determine the target dataset corresponding to the total energy consumption of the building that meets the preset conditions, and determine the energy-saving scheme based on the target dataset; the energy-saving scheme is used to indicate the method of adjusting the at least one first control parameter and / or the at least one second control parameter.

[0082] In practical applications, considering the difficulty of accumulating historical data, if only the collected historical data is used, the space of historical data parameters is relatively small. In order to expand the data space, the data can be expanded.

[0083] Based on this, in some embodiments, the method further includes:

[0084] An adjustable parameter is determined as a control parameter; the control parameter includes: a first control parameter and a second control parameter;

[0085] Query the historical data of the control parameter to determine the historical maximum and minimum values ​​of the control parameter;

[0086] Determine a first multiple of the historical maximum value as the first limit;

[0087] Determine the second multiple of the historical minimum value as the second limit value;

[0088] Accordingly, obtaining at least one dataset includes:

[0089] Using a genetic algorithm, the combination of control parameters for each dataset is determined based on the first and second limits of each control parameter;

[0090] The combination of control parameters includes: at least one first control parameter and at least one second control parameter.

[0091] For example, the first multiple is 1.05 times; the second multiple is 0.95 times. That is, after finding the maximum value m1 and the minimum value m2 of the control parameter from historical data, the upper and lower limits of the parameter space are set as m1*1.05 and m2*0.95 respectively, as the first limit and the second limit.

[0092] Here, the control parameters include: a first control parameter and a second control parameter; that is, the above operation can be performed on each first control parameter and each second control parameter. When obtaining at least one dataset, a genetic algorithm can be used to generate a combination of N control parameters {n1, n2, ..., n} based on historical data. N}, where n i This represents a combination of control parameters, with each control parameter having a value range of [m1*1.05, m2*0.95]. Each dataset contains a combination of control parameters.

[0093] The method in this invention considers that 70% of the non-IT energy consumption of a data center comes from the cooling system. The energy consumption of the cooling system is affected by the operating status of the cooling station equipment and the terminal air conditioners. The cooling station and the terminal air conditioners are mutually influential and coupled; for example, the greater the cooling capacity provided by the cooling station, the greater the energy consumption, but the less energy is required by the air conditioner terminals. Another important factor affecting data center energy consumption is IT energy consumption. The workload carried by server equipment is not constant; the greater the workload, the higher the total IT load, and the higher the IT energy consumption. Therefore, intelligent energy saving in data centers needs to consider the impact of the operating status of the cooling station and terminal air conditioners on the total energy consumption of the building, as well as the changes in IT energy consumption caused by changes in workload. Based on this, this invention modeles the total energy consumption of the data center building. The resulting total energy consumption model can predict the total energy consumption of the cooling equipment in various states, and based on the minimum energy consumption prediction results, it inversely deduces energy-saving schemes, ultimately outputting the control parameter values ​​of each cooling device.

[0094] Based on this, in some embodiments, the method further includes:

[0095] Generate the preset energy consumption prediction model for the refrigeration station;

[0096] Generate the preset terminal air conditioner energy consumption prediction model;

[0097] Generate the preset total energy consumption model for the building.

[0098] In some embodiments, generating the preset energy consumption prediction model for the cooling station includes:

[0099] Obtain a first training sample set; the first training sample set includes: at least one first training set and a first label corresponding to each first training set in the at least one first training set; the first label represents the historical total energy consumption of the cooling station corresponding to the first training set;

[0100] The first training sample set is trained using the first neural network to obtain the trained first neural network, which serves as the energy consumption prediction model for the refrigeration station.

[0101] The first training set includes: historical environmental state data, historical data of at least one first control parameter, and historical data of at least one second parameter.

[0102] In some embodiments, generating the preset terminal air conditioning energy consumption prediction model includes:

[0103] Obtain a second training sample set; the first training sample set includes: at least one second training set and a second label corresponding to each second training set in the at least one second training set; the second label represents the historical total energy consumption of the terminal air conditioner corresponding to the second training set;

[0104] The second training sample set is trained using the second neural network to obtain the trained second neural network, which serves as the terminal air conditioner energy consumption prediction model.

[0105] The second training set includes: historical environmental status data, historical total IT load, historical data of at least one second control parameter, and historical data of at least one third parameter.

[0106] In some embodiments, generating the preset total energy consumption model for the terminal building includes:

[0107] During the training of the energy consumption prediction model of the refrigeration station and the energy consumption prediction model of the terminal air conditioner, the historical first hidden layer data of the energy consumption prediction model of the refrigeration station and the historical second hidden layer data of the energy consumption prediction model of the terminal air conditioner are extracted.

[0108] Determine the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the computer building; determine a first ratio based on the historical energy consumption of the refrigeration station and the historical total energy consumption of the computer building; and determine a second ratio based on the historical total energy consumption of the terminal air conditioning and the historical total energy consumption of the computer building.

[0109] Using the first ratio as the initial weight of the historical first hidden layer data and the second ratio as the initial weight of the historical second hidden layer data, fully connected layers are constructed for the historical first hidden layer data and the historical second hidden layer data, respectively. The constructed fully connected layers are then connected to a neural network (such as a long short-term memory network (LSTM)) to obtain the target training model.

[0110] The target training model is trained by taking the historical first hidden layer data, the historical second hidden layer data, the historical environmental status data, and the historical total IT load as inputs and the historical total energy consumption of the computer room as outputs. The trained target training model is then used as the total energy consumption model of the computer room.

[0111] Specifically, during the training process of the refrigeration plant energy consumption prediction model and the terminal air conditioning energy consumption prediction model, the hidden layer data of the refrigeration plant energy consumption prediction model and the terminal air conditioning energy consumption prediction model are extracted to obtain historical first hidden layer data and historical second hidden layer data. The first ratio and the second ratio are used as initial weights to assign values, and fully connected layers are constructed on the historical first hidden layer data and the historical second hidden layer data, respectively. A neural network (such as LSTM) model is connected after the fully connected layer. The historical first hidden layer data, historical second hidden layer data, historical environmental state data, and historical total IT load are combined as inputs, and the historical total energy consumption of the computer building is used as output. The fully connected layer + LSTM model is trained to obtain the total energy consumption model of the computer building.

[0112] In some embodiments, determining the historical total energy consumption of the chiller plant, the historical total energy consumption of terminal air conditioning, and the historical total energy consumption of the server room, determining a first ratio based on the historical chiller plant energy consumption and the historical server room total energy consumption, and determining a second ratio based on the historical terminal air conditioning total energy consumption and the historical server room total energy consumption, includes:

[0113] The average values ​​of the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the computer building are obtained by averaging the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the computer building.

[0114] The first ratio is obtained by dividing the average total energy consumption of the historical refrigeration station by the average total energy consumption of the historical computer building.

[0115] The second ratio is obtained by dividing the historical average total air conditioning energy consumption of the terminal by the historical average total energy consumption of the building.

[0116] Here, the first ratio represents the proportion of refrigeration station energy consumption in the total energy consumption of the building; the second ratio represents the proportion of terminal air conditioning total energy consumption in the total energy consumption of the building.

[0117] The data center energy-saving method provided in this invention can be applied to a smart device, such as a server or cloud server. The method can be applied to any data center, and based on the actual situation of each data center, the energy-saving scheme can be determined using steps 101-104 above. The acquired environmental status data, total IT load, and at least one dataset are collected in real time based on the actual situation.

[0118] The first subset includes: refrigeration equipment operation status data; the second subset includes: terminal air conditioner operation status data.

[0119] Different data centers may be equipped with different cooling equipment and terminal air conditioners. For example, the cooling equipment may include other equipment besides the terminal air conditioners in Table 1 below; correspondingly, the first subset of data includes the status data of other equipment besides the terminal air conditioners in Table 1.

[0120] The terminal air conditioner operating status data includes: terminal air conditioner speed and outlet air temperature.

[0121] The environmental status data includes: outdoor relative humidity, outdoor dry bulb temperature, and outdoor wet bulb temperature.

[0122] The control parameters (including a first control parameter and a second control parameter) are adjustable parameters. For example, the adjustable parameters may include: chiller cooling water outlet temperature, chiller chilled water outlet temperature, chilled water pump frequency, cooling pump frequency, cooling tower outlet water temperature, distributor water pressure, and terminal air conditioner speed. The chiller cooling water outlet temperature, chiller chilled water outlet temperature, chilled water pump frequency, cooling pump frequency, cooling tower outlet water temperature, and distributor water pressure can be considered the first control parameter, and the terminal air conditioner speed can be considered the second control parameter.

[0123] The second and third parameters are non-adjustable. For example, the second control parameter may include: chiller compressor running time, chiller condensing saturation temperature, chilled water pump inlet pressure, chilled water pump inlet and outlet pressure, etc. The third parameter may include: the outlet air temperature of the terminal air conditioner.

[0124] To collect the above data, data acquisition devices such as temperature sensors and humidity sensors can be installed in the data center. The specific settings are based on actual needs and are not limited here.

[0125]

[0126]

[0127] Table 1

[0128] Accordingly, the first training sample set, the second training sample set, historical environmental status data, and historical total IT load required to train each model can be obtained based on historical data from the data center. For example, historical cooling station equipment operation status data, equipment energy consumption data, and historical environmental status data of a certain data center can be obtained; the various historical data can be grouped and organized to obtain the first training sample set, the second training sample set, historical environmental status data, and historical total IT load.

[0129] Among them, the historical energy consumption data of the equipment can include the historical energy consumption data of each piece of equipment, such as the energy consumption of chiller units, chilled water pumps, cooling pumps, cooling towers, terminal air conditioning units, IT loads, and the total energy consumption of the building.

[0130] The first label used in the model training process represents the total energy consumption of the historical refrigeration station corresponding to the first training set. The total energy consumption of the historical refrigeration station can be the sum of the energy consumption of each piece of equipment in the refrigeration station, such as the sum of the energy consumption of the chiller unit, the energy consumption of the chilled pump, the energy consumption of the cooling pump, the energy consumption of the cooling tower, etc.

[0131] The second label represents the total historical energy consumption of the terminal air conditioners corresponding to the second training set, and the total energy consumption of the terminal air conditioners can be the sum of the energy consumption of each terminal air conditioner;

[0132] When training the total energy consumption model of the computer building, the historical total energy consumption of the computer building is used as the output (i.e., the label). The historical total energy consumption of the computer building can be the sum of the energy consumption of each piece of equipment, such as the sum of the energy consumption of the chiller unit, the energy consumption of the chilled pump, the energy consumption of the cooling pump, the energy consumption of the cooling tower, the energy consumption of the terminal air conditioning, the energy consumption of the IT load, and the total energy consumption of the computer building.

[0133] Historical total IT load can be the sum of the loads of all IT components.

[0134] For different data centers, the aforementioned first label, second label, historical total IT load, and historical environmental status data are compiled based on historical data collected for that data center, and no specific restrictions are placed on the equipment involved or the values ​​used.

[0135] In some embodiments, determining the target dataset corresponding to the total energy consumption of the building that meets preset conditions includes:

[0136] After determining the total power consumption of the terminal building predicted for each dataset, compare the total power consumption of the terminal building predicted for each dataset.

[0137] The dataset corresponding to the minimum total power consumption of the building is determined as the target dataset.

[0138] In some embodiments, the method further includes:

[0139] Based on the control parameter data of the target dataset, an energy-saving scheme is determined;

[0140] The control parameter data includes: the data of the first control parameter and the data of the second control parameter in the target dataset;

[0141] The energy-saving scheme includes: at least one device to be adjusted, and adjustment values ​​of at least one control parameter for each of the at least one device to be adjusted;

[0142] In some embodiments, the method further includes:

[0143] The energy-saving scheme is sent to the group control system; the group control system is used to adjust the device to be adjusted according to the energy-saving scheme so that the control parameters of the device to be adjusted reach the corresponding adjustment values.

[0144] The method provided in this invention uses artificial intelligence to replace manual control. It only requires modification of the existing data center's collection points and does not need to collect data from the entire sample space; training based on existing historical data is sufficient. The adopted total energy consumption model for the data center comprehensively considers factors such as the performance of various equipment in the cooling station and terminal air conditioners, the outdoor environment, and the operating status of other equipment in the cooling station. It combines the cooling station and terminal air conditioners—the building's most energy-consuming modules—for state optimization, thereby achieving energy conservation and emission reduction for the entire building. Furthermore, through deep learning, it simultaneously captures the relational features of data at adjacent time points and the periodicity of the data, providing the optimal energy-saving solution.

[0145] The data center energy-saving method provided in this invention is based on deep learning and machine learning to achieve energy conservation and emission reduction in data centers. The method involves offline model training and updating, with online model inference only. In application, inference can be performed once every fixed time period, and the control parameters output by the model can be sent to the group control system.

[0146] Figure 2 A schematic diagram of a model building method provided in an application embodiment of the present invention, as shown below. Figure 2 As shown, the method provided in this embodiment of the invention is based on multiple deep learning models. First, a deep neural network model is constructed to predict the relationship between the status of each piece of equipment in the refrigeration station and its energy consumption (i.e., the refrigeration station energy consumption prediction model), and a deep neural network model is constructed to predict the relationship between the operating status of the terminal air conditioners and their energy consumption (i.e., the terminal air conditioner energy consumption prediction model). Then, using the hidden layer information of the two models and combining time features, a total energy consumption model for the air conditioning building is constructed. This total energy consumption model for the air conditioning building can predict the energy consumption of the building over a future period. Through this multi-level nested deep neural network cascade method, the energy consumption of the entire air conditioning building under a certain state is predicted, thereby finding the scheme that minimizes the total energy consumption of the air conditioning building from multiple sets of alternative control parameter schemes.

[0147] Figure 2 The inputs to the energy consumption prediction model for the refrigeration station may include: historical environmental status data (including: outdoor relative humidity, outdoor dry bulb temperature, and outdoor wet bulb temperature), and historical operating status data of the refrigeration equipment (including: data of various status parameters of equipment other than the terminal air conditioners in Table 1); the output may include: historical total energy consumption of the refrigeration station.

[0148] The inputs to the terminal air conditioning energy consumption prediction model may include: historical environmental status data, historical total IT load, and historical operating status data of the terminal air conditioning (including: terminal air conditioning speed and outlet air temperature); the output may include: historical total energy consumption of the terminal air conditioning.

[0149] Figure 3This is a flowchart illustrating a model training method provided in an application embodiment of the present invention. Before describing the specific steps, the following explains the terms involved:

[0150] Control parameters: These are parameters of the equipment that can be adjusted in real time. For example, control parameters may include: chiller cooling water outlet temperature, chiller chilled water outlet temperature, chilled water pump frequency, cooling water pump frequency, cooling tower frequency, and terminal air conditioner speed, etc. It should be noted that control parameters can be adjusted according to the actual situation of the data center.

[0151] Parameter space: The adjustable range of control parameters for each piece of equipment in the refrigeration plant. Because the historical data sample space is small, the upper and lower limits of each parameter are expanded by 5% during parameter space exploration. In application, a genetic algorithm is used to transform the infinite parameter space into a finite one. The optimal solution is found by setting the total energy consumption of the chiller tower as the objective function, and then the solution is output.

[0152] Parameter distribution: Data is continuously collected, and model inference is performed at fixed time intervals. The resulting optimal parameter combination is then distributed to the group control system, which controls the set values ​​of each physical device.

[0153] Data Acquisition: Collect data on the operational status of data center cooling equipment, equipment energy consumption, and environmental conditions over a specified period (e.g., at least five-minute intervals within a year). During the acquisition process, the acquisition time and specific values ​​should be recorded with minute-level precision. Specific fields to be collected for cooling equipment operational status are shown in Table 1. Equipment energy consumption data may include: chiller energy consumption, chilled pump energy consumption, cooling pump energy consumption, cooling tower energy consumption, terminal air conditioning energy consumption, IT load energy consumption, and total building energy consumption. Environmental conditions to be collected include: outdoor relative humidity, outdoor dry-bulb temperature, and outdoor wet-bulb temperature.

[0154] Feature Construction: First, train the energy consumption prediction model for the refrigeration plant and the terminal air conditioning unit. Then, extract the hidden layers of these two models as features for the total energy consumption model of the server room. The feature construction methods for the refrigeration plant energy consumption prediction model, the terminal air conditioning unit energy consumption prediction model, and the total energy consumption model of the server room are as follows: Figure 2 As shown.

[0155] The characteristic data of the refrigeration plant energy consumption prediction model include: total energy consumption of refrigeration plant equipment, refrigeration plant equipment operating status data, and environmental status data. The refrigeration plant equipment operating status data and environmental status data can be obtained directly from collected historical data. The total energy consumption of the refrigeration plant equipment is obtained by summing the energy consumption of the chiller, refrigeration pump, cooling pump, and cooling tower. Then, using the refrigeration equipment operating status data and environmental status data as inputs, and the total energy consumption data of the refrigeration equipment as output, an energy consumption model expressing the relationship between the chiller operating status and energy consumption is constructed; this is the refrigeration plant energy consumption prediction model.

[0156] The characteristic data for the terminal air conditioning energy consumption prediction model includes: terminal air conditioning status data, environmental status data, total IT load, time, and terminal air conditioning energy consumption. All data can be obtained directly from collected historical data. The terminal air conditioning energy consumption data is used as the model output, and the remaining values ​​are used as model inputs to construct the terminal air conditioning energy consumption prediction model.

[0157] Building Total Energy Consumption Model Features: The hidden layers of the cooling station energy consumption prediction model and the terminal air conditioning energy consumption prediction model are extracted and used as energy consumption features for the cooling station and terminal air conditioning. These features are then combined with data collection time, environmental status data, and total IT load to construct the air conditioning system feature data for the data center. The total energy consumption of the data center is obtained by adding the total energy consumption of the cooling station equipment, the terminal air conditioning energy consumption, the total IT load energy consumption, and the energy consumption of basic electrical equipment supporting lighting, office power, etc. The air conditioning system feature data is used as input, and the total energy consumption of the data center is used as output to construct the total energy consumption model of the data center.

[0158] The following describes the algorithm model used in the energy-saving method of this invention:

[0159] 1. Genetic Algorithm: The operating state of data center cooling station equipment belongs to a continuous parameter space, which can exhaustively enumerate countless different combinations of state parameters. However, the control parameters for the next moment should be fine-tuned based on the current operating state. Genetic algorithms have the characteristics of strong robustness, ability to find the global optimal solution to optimization problems, and suitability for solving complex data problems. Therefore, the genetic algorithm is used as a parameter generation engine to generate N sets of parameter combinations based on the historical data of the cooling station and put them into the strategy pool. The objective function is set as the minimum predicted total energy consumption of the building, thereby obtaining the optimal parameter combination solution that meets the objective function.

[0160] 2. Refrigeration Plant Energy Consumption Prediction Model: The total energy consumption of the building is strongly correlated with the current operating status and control parameter settings of the refrigeration plant equipment. The refrigeration plant equipment includes: chillers, chilled water pumps, cooling water pumps, water distributors, water collectors, chilled water mains, cooling water mains, cooling towers, fans, etc. There are many types of equipment, and the monitoring values ​​of different equipment at different measuring points vary slightly; for example, the temperature measured by a water pump differs at different measuring points. Therefore, traditional methods of decoupling the various equipment in the refrigeration plant based on physical structure and HVAC knowledge and constructing corresponding mathematical formulas are insufficient to accurately describe the energy consumption of the refrigeration plant. Considering the complex connections between various equipment, the large number of devices, and the varying measuring points, this embodiment of the invention uses a neural network to construct an artificial intelligence (AI) model capable of expressing the energy consumption of the refrigeration plant. This model encodes the energy consumption of the refrigeration plant and is used to train the final total energy consumption model of the building.

[0161] 3. Terminal Air Conditioning Energy Consumption Prediction Model: Terminal air conditioners in the data center directly cool the servers. The cooling capacity and energy consumption per unit time depend on the current server temperature, IT load, etc. From a physical perspective, the chiller outlet water is directly connected to the terminal air conditioner via a distributor and acts on the terminal air conditioner for cooling. From the cooling principle perspective, the air conditioner speed also directly affects the terminal outlet air temperature. Considering that server energy consumption, chiller outlet water temperature, and terminal air conditioner speed are all strongly correlated and coupled with terminal air conditioner energy consumption, this embodiment of the invention uses a neural network to model and represent the energy consumption of the terminal air conditioner, and incorporates the hidden layer representing the terminal air conditioner energy consumption as a feature into the final data center energy consumption prediction model.

[0162] 4. Total Energy Consumption Model of the Terminal Building: The energy consumption of the cooling station and terminal air conditioning accounts for the largest proportion of the total energy consumption of the terminal building. However, the energy consumption of the terminal building also needs to support basic electricity consumption such as office work and lighting. Therefore, in this embodiment of the invention, the proportion of energy consumption of the cooling station and terminal air conditioning is calculated based on historical data, and the proportion is used as the basic weight to construct a fully connected layer. An LSTM model is then used to model the total energy consumption of the terminal building from the perspective of time series prediction. During the model training process, the basic weight values ​​are continuously optimized through iteration to obtain the final total energy consumption model of the terminal building. Specifically, as follows... Figure 2 As shown.

[0163] The method provided in this embodiment of the invention includes the following inputs: historical data on the operating status of each device in the cooling station of a data center with dimension LxN, historical data on device energy consumption, and historical data on environmental status for training; the outputs include: specific suggested values ​​for the control parameters of each device in the data center cooling station at the next control moment, and the total energy consumption of the building after adjustment according to the suggested values.

[0164] The method specifically includes: a model training and / or update phase, and a data center energy-saving solution determination phase.

[0165] The following combination Figure 2 and Figure 3 The model training method is described below, and the model training method includes:

[0166] Step 301: Collect historical data and group the collected historical data according to the dimensions of refrigeration station and air conditioning.

[0167] The historical data includes three types of data: equipment status, equipment energy consumption, and environmental status. Specifically, the historical data collection can be as described above, collecting at least one year's worth of data on the data center's cooling equipment operating status, equipment energy consumption, and environmental status at a five-minute granularity. During the collection process, the collection time and specific collection values ​​should be recorded with a minute-level precision; that is, the vertical dimension is time, and the horizontal dimension is the specific value at the collection point.

[0168] For example, the specific fields of the collected refrigeration equipment operating status data can be shown in Table 1;

[0169] The collected equipment energy consumption data may include: chiller energy consumption, chilled pump energy consumption, cooling pump energy consumption, cooling tower energy consumption, terminal air conditioning energy consumption, IT equipment energy consumption, and total building energy consumption.

[0170] The collected environmental data may include: outdoor relative humidity, outdoor dry bulb temperature, and outdoor wet bulb temperature.

[0171] Accordingly, the historical data collected is grouped and organized according to the dimensions of refrigeration stations and air conditioning systems, including:

[0172] When organizing by group, time is used as the vertical dimension and the specific values ​​of the collection points are used as the horizontal dimension;

[0173] Determine the historical operating status data of the refrigeration equipment, and record it as A. T*k This may include the status parameters of other equipment besides the terminal air conditioner in Table 1;

[0174] Determine the historical operating status data of the terminal air conditioner, and record it as G. T*k This can include the terminal air conditioner speed and outlet air temperature;

[0175] Determine the historical energy consumption value of the refrigeration equipment and denote it as B. T*p This includes energy consumption of chillers, refrigeration pumps, cooling pumps, and cooling towers.

[0176] Determine the historical total energy consumption of the terminal air conditioner, denoted as C. T It can be the sum of the historical energy consumption of each terminal air conditioner collected at the same collection time point.

[0177] Determine the total historical energy consumption of the computer building, denoted as D.T ;

[0178] Determine historical environmental state data, denoted as E T*q ;

[0179] Identify historical IT load, denoted as F T .

[0180] Step 302: Train the energy consumption prediction model for the refrigeration station;

[0181] Specifically, step 302 includes:

[0182] Construct the energy consumption characteristics of the refrigeration station and record the historical energy consumption values ​​of the refrigeration equipment. T*p Summing over time yields the total historical energy consumption B' of the refrigeration station. T ;

[0183] Historical environmental status data E T*q Historical operating status data of refrigeration equipment A T*k By merging, we obtain the input AE of the refrigeration station energy consumption prediction model. T*(q+k) ;

[0184] With AE T*(q+k) As input, B' T As the output, it is input into the LSTM model for training, resulting in the energy consumption prediction model for the cooling station, denoted as M1.

[0185] Step 303: Train the terminal air conditioning energy consumption prediction model;

[0186] Specifically, step 303 includes:

[0187] Constructing the energy consumption characteristics of terminal air conditioners, and using historical operating status data of terminal air conditioners (G) T*k Historical environmental status data E T*q Historical IT load F T By merging data along the time dimension, a new historical status data of terminal air conditioners, GEF, is formed. T*(k+q+1) ;

[0188] With GEF T*(k+q+1) As input, the historical total energy consumption of the terminal air conditioning system, C T As output, an LSTM model is trained to obtain a terminal air conditioning energy consumption prediction model, denoted as M2.

[0189] Step 304: Training the total energy consumption model of the computer building;

[0190] Specifically, step 304 includes:

[0191] The historical total energy consumption of the refrigeration station B' T Historical total energy consumption of terminal air conditioning C TTotal historical energy consumption of the building (D) T The average values ​​are calculated over time to obtain the average energy consumption of the refrigeration station (b), the average energy consumption of the terminal air conditioners (c), and the average energy consumption of the computer building (d) during the collection period; b / d and c / d are calculated to obtain the energy consumption ratio of the refrigeration station and the energy consumption ratio of the terminal air conditioners.

[0192] During the training of models M1 and M2, their hidden layer data are extracted and named H1 and H2 respectively. b / d and c / d are used as initial weights, and fully connected layers are constructed for H1 and H2 respectively. An LSTM model is then connected after each fully connected layer to store the hidden layer data H1, H2, and historical environment state data E. T*q Historical IT load F T Combined as input, the total historical energy consumption of the computer building, D T As output, a fully connected layer + LSTM model is trained to obtain the total energy consumption model of the building, denoted as M3.

[0193] Figure 4 A flowchart illustrating a data center energy-saving method provided as an application embodiment of the present invention; as shown below. Figure 4 As shown, the method employs Figure 3 The methods described above train models M1, M2, and M3, from which data center energy-saving solutions are derived. These methods include:

[0194] Step 401: Collect real-time data, including: data on the operating status of the data center's cooling station equipment (a) 1*k Terminal air conditioning operation status data g 1*k Environmental status data e 1*q Total IT load f.

[0195] Step 402: Using the adjustable parameter as the control parameter, calculate its upper and lower limits in historical data, and expand the upper and lower limits by 5% respectively.

[0196] Specifically, the calculation of the upper and lower limits in historical data and the expansion of the upper and lower limits by 5% respectively includes: finding the maximum value m1 and minimum value m2 of each control parameter from historical data, and setting the upper and lower limits of the parameter space to m1*1.05 and m2*0.95 respectively.

[0197] In a certain data center, manually adjustable parameters include: chiller cooling water outlet temperature, chiller chilled water outlet temperature, chilled water pump frequency, cooling pump frequency, cooling tower frequency, cooling tower outlet temperature, distributor water pressure, and terminal air conditioner speed.

[0198] Step 403: Use a genetic algorithm to generate N combinations of control parameters {n1, n2, ..., n} based on historical data. NThe range of values ​​for each control parameter is [m1*1.05, m2*0.95].

[0199] Step 404: Take the total energy consumption of the terminal building as the objective function, that is, the optimization objective is to minimize the output of M3, and build a genetic algorithm model.

[0200] Specifically, for each combination of control parameters, the energy consumption of the chiller station and the terminal air conditioning system under that combination are calculated. The hidden layers of the chiller station energy consumption prediction model and the terminal air conditioning system energy consumption prediction model are output respectively. The two hidden layers are concatenated with the real-time environmental status, IT load, and time, respectively, and input into the total energy consumption model of the computer building to obtain the total energy consumption prediction result of the computer building. From the N combinations of control parameters, the control parameter combination that minimizes the total energy consumption M3 of the computer building is selected as the output scheme.

[0201] The reasoning process of M3 consists of the following steps:

[0202] Step 4041: Retrieve the current operating status data of each device in the current cooling station, the current environmental status data of the current data center, and the current total IT load of the current computer room, excluding control parameters. For each generated combination of control parameters, proceed to steps 4042-4044.

[0203] Step 4042: Combine the retrieved current status data (excluding terminal air conditioners), the current data center environmental status data, and the control parameters such as chiller cooling water outlet temperature, chiller chilled water outlet temperature, chilled water pump frequency, cooling pump frequency, cooling tower frequency, cooling tower outlet water temperature, and distributor water pressure, and input them into the refrigeration plant energy consumption prediction model to obtain the future energy consumption of the refrigeration plant. Simultaneously, output the hidden layer data h. 1i , i∈{1,N}.

[0204] Step 4043: Combine the retrieved current time data on the terminal air conditioner's outlet air temperature, total IT load, and environmental status with the terminal air conditioner's rotational speed from the control parameters, and input this data into the terminal air conditioner energy consumption prediction model to obtain the future terminal air conditioner energy consumption. Simultaneously, output the hidden layer data h. 2i , i∈{1,N}.

[0205] Step 4044, h 1i h 2i The current environmental status data and total IT load are combined to form a new set of data, which is then input into the total energy consumption model of the computer room to obtain the total energy consumption m of the computer room at future time.

[0206] Step 405: Obtain the total energy consumption of the building for each set of control parameter combinations through steps 4042-4044, and compare the total energy consumption of the building corresponding to each set of control parameter combinations.

[0207] Step 406: Determine the specific values ​​of each control parameter in the control parameter combination corresponding to the minimum total energy consumption of the computer room, and output the energy-saving scheme; the energy-saving scheme shall at least include the specific values ​​of each control parameter in the control parameter combination corresponding to the minimum total energy consumption of the computer room.

[0208] The method provided in this invention adopts a strategy of treating refrigeration equipment and terminal air conditioning as a whole for energy saving. This avoids the problem that if only the energy consumption of the refrigeration unit or the terminal air conditioning is considered, reducing the energy consumption of only one of them may increase the energy consumption of the other, thus making it impossible to achieve energy saving.

[0209] Figure 5 This is a schematic diagram of the structure of a data center energy-saving device provided in an embodiment of the present invention; as shown below. Figure 5 As shown, the device, applied to a server, includes: an acquisition module, a first processing module, a second processing module, and a third processing module; wherein,

[0210] The acquisition module is used to acquire environmental status data, total IT load, and at least one dataset; each of the at least one dataset includes: a first sub-dataset and a second sub-dataset; the first sub-dataset includes: data of at least one first control parameter and data of at least one second parameter; the second sub-dataset includes: data of at least one second control parameter and data of at least one third parameter.

[0211] The first processing module is configured to, for each of the at least one datasets, input the environmental status data and the first sub-dataset of each dataset into a preset cooling station energy consumption prediction model to obtain the first hidden layer data of the cooling station energy consumption prediction model; and input the environmental status data, total IT load and the second sub-dataset of each dataset into a preset terminal air conditioning energy consumption prediction model to obtain the second hidden layer data of the terminal air conditioning energy consumption prediction model.

[0212] The second processing module is used to input the first hidden layer data, the second hidden layer data, the environmental status data, and the total IT load corresponding to each dataset into a preset total energy consumption model of the computer room, so as to obtain the total energy consumption of the computer room corresponding to each dataset.

[0213] The third processing module is used to determine the target dataset corresponding to the total energy consumption of the building that meets the preset conditions, and to determine an energy-saving scheme based on the target dataset; the energy-saving scheme is used to indicate the method of adjusting the at least one first control parameter and / or the at least one second control parameter.

[0214] In some embodiments, the apparatus further includes: a preprocessing module;

[0215] The preprocessing module is used to obtain a first training sample set; the first training sample set includes: at least one first training set and a first label corresponding to each first training set in the at least one first training set; the first label represents the total energy consumption of the historical cooling station corresponding to the first training set;

[0216] The first training sample set is trained using the first neural network to obtain the trained first neural network, which serves as the energy consumption prediction model for the refrigeration station.

[0217] The first training set includes: historical environmental state data, historical data of at least one first control parameter, and historical data of at least one second parameter.

[0218] In some embodiments, the preprocessing module is further configured to obtain a second training sample set; the first training sample set includes: at least one second training set and a second label corresponding to each second training set in the at least one second training set; the second label represents the historical total energy consumption of the terminal air conditioner corresponding to the second training set;

[0219] The second training sample set is trained using the second neural network to obtain the trained second neural network, which serves as the terminal air conditioner energy consumption prediction model.

[0220] The second training set includes: historical environmental status data, historical total IT load, historical data of at least one second control parameter, and historical data of at least one third parameter.

[0221] In some embodiments, the preprocessing module is further configured to extract historical first hidden layer data of the refrigeration station energy consumption prediction model and historical second hidden layer data of the terminal air conditioning energy consumption prediction model during the training of the refrigeration station energy consumption prediction model and the terminal air conditioning energy consumption prediction model.

[0222] Determine the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the computer building; determine a first ratio based on the historical energy consumption of the refrigeration station and the historical total energy consumption of the computer building; and determine a second ratio based on the historical total energy consumption of the terminal air conditioning and the historical total energy consumption of the computer building.

[0223] Using the first ratio as the initial weight of the historical first hidden layer data and the second ratio as the initial weight of the historical second hidden layer data, fully connected layers are constructed on the historical first hidden layer data and the historical second hidden layer data, respectively. The constructed fully connected layers are then connected to the neural network to obtain the target training model.

[0224] The target training model is trained by taking the historical first hidden layer data, the historical second hidden layer data, the historical environmental status data, and the historical total IT load as inputs and the historical total energy consumption of the computer room as outputs. The trained target training model is then used as the total energy consumption model of the computer room.

[0225] In some embodiments, the preprocessing module is used to average the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the computer building determined within the collection time, to obtain the average historical total energy consumption of the refrigeration station, the average historical total energy consumption of the terminal air conditioning, and the average historical total energy consumption of the computer building.

[0226] The first ratio is obtained by dividing the average total energy consumption of the historical refrigeration station by the average total energy consumption of the historical computer building.

[0227] The second ratio is obtained by dividing the historical average total air conditioning energy consumption of the terminal by the historical average total energy consumption of the building.

[0228] In some embodiments, the acquisition module is further configured to determine adjustable parameters as control parameters; the control parameters include: a first control parameter and a second control parameter;

[0229] Query the historical data of the control parameter to determine the historical maximum and minimum values ​​of the control parameter;

[0230] Determine a first multiple of the historical maximum value as the first limit;

[0231] Determine the second multiple of the historical minimum value as the second limit value;

[0232] Accordingly, the acquisition module is used to use a genetic algorithm to determine the combination of control parameters for each dataset based on the first and second limits of each control parameter;

[0233] The combination of control parameters includes: at least one first control parameter and at least one second control parameter.

[0234] In some embodiments, the energy-saving scheme includes: at least one device to be adjusted, and an adjustment value of at least one control parameter for each of the at least one device to be adjusted;

[0235] The device further includes: a communication module; the communication module is used to send the energy-saving scheme to the group control system; the group control system is used to adjust the device to be adjusted according to the energy-saving scheme so that the control parameters of the device to be adjusted reach the adjustment values ​​of the corresponding control parameters.

[0236] It should be noted that the data center energy-saving device provided in the above embodiments is only illustrated by the division of the above program modules when implementing the corresponding data center energy-saving method. In actual applications, the above processing can be assigned to different program modules as needed, that is, the internal structure of the application server can be divided into different program modules to complete all or part of the processing described above. In addition, the device and the corresponding method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0237] Figure 6 This is a schematic diagram of another data center energy-saving device provided in an embodiment of the present invention, as shown below. Figure 6 As shown, the data center energy-saving device 60 includes: a processor 601 and a memory 602 for storing computer programs capable of running on the processor; when the processor 601 runs the computer program, it performs the following: acquiring environmental status data, total IT load, and at least one dataset; each of the at least one dataset includes: a first sub-dataset and a second sub-dataset; the first sub-dataset includes: data of at least one first control parameter and data of at least one second parameter; the second sub-dataset includes: data of at least one second control parameter and data of at least one third parameter;

[0238] For each of the at least one datasets, the environmental status data and the first subset of each dataset are input into a preset cooling station energy consumption prediction model to obtain the first hidden layer data of the cooling station energy consumption prediction model; the environmental status data, total IT load and the second subset of each dataset are input into a preset terminal air conditioning energy consumption prediction model to obtain the second hidden layer data of the terminal air conditioning energy consumption prediction model.

[0239] Input the first hidden layer data, the second hidden layer data, the environmental status data, and the total IT load corresponding to each dataset into a preset total energy consumption model for the computer room to obtain the total energy consumption of the computer room corresponding to each dataset.

[0240] A target dataset corresponding to the total energy consumption of the building that meets preset conditions is determined, and an energy-saving scheme is determined based on the target dataset; the energy-saving scheme is used to indicate the method of adjusting the at least one first control parameter and / or the at least one second control parameter.

[0241] Specifically, the data center energy-saving device can also perform the following: Figure 1 The method shown is the same as Figure 1 The data center energy-saving method embodiments shown are based on the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0242] In practical applications, the data center energy-saving device 60 may further include at least one network interface 603. The various components of the data center energy-saving device 60 are coupled together via a bus system 604. It is understood that the bus system 604 is used to realize the connection and communication between these components. In addition to a data bus, the bus system 604 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 6 All buses are labeled as bus system 604. The number of processors 601 can be at least one. Network interface 603 is used for wired or wireless communication between the data center energy-saving device 60 and other devices.

[0243] The memory 602 in this embodiment of the invention is used to store various types of data to support the operation of the data center energy-saving device 60.

[0244] The methods disclosed in the above embodiments of the present invention can be applied to processor 601, or implemented by processor 601. Processor 601 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 601 or by instructions in the form of software. The processor 601 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 601 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the methods disclosed in the embodiments of the present invention can be directly manifested as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in memory 602. Processor 601 reads the information in memory 602 and combines its hardware to complete the steps of the aforementioned method.

[0245] In an exemplary embodiment, the data center energy-saving device 60 may be implemented by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers (MCUs), microprocessors, or other electronic components to perform the aforementioned method.

[0246] This invention also provides a computer-readable storage medium storing a computer program thereon; the computer-readable storage medium is applied to a server, and the computer program, when run by a processor, executes: acquiring environmental status data, total IT load, and at least one dataset; each of the at least one dataset includes: a first subset and a second subset; the first subset includes: data for at least one first control parameter and data for at least one second parameter; the second subset includes: data for at least one second control parameter and data for at least one third parameter;

[0247] For each of the at least one datasets, the environmental status data and the first subset of each dataset are input into a preset cooling station energy consumption prediction model to obtain the first hidden layer data of the cooling station energy consumption prediction model; the environmental status data, total IT load and the second subset of each dataset are input into a preset terminal air conditioning energy consumption prediction model to obtain the second hidden layer data of the terminal air conditioning energy consumption prediction model.

[0248] Input the first hidden layer data, the second hidden layer data, the environmental status data, and the total IT load corresponding to each dataset into a preset total energy consumption model for the computer room to obtain the total energy consumption of the computer room corresponding to each dataset.

[0249] A target dataset corresponding to the total energy consumption of the building that meets preset conditions is determined, and an energy-saving scheme is determined based on the target dataset; the energy-saving scheme is used to indicate the method of adjusting the at least one first control parameter and / or the at least one second control parameter.

[0250] Specifically, the computer program can also perform actions such as Figure 1 The method shown is the same as Figure 1The data center energy-saving method embodiments shown are based on the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0251] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0252] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0253] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0254] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0255] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0256] It should be noted that terms such as "first" and "second" are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0257] Furthermore, the technical solutions described in the embodiments of this application can be combined arbitrarily without conflict.

[0258] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A data center energy-saving method, characterized in that, The method includes: Acquire environmental status data, total IT load of Internet devices, and at least one dataset; each of the at least one dataset includes: a first sub-dataset and a second sub-dataset; the first sub-dataset includes: data of at least one first control parameter and data of at least one second parameter; the second sub-dataset includes: data of at least one second control parameter and data of at least one third parameter; For each of the at least one datasets, the environmental status data and the first subset of each dataset are input into a preset cooling station energy consumption prediction model to obtain the first hidden layer data of the cooling station energy consumption prediction model; the environmental status data, total IT load and the second subset of each dataset are input into a preset terminal air conditioning energy consumption prediction model to obtain the second hidden layer data of the terminal air conditioning energy consumption prediction model. Input the first hidden layer data, the second hidden layer data, the environmental status data, and the total IT load corresponding to each dataset into a preset total energy consumption model for the computer room to obtain the total energy consumption of the computer room corresponding to each dataset. A target dataset corresponding to the total energy consumption of the building that meets preset conditions is determined, and an energy-saving scheme is determined based on the target dataset; the energy-saving scheme is used to indicate the method of adjusting the at least one first control parameter and / or the at least one second control parameter.

2. The method according to claim 1, characterized in that, Generating the preset energy consumption prediction model for the refrigeration station includes: Obtain a first training sample set; the first training sample set includes: at least one first training set and a first label corresponding to each first training set in the at least one first training set; the first label represents the historical total energy consumption of the cooling station corresponding to the first training set; The first training sample set is trained using the first neural network to obtain the trained first neural network, which serves as the energy consumption prediction model for the refrigeration station. The first training set includes: historical environmental state data, historical data of at least one first control parameter, and historical data of at least one second parameter.

3. The method according to claim 1, characterized in that, Generating the preset terminal air conditioning energy consumption prediction model includes: Obtain a second training sample set; the second training sample set includes: at least one second training set and a second label corresponding to each second training set in the at least one second training set; the second label represents the historical total energy consumption of the terminal air conditioner corresponding to the second training set; The second training sample set is trained using the second neural network to obtain the trained second neural network, which serves as the terminal air conditioner energy consumption prediction model. The second training set includes: historical environmental status data, historical total IT load, historical data of at least one second control parameter, and historical data of at least one third parameter.

4. The method according to claim 2 or 3, characterized in that, Generating the preset total energy consumption model for the terminal building includes: During the training of the energy consumption prediction model of the refrigeration station and the energy consumption prediction model of the terminal air conditioner, the historical first hidden layer data of the energy consumption prediction model of the refrigeration station and the historical second hidden layer data of the energy consumption prediction model of the terminal air conditioner are extracted. Determine the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the computer building; determine a first ratio based on the historical energy consumption of the refrigeration station and the historical total energy consumption of the computer building; and determine a second ratio based on the historical total energy consumption of the terminal air conditioning and the historical total energy consumption of the computer building. Using the first ratio as the initial weight of the historical first hidden layer data and the second ratio as the initial weight of the historical second hidden layer data, fully connected layers are constructed on the historical first hidden layer data and the historical second hidden layer data, respectively. The constructed fully connected layers are then connected to the neural network to obtain the target training model. The target training model is trained by taking the historical first hidden layer data, the historical second hidden layer data, the historical environmental status data, and the historical total IT load as inputs and the historical total energy consumption of the computer room as outputs. The trained target training model is then used as the total energy consumption model of the computer room.

5. The method according to claim 4, characterized in that, The process of determining the historical total energy consumption of the chiller plant, the historical total energy consumption of terminal air conditioning units, and the historical total energy consumption of the server room, and determining a first ratio based on the historical energy consumption of the chiller plant and the historical total energy consumption of the server room, and determining a second ratio based on the historical total energy consumption of terminal air conditioning units and the historical total energy consumption of the server room, includes: The average values ​​of the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the computer building are obtained by averaging the historical total energy consumption of the refrigeration station, the historical total energy consumption of the terminal air conditioning, and the historical total energy consumption of the computer building. The first ratio is obtained by dividing the average total energy consumption of the historical refrigeration station by the average total energy consumption of the historical computer building. The second ratio is obtained by dividing the historical average total air conditioning energy consumption of the terminal by the historical average total energy consumption of the building.

6. The method according to claim 1, characterized in that, The method further includes: An adjustable parameter is determined as a control parameter; the control parameter includes: a first control parameter and a second control parameter; Query the historical data of the control parameter to determine the historical maximum and minimum values ​​of the control parameter; Determine a first multiple of the historical maximum value as the first limit; Determine the second multiple of the historical minimum value as the second limit value; Accordingly, obtaining at least one dataset includes: Using a genetic algorithm, the combination of control parameters for each dataset is determined based on the first and second limits of each control parameter; The combination of control parameters includes: at least one first control parameter and at least one second control parameter.

7. The method according to claim 1, characterized in that, The energy-saving scheme includes: at least one device to be adjusted, and adjustment values ​​of at least one control parameter for each of the at least one device to be adjusted; The method further includes: The energy-saving scheme is sent to the group control system; the group control system is used to adjust the device to be adjusted according to the energy-saving scheme so that the control parameters of the device to be adjusted reach the corresponding adjustment values.

8. A data center energy-saving device, characterized in that, The device includes: an acquisition module, a first processing module, a second processing module, and a third processing module; wherein, The acquisition module is used to acquire environmental status data, total IT load, and at least one dataset; each of the at least one dataset includes: a first sub-dataset and a second sub-dataset; the first sub-dataset includes: data of at least one first control parameter and data of at least one second parameter; the second sub-dataset includes: data of at least one second control parameter and data of at least one third parameter. The first processing module is configured to, for each of the at least one datasets, input the environmental status data and the first sub-dataset of each dataset into a preset cooling station energy consumption prediction model to obtain the first hidden layer data of the cooling station energy consumption prediction model; and input the environmental status data, total IT load and the second sub-dataset of each dataset into a preset terminal air conditioning energy consumption prediction model to obtain the second hidden layer data of the terminal air conditioning energy consumption prediction model. The second processing module is used to input the first hidden layer data, the second hidden layer data, the environmental status data, and the total IT load corresponding to each dataset into a preset total energy consumption model of the computer room, so as to obtain the total energy consumption of the computer room corresponding to each dataset. The third processing module is used to determine the target dataset corresponding to the total energy consumption of the building that meets the preset conditions, and to determine an energy-saving scheme based on the target dataset; the energy-saving scheme is used to indicate the method of adjusting the at least one first control parameter and / or the at least one second control parameter.

9. A data center energy-saving device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images