Control methods, devices and related equipment for computer room cooling systems

Abstract

This application provides a control method, apparatus, and related equipment for a data center cooling system. The method includes: acquiring N first user data points matching a target data center; calculating the load forecast value of the target data center based on the N first user data points; inputting the load forecast value, environmental variables corresponding to the target data center, runtime parameters corresponding to the target data center, and switching parameters corresponding to the target data center into a target deep learning model for prediction to obtain a cooling control strategy; and controlling the cooling system of the target data center based on the cooling control strategy. This application improves the control effect of the cooling system by acquiring N first user data points matching the target data center, calculating the load forecast value of the target data center based on the N first user data points, and inputting the load forecast value, environmental variables, runtime parameters, and switching parameters into a target deep learning model for prediction to obtain a cooling control strategy.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and specifically to a control method, device, and related equipment for a computer room cooling system. Background Technology

[0002] Current energy-saving solutions for communication equipment rooms primarily utilize AI algorithms to build a power load model of the room based on historical power load data of various devices, operating hours of cooling equipment, historical meteorological data, and historical temperature data. This model is then used to establish the optimal air conditioning and cooling strategy. Existing solutions build upon this foundation by collecting more data on temperature around the server racks and operating data of the cooling equipment, and by optimizing more specific aspects.

[0003] However, existing technologies passively issue cooling system control strategies based on the operating status and heat generation of equipment in the computer room, resulting in poor control effects of existing cooling system control strategies. Summary of the Invention

[0004] This invention provides a control method, device, and related equipment for a computer room cooling system, which solves the problem of poor control effect of the cooling system in the prior art.

[0005] To solve the above problems, the present invention is implemented as follows:

[0006] In a first aspect, embodiments of this application provide a control method for a computer room cooling system, the method comprising:

[0007] Obtain N first user data that match the target data center. The first user data is the internet access behavior data generated by the corresponding user based on the user account during broadband use. N is a positive integer.

[0008] The load forecast value of the target computer room is obtained by calculating the N first user data, and the load forecast value is the cooling load required in the target computer room;

[0009] The load forecast, the environmental variables corresponding to the target data center, the operating time parameters corresponding to the target data center, and the switching parameters corresponding to the target data center are input into the target deep learning model for prediction to obtain a cooling control strategy. The environmental variables include the temperature and humidity parameters corresponding to the target data center. The cooling control strategy is used to adjust the operating parameters corresponding to the cooling system of the target data center. The operating time parameter is the service runtime of the target data center, and the switching parameters are the number of times the cooling system of the target data center is switched on and off within the service runtime.

[0010] The cooling system of the target computer room is controlled based on the cooling control strategy.

[0011] Optionally, obtaining N first user data points matching the target data center includes:

[0012] Obtain M second user data, each of which corresponds to one of M user accounts. The second user data is the internet behavior data generated by the corresponding user during broadband use based on the user account. M is greater than or equal to N, and M is a positive integer.

[0013] The M second user data are parsed to obtain M parsing results. The M parsing results correspond one-to-one with the M second user data. The parsing results are used to indicate whether the data center information corresponding to the second user data matches the data center information corresponding to the target data center.

[0014] Based on the M parsing results, N first user data are determined from the M second user data, and the data center information indicated by the parsing results corresponding to the first user data matches the data center information corresponding to the target data center.

[0015] Optionally, parsing the M pieces of second user data to obtain M parsing results includes:

[0016] The M sets of second user data are cleaned and parsed to obtain the M parsing results. Each of the M parsing results includes a first field and a second field. The first field indicates the user access line identifier information of the corresponding second user data, and the second field indicates the user access location information of the corresponding second user data. The user access line identifier information indicates the device identifier of the user accessing the network, and the user access location information indicates the geographical location of the user accessing the network. If the device identifier and the geographical location are the same as the data center information corresponding to the target data center, the data center information indicated by the parsing result matches the data center information corresponding to the target data center. If the device identifier and the geographical location are different from the data center information corresponding to the target data center, the data center information indicated by the parsing result does not match the data center information corresponding to the target data center.

[0017] The step of determining the N first user data from the M second user data based on the M parsing results includes:

[0018] For each piece of second user data, data center information corresponding to each piece of second user data is generated based on the first field and the second field corresponding to the second user data. The data center information includes the device identifier and the geographical location.

[0019] Based on the data center information corresponding to each of the second user data, the N first user data are determined from the M second user data.

[0020] Optionally, the step of calculating the load forecast value of the target data room from the N first user data includes:

[0021] Iterate through the N first user data sets to obtain the internet access information corresponding to each first user data set, including the user's internet access time period and the user's internet traffic.

[0022] Obtain the target time period and data center bandwidth capacity corresponding to the target data center, wherein the target time period is the service opening time period of the target data center, and the data center bandwidth capacity is the maximum bandwidth that the target data center can provide;

[0023] The bandwidth utilization rate of the target data center is obtained by calculating the bandwidth utilization rate of the target data center based on the N sets of internet access information, the target time period, and the data center bandwidth capacity.

[0024] The bandwidth utilization rate is determined as the predicted load value for the target data center.

[0025] Optionally, before inputting the load forecast value, the environmental variables corresponding to the target computer room, and other variables corresponding to the target computer room into the target deep learning model for prediction to obtain the cooling control strategy, the method further includes:

[0026] Obtain a training dataset, which includes historical load prediction values, historical environmental variables, historical running time parameters, historical switching parameters, and training labels corresponding to the target data center. The training labels are the historical cooling control strategies corresponding to the target data center. The historical load prediction values, historical environmental variables, historical running time parameters, and historical switching parameters are data within the same historical time period.

[0027] The deep learning model is trained based on the training dataset to obtain the target deep learning model.

[0028] Optionally, the refrigeration control strategy includes at least one of the following: refrigeration equipment supply / return air temperature regulation strategy, refrigeration equipment fan speed adjustment strategy, refrigeration equipment compressor speed adjustment strategy, refrigeration equipment on / off quantity control strategy, humidification / dehumidification control strategy, refrigerant switching strategy, fresh air switching strategy, dual cold source switching strategy, and compressor load setting strategy.

[0029] Optionally, the environmental variables include at least one of the following: Internet power, total power of the computer room, the ratio of energy consumed to energy consumed by the Internet load, supply air temperature of the refrigeration equipment, return air temperature of the refrigeration equipment, cold aisle temperature and humidity, hot aisle temperature and humidity, outdoor temperature and humidity, refrigeration equipment start / stop status, fan start / stop status, compressor start / stop status, fan speed, compressor speed, and outdoor temperature.

[0030] Secondly, embodiments of this application provide a control device for a computer room cooling system, the device comprising:

[0031] The acquisition module is used to acquire N first user data that match the target computer room. The first user data is the internet access behavior data generated by the corresponding user during the broadband usage process based on the user account, and N is a positive integer.

[0032] The calculation module is used to calculate the load forecast value of the target computer room based on the N first user data, and the load forecast value is the cooling load required in the target computer room;

[0033] The prediction module is used to input the predicted load value, the environmental variables corresponding to the target data center, the runtime parameters corresponding to the target data center, and the switching parameters corresponding to the target data center into the target deep learning model for prediction, and obtain a cooling control strategy. The environmental variables include the temperature and humidity parameters corresponding to the target data center. The cooling control strategy is used to adjust the runtime parameters corresponding to the cooling system of the target data center. The runtime parameters are the service runtime of the target data center, and the switching parameters are the number of times the cooling system of the target data center is switched on and off within the service runtime.

[0034] The control module is used to control the cooling system of the target computer room based on the cooling control strategy.

[0035] Thirdly, this application also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method described in the first aspect above.

[0036] Fourthly, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described in the first aspect above.

[0037] Fifthly, this application also provides a computer program product, including computer instructions that, when executed by a processor, implement the steps of the method described in the first aspect above.

[0038] This application provides a control method, apparatus, and related equipment for a data center cooling system. The method includes: acquiring N first user data points matching a target data center; calculating the load forecast value of the target data center based on the N first user data points; inputting the load forecast value, environmental variables corresponding to the target data center, runtime parameters corresponding to the target data center, and switching parameters corresponding to the target data center into a target deep learning model for prediction to obtain a cooling control strategy; and controlling the cooling system of the target data center based on the cooling control strategy. This application improves the control effect of the cooling system by acquiring N first user data points matching the target data center, calculating the load forecast value of the target data center based on the N first user data points, and inputting the load forecast value, environmental variables, runtime parameters, and switching parameters into a target deep learning model for prediction to obtain a cooling control strategy. Attached Figure Description

[0039] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description of the present invention will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 A flowchart illustrating a control method for a computer room cooling system provided in an embodiment of this application;

[0041] Figure 2 This is a schematic diagram of the cooling control strategy provided in the embodiments of this application;

[0042] Figure 3 A flowchart of the parsing algorithm for logicalportno provided in this application embodiment;

[0043] Figure 4 This is a schematic diagram of the structure of the DCN model provided in the embodiments of this application;

[0044] Figure 5 A schematic diagram of the structure of a control device for a computer room cooling system provided in this application embodiment;

[0045] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0046] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0047] The terms "first," "second," etc., used in the embodiments of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices. Additionally, the use of "and / or" in this application indicates at least one of the connected objects, such as A and / or B and / or C, representing seven possibilities: including A alone, B alone, C alone, and the presence of both A and B, both B and C, both A and C, and the presence of A, B, and C.

[0048] See Figure 1 , Figure 1 This is a flowchart illustrating the control method for a computer room cooling system provided in an embodiment of this application. Figure 1 As shown, the control method for the computer room cooling system may include the following steps:

[0049] Step 101: Obtain N first user data that match the target data center. The first user data is the internet access behavior data generated by the corresponding user during the broadband usage process based on the user account. N is a positive integer.

[0050] In this embodiment, the target equipment room is a equipment room containing an Optical Line Terminal (OTL) device. An OLT is a device used in fiber optic access networks, primarily responsible for converting fiber optic signals into electrical signals and performing data processing and management. It should be noted that this application predicts the load on the equipment room based on the internet access behavior of broadband users to adjust the air conditioning accordingly. Therefore, selecting an equipment room containing an OLT device, especially one primarily carrying broadband services, will result in better energy-saving air conditioning performance.

[0051] The N first-user data points are internet access behavior data generated when a user uses broadband through their user account. For example, N first-user data points can be extracted from Radius broadband AAA call detail record (CDR) data. AAA is a network access control security management framework embedded in the process of a user terminal initiating a PPPoE internet access request. It determines which users can access the network and which resources or services they can access. AAA includes three elements: Authentication, Authorization, and Accounting, and can be implemented through various protocols, including RADIUS, TACACS, LDAP, and Diameter. RADIUS (Remote Authentication Dial-Up User Service) is currently the standard protocol. The RADIUS protocol can be divided into authentication and accounting protocols, defined by IETF RFC 2865 and RFC 2866, respectively.

[0052] The Radius home broadband AAA call detail records (CDRs) contain customer data and call detail record data. Each time a home broadband user accesses the internet, a CDR record is generated. By extracting the user account, start time, end time, uplink traffic, downlink traffic, the IP address of the user's uplink OLT, and the IP address of the user's uplink BRAS from each CDR record, a home broadband user internet behavior model can be established. Based on this model, using the user account field as a unique identifier, the internet behavior data of a single home broadband user over a period of time can be obtained, i.e., the first user data.

[0053] Step 102: Calculate the load forecast value of the target computer room by processing the N first user data. The load forecast value is the cooling load required in the target computer room.

[0054] In this embodiment, the acquired N sets of first user data are used to calculate the load forecast value of the target computer room. The load forecast value is the cooling load required in the target computer room. Cooling load refers to the cooling capacity required to maintain a specific environment (such as an indoor space), expressed as the heat that needs to be removed from that space. Cooling load is typically measured in kilowatts (kW) or tons of cooling capacity (TR) and is used to calculate and design the capacity of air conditioning and refrigeration systems.

[0055] Step 103: Input the load forecast value, the environmental variables corresponding to the target data center, the operating time parameters corresponding to the target data center, and the switching parameters corresponding to the target data center into the target deep learning model for prediction to obtain a cooling control strategy. The environmental variables include the temperature and humidity parameters corresponding to the target data center. The cooling control strategy is used to adjust the operating parameters corresponding to the cooling system of the target data center. The operating time parameter is the service runtime of the target data center, and the switching parameters are the number of times the cooling system of the target data center is switched on and off within the service runtime.

[0056] In this embodiment, after obtaining the load forecast value, it is also necessary to obtain the environmental variables, the running time parameters, and the switching parameters corresponding to the target computer room. These parameters are then input into the target deep learning model for prediction to obtain the cooling control strategy.

[0057] Specifically, the environmental variables include the temperature and humidity parameters corresponding to the target computer room. Specifically, the environmental variables include at least one of the following: Internet power, total power of the computer room, the ratio of energy consumed to energy consumed by the Internet load, supply air temperature of the cooling equipment, return air temperature of the cooling equipment, cold aisle temperature and humidity, hot aisle temperature and humidity, outdoor temperature and humidity, start / stop status of the cooling equipment, start / stop status of the fan, start / stop status of the compressor, fan speed, compressor speed, and outdoor temperature.

[0058] The runtime parameter is the service runtime of the target data center, i.e., the normal operating time of the target data center. The switching parameter is the number of times the cooling system of the target data center is switched on and off within the service runtime, for example, the number of times each cooling unit is switched on and off.

[0059] The cooling control strategy is used to adjust the operating parameters of the cooling system of the target computer room, so that the target computer room can meet the current cooling load requirements while saving energy.

[0060] Step 104: Control the cooling system of the target computer room based on the cooling control strategy.

[0061] In this embodiment, the cooling system of the target computer room is controlled according to a cooling control strategy. Specifically, as follows: Figure 2 As shown, Figure 2This diagram illustrates the cooling control strategy in this embodiment. The cooling control strategy includes at least one of the following: supply / return air temperature control strategy for cooling equipment, fan speed adjustment strategy for cooling equipment, compressor speed adjustment strategy for cooling equipment, on / off control strategy for cooling equipment, humidification / dehumidification control strategy, refrigerant switching strategy, fresh air switching strategy, dual cold source switching strategy, and compressor load setting strategy. These control variables directly affect the computer room environment, thereby achieving the control objective of maintaining stable computer room temperature and optimizing energy consumption.

[0062] By controlling the cooling system of the target computer room through a cooling control strategy, the system combines existing data room environmental variables, time and on / off variables, and load forecast values. Based on artificial intelligence deep reinforcement learning, it automatically learns and optimizes the control strategy, thereby reducing the power consumption of the computer room cooling system while ensuring stable computer room temperature.

[0063] This application obtains N first user data matching the target computer room, calculates the load forecast value of the target computer room based on the N first user data, inputs the load forecast value, environmental variables, running time parameters and switching parameters into the target deep learning model for prediction, and obtains the cooling control strategy, thereby improving the control effect of the cooling system control strategy on the cooling system.

[0064] In some feasible implementations, optionally, acquiring N first user data points matching the target data center includes:

[0065] Obtain M second user data, each of which corresponds to one of M user accounts. The second user data is the internet behavior data generated by the corresponding user during broadband use based on the user account. M is greater than or equal to N, and M is a positive integer.

[0066] The M second user data are parsed to obtain M parsing results. The M parsing results correspond one-to-one with the M second user data. The parsing results are used to indicate whether the data center information corresponding to the second user data matches the data center information corresponding to the target data center.

[0067] Based on the M parsing results, N first user data are determined from the M second user data, and the data center information indicated by the parsing results corresponding to the first user data matches the data center information corresponding to the target data center.

[0068] In this embodiment, M second user data points are obtained from a certain cell or a certain communication range. Specifically, the M second user data points correspond one-to-one with M user accounts. The second user data points are the internet access behavior data generated by the corresponding user during broadband use based on the user account.

[0069] By parsing M pieces of second user data, M parsing results are obtained. These results determine whether the data center information of the second user data matches the data center information of the target data center. If a match is found, it indicates that the second user data was processed by the target data center. Therefore, N pieces of first user data are identified from the M pieces of second user data. The data center information indicated by the parsing results of these first user data matches the data center information corresponding to the target data center. This means the first user data was transmitted by a single OTL device, and the OTL is located in the target data center. Thus, N pieces of first user data associated with the target data center are identified.

[0070] Optionally, parsing the M pieces of second user data to obtain M parsing results includes:

[0071] The M sets of second user data are cleaned and parsed to obtain the M parsing results. Each of the M parsing results includes a first field and a second field. The first field indicates the user access line identifier information of the corresponding second user data, and the second field indicates the user access location information of the corresponding second user data. The user access line identifier information indicates the device identifier of the user accessing the network, and the user access location information indicates the geographical location of the user accessing the network. If the device identifier and the geographical location are the same as the data center information corresponding to the target data center, the data center information indicated by the parsing result matches the data center information corresponding to the target data center. If the device identifier and the geographical location are different from the data center information corresponding to the target data center, the data center information indicated by the parsing result does not match the data center information corresponding to the target data center.

[0072] The step of determining the N first user data from the M second user data based on the M parsing results includes:

[0073] For each piece of second user data, data center information corresponding to each piece of second user data is generated based on the first field and the second field corresponding to the second user data. The data center information includes the device identifier and the geographical location.

[0074] Based on the data center information corresponding to each of the second user data, the N first user data are determined from the M second user data.

[0075] In this embodiment, when M pieces of second user data are obtained, data cleaning and parsing are performed on the M pieces of second user data to obtain M parsing results. Specifically, the parsing results include a first field and a second field. The first field indicates the user access line identifier information of the corresponding second user data, and the second field indicates the user access location information of the corresponding second user data. The first field is the logicalportno field, and the second field is the nas-identifier field. The user access line identifier information represents the device identifier of the user accessing the network, and the user access location information represents the geographical location of the user accessing the network. If the device identifier and the geographical location are the same as the data center information corresponding to the target data center, then the parsing results are determined to match the data center information of the target data center based on the first and second fields.

[0076] It should be noted that when JiaKuan Radius generates daily call detail records for AAA users, it needs to map each user record to a single OLT to pinpoint the target data center. However, existing AAA call detail records cannot directly identify the OLT device corresponding to each user for two main reasons: first, the call detail records lack an OLT IP field; second, different OLTs in different regions may share the same IP address. Therefore, to uniquely identify a specific OLT device, in addition to obtaining the device's IP address, it is also necessary to obtain the city / district information to which the device belongs. Using the IP and city fields as a composite primary key, a single OLT device can be uniquely identified.

[0077] The parsing of the first field, `logicalportno`, is explained below. An example format for `logicalportno` is: `trunk3 / 0 / 7:2125.18 172.31.251.84 / 0 / 0 / 7 / 0 / 16 / CMDCB238F0FA GP`. First, it checks if the length of `logicalportno` is greater than 30 characters. If `logicalportno` is less than or equal to 30 characters, the process ends, and this call detail record is considered invalid and not added to the OLT peak user record. Then, the positions of each element are retrieved sequentially. If retrieval fails, it is considered an abnormal character, and the process ends. If each element can be successfully retrieved, the OLT_IP can be extracted from this call detail record. Specifically, as shown... Figure 3 As shown, Figure 3 The flowchart shows the parsing algorithm for logicalportno. It determines whether the OLT_ip has been completely captured by obtaining the position of each element in turn.

[0078] The parsing of the second field, nas-identifier, is explained below. An example format for nas-identifier is: HBJZO-MC-CMNET-BAS03-ESN-7750. The string before the first "-" is extracted as the city, nas_city. For example, the parsing result is: HBJZO. Since different cities may share the same OLT_ip, nas_city and OLT_ip are concatenated to form a composite primary key to ensure the uniqueness of the OLT in each call detail record (CDR) record. This is used to associate Radius broadband AAA CDRs with the target data center. For example, nas_cityolt_ip:

[0079] HAJIZ172.31.251.84、

[0080] HAJIZ172.31.247.60

[0081] HAJIZ172.31.247.60.

[0082] By using the first and second fields, it can be accurately determined whether the second user data matches the target data center, thus improving the matching accuracy.

[0083] Optionally, the step of calculating the load forecast value of the target data room from the N first user data includes:

[0084] Iterate through the N first user data sets to obtain the internet access information corresponding to each first user data set, including the user's internet access time period and the user's internet traffic.

[0085] Obtain the target time period and data center bandwidth capacity corresponding to the target data center, wherein the target time period is the service opening time period of the target data center, and the data center bandwidth capacity is the maximum bandwidth that the target data center can provide;

[0086] The bandwidth utilization rate of the target data center is obtained by calculating the bandwidth utilization rate of the target data center based on the N sets of internet access information, the target time period, and the data center bandwidth capacity.

[0087] The bandwidth utilization rate is determined as the predicted load value for the target data center.

[0088] In this embodiment, when calculating the load prediction value of the target data center, the internet access information corresponding to each of the N first user data sets is obtained by traversing the data. The internet access information includes the user's internet access time period and the user's internet traffic. For example, the AAA call detail record (CDR) data within a continuous three-day time range is traversed. This is because the CDR data for the current day may only contain a portion of the CDR records. There is a situation where the device's active end time is a certain time on the next day. Such CDR records are recorded in the CDR data of the next day and will not appear on the current day. Similarly, the CDR records of the previous day are traversed for the same reason. Therefore, to solve the problem of inaccurate CDR segmentation, the data of the last three days is used as the data source. The AAA CDR data within a continuous three-day time range, after being cleaned and parsed, is traversed to obtain relevant information such as the user's active start time, active end time, user uplink traffic, and user downlink traffic on each OLT device. The key information is shown in Table 1 below.

[0089] Table 1

[0090]

[0091]

[0092] In this context, OLT represents the OLT device; begintime represents the start timestamp of the user's activity on the current device; endtime represents the end timestamp of the user's activity on the current device; dataTime represents the time of the current call detail record; sendkbytes represents the user's cumulative uplink traffic during the active period, in KB; and receivekbytes represents the user's cumulative downlink traffic during the active period, in KB. Specifically, the user's internet access time period is determined based on endtime and dataTime, while the user's internet traffic is determined based on sendkbytes and receivekbytes.

[0093] Therefore, after identifying the target data center, the identifiers of all OLT devices in the target data center can be determined. This allows the identification of all broadband internet users carried by the OLT devices in the target data center to be determined based on the Radius AAA call detail record data.

[0094] In this embodiment, the OLT bandwidth utilization rate is determined as the load forecast value of the target data center, wherein, The cooling load of the target data center is measured by OLT bandwidth utilization.

[0095] In this embodiment, based on Radius broadband AAA call detail record data, the load of the data center is predicted by predicting users' internet access behavior. Then, combined with traditional data center environmental variable parameters, cooling equipment parameters, equipment load parameters, and other parameters, a cooling system strategy is formulated. This is a proactive control strategy for the data center cooling system, which improves the control effect.

[0096] Optionally, before inputting the load forecast value, the environmental variables corresponding to the target computer room, and other variables corresponding to the target computer room into the target deep learning model for prediction to obtain the cooling control strategy, the method further includes:

[0097] Obtain a training dataset, which includes historical load prediction values, historical environmental variables, historical running time parameters, historical switching parameters, and training labels corresponding to the target data center. The training labels are the historical cooling control strategies corresponding to the target data center. The historical load prediction values, historical environmental variables, historical running time parameters, and historical switching parameters are data within the same historical time period.

[0098] The deep learning model is trained based on the training dataset to obtain the target deep learning model.

[0099] In this embodiment, the deep learning model is a Data Center Network (DCN), a structure used for designing and architecting data center networks. The main purpose of the DCN model is to provide efficient, high-speed, and reliable data transmission to meet the network performance and scalability requirements of modern data centers.

[0100] Based on a single Radius broadband AAA call detail record (CDR), the exact OLT identifier, start time ts, end time te, and total user traffic (uplink traffic tup + downlink traffic tup) for each AAA CDR can be located. Furthermore, based on all AAA CDR data, the load on the data center from all user actions within a given time window can be obtained, i.e., the OLT bandwidth utilization of the target data center. Associating AAA user data with CDR data based on user ID yields the final DCN model input. Assume user u... i The call details are shown in Table 2 below:

[0101] Table 2

[0102]

[0103]

[0104] Then, the user's upload and download traffic per minute during this period are as follows: and Assume there are n users {u1, u2, ..., u3} under the OLT device. n}, then the uplink and downlink traffic per minute for this OLT device are respectively: and

[0105] The structural diagram of the DCN model in this embodiment is as follows: Figure 4 As shown, it mainly consists of two parts: a Cross network and a deep learning network, which increases the interaction between features. Each cross layer of the Cross network is used to capture the cross relationships between features. Each cross layer performs a cross operation between the output and input of the previous layer, and then uses the result as the input of the next layer. The specific stacking method is as follows:

[0106]

[0107] Where, x l+1 It is the output of the (l+1)th layer; x l It is the output of the l-th layer; w l It is the weight matrix connecting layer l and layer l+1; b l x0 is the bias value; x0 is the initial value.

[0108] A deep network is a fully connected feedforward neural network, with each layer containing multiple neurons. Each layer is constructed as follows: a l+1 =∫(w l a l +b l ), where a l+1 It is the output of the (l+1)th layer of the neural network; a l It is the output of the l-th layer of the neural network; f(·) is the activation function; w l It is the weight matrix connecting layer l and layer l+1; b l It is the bias of the l-th layer, used to adjust the activation value of each neuron to better fit the data.

[0109] In this embodiment, the training dataset includes historical load prediction values, historical environmental variables, historical running time parameters, historical switching parameters, and training labels corresponding to the target data center. The training labels are the historical cooling control strategies corresponding to the target data center. The historical load prediction values, historical environmental variables, historical running time parameters, and historical switching parameters are data within the same historical time period, which can be a three-day period, etc.

[0110] It should be noted that this embodiment uses an "actor-critic" deep reinforcement learning algorithm during the training of the DCN model. The "actor-critic" model includes: Actors: The actor is responsible for generating the control strategy, that is, deciding how to adjust various environmental variables based on the current environmental state. For example, determining the air conditioner's supply air temperature or the fan speed based on the current temperature and power status. Critic: The critic is responsible for evaluating the effectiveness of the current strategy. It measures the effectiveness of the current strategy based on a reward function, that is, providing feedback (reward or penalty) by comparing the actual data center environment state with the expected goal. It is important to note that there are only two types of feedback: positive (reward) or negative (penalty).

[0111] Specifically, the main interactive processes include: Strategy Generation: Based on current environmental variables, the actors generate an overall cooling system execution strategy (such as air conditioning temperature, fan speed, and chiller parameters). Strategy Evaluation: Critics evaluate the effectiveness of this strategy. If the strategy can stabilize the computer room temperature and optimize energy consumption, the system will receive a positive reward; otherwise, it will receive negative feedback. Strategy Adjustment: Based on the critics' feedback, the actors will adjust the strategy to optimize future decisions, thereby improving environmental control through continuous feedback.

[0112] This application obtains N first user data matching the target computer room, calculates the load forecast value of the target computer room based on the N first user data, inputs the load forecast value, environmental variables, running time parameters and switching parameters into the target deep learning model for prediction, and obtains the cooling control strategy, thereby improving the control effect of the cooling system control strategy on the cooling system.

[0113] See Figure 5 , Figure 5 This is a structural diagram of the control device for the computer room cooling system provided in an embodiment of this application. Figure 5 As shown, the control device 500 of the computer room cooling system includes:

[0114] The acquisition module 510 is used to acquire N first user data that match the target computer room. The first user data is the internet access behavior data generated by the corresponding user during the broadband usage process based on the user account. N is a positive integer.

[0115] The calculation module 520 is used to calculate the N first user data to obtain the load prediction value of the target computer room, wherein the load prediction value is the cooling load required in the target computer room;

[0116] The prediction module 530 is used to input the load prediction value, the environmental variables corresponding to the target computer room, the running time parameters corresponding to the target computer room, and the switching parameters corresponding to the target computer room into the target deep learning model for prediction to obtain a cooling control strategy. The environmental variables include the temperature and humidity parameters corresponding to the target computer room. The cooling control strategy is used to adjust the running parameters corresponding to the cooling system of the target computer room. The running time parameter is the service running time of the target computer room, and the switching parameters are the number of times the cooling system of the target computer room is switched on and off within the service running time.

[0117] The control module 540 is used to control the cooling system of the target computer room based on the cooling control strategy.

[0118] Optionally, the computing module 520 includes:

[0119] The first acquisition submodule is used to acquire M second user data, wherein the M second user data correspond one-to-one with M user accounts, and the second user data is the internet access behavior data generated by the corresponding user during broadband use based on the user account, wherein M is greater than or equal to N, and M is a positive integer.

[0120] The parsing submodule is used to parse the M second user data to obtain M parsing results. The M parsing results correspond one-to-one with the M second user data. The parsing results are used to indicate whether the data center information corresponding to the second user data matches the data center information corresponding to the target data center.

[0121] The first determining submodule is used to determine the N first user data from the M second user data based on the M parsing results, wherein the data center information indicated by the parsing results corresponding to the first user data matches the data center information corresponding to the target data center.

[0122] Optionally, the parsing submodule includes:

[0123] The parsing unit is used to perform data cleaning and data parsing on the M second user data to obtain the M parsing results. Each of the M parsing results includes a first field and a second field. The first field is used to indicate the user access line identification information of the corresponding second user data, and the second field is used to indicate the user access location information of the corresponding second user data. The user access line identification information is used to represent the device identifier of the user accessing the network, and the user access location information is used to represent the geographical location of the user accessing the network. If the device identifier and the geographical location are the same as the data center information corresponding to the target data center, the data center information indicated by the parsing result matches the data center information corresponding to the target data center. If the device identifier and the geographical location are different from the data center information corresponding to the target data center, the data center information indicated by the parsing result does not match the data center information corresponding to the target data center.

[0124] The first determination submodule includes:

[0125] The generation unit is configured to generate data center information corresponding to each second user data based on the first field and the second field corresponding to the second user data, wherein the data center information includes the device identifier and the geographical location;

[0126] The determining unit is used to determine the N first user data from the M second user data based on the computer room information corresponding to each second user data.

[0127] Optionally, the acquisition module 510 includes:

[0128] The second acquisition submodule is used to traverse the N first user data and acquire the internet access information corresponding to each first user data, the internet access information including the user's internet access time period and the user's internet traffic.

[0129] The third acquisition submodule is used to acquire the target time period and the bandwidth capacity of the target data center. The target time period is the service opening time period of the target data center, and the bandwidth capacity is the maximum bandwidth that the target data center can provide.

[0130] The calculation submodule is used to calculate the bandwidth utilization rate of the target data center based on the N Internet access information, the target time period, and the data center bandwidth capacity.

[0131] The second determining submodule is used to determine the bandwidth utilization rate as the load prediction value of the target data center.

[0132] Optional, also includes:

[0133] The data acquisition module is used to acquire a training dataset, which includes historical load prediction values, historical environmental variables, historical running time parameters, historical switching parameters, and training labels corresponding to the target data center. The training labels are the historical cooling control strategies corresponding to the target data center. The historical load prediction values, historical environmental variables, historical running time parameters, and historical switching parameters are data within the same historical time period.

[0134] The training module is used to train the deep learning model based on the training dataset to obtain the target deep learning model.

[0135] Optionally, the refrigeration control strategy includes at least one of the following: refrigeration equipment supply / return air temperature regulation strategy, refrigeration equipment fan speed adjustment strategy, refrigeration equipment compressor speed adjustment strategy, refrigeration equipment on / off quantity control strategy, humidification / dehumidification control strategy, refrigerant switching strategy, fresh air switching strategy, dual cold source switching strategy, and compressor load setting strategy.

[0136] Optionally, the environmental variables include at least one of the following: Internet power, total power of the computer room, the ratio of energy consumed to energy consumed by the Internet load, supply air temperature of the refrigeration equipment, return air temperature of the refrigeration equipment, cold aisle temperature and humidity, hot aisle temperature and humidity, outdoor temperature and humidity, refrigeration equipment start / stop status, fan start / stop status, compressor start / stop status, fan speed, compressor speed, and outdoor temperature.

[0137] This application obtains N first user data matching the target computer room, calculates the load forecast value of the target computer room based on the N first user data, inputs the load forecast value, environmental variables, running time parameters and switching parameters into the target deep learning model for prediction, and obtains the cooling control strategy, thereby improving the control effect of the cooling system control strategy on the cooling system.

[0138] This application also provides an electronic device. Please refer to [link to relevant documentation]. Figure 6 The electronic device may include a processor 601, a memory 602, and a program 6021 stored in the memory 602 and executable on the processor 601.

[0139] When program 6021 is executed by processor 601, it can achieve the following: Figure 1 Any step in the corresponding method embodiment:

[0140] Obtain N first user data that match the target data center. The first user data is the internet access behavior data generated by the corresponding user based on the user account during broadband use. N is a positive integer.

[0141] The load forecast value of the target computer room is obtained by calculating the N first user data, and the load forecast value is the cooling load required in the target computer room;

[0142] The load forecast, the environmental variables corresponding to the target data center, the operating time parameters corresponding to the target data center, and the switching parameters corresponding to the target data center are input into the target deep learning model for prediction to obtain a cooling control strategy. The environmental variables include the temperature and humidity parameters corresponding to the target data center. The cooling control strategy is used to adjust the operating parameters corresponding to the cooling system of the target data center. The operating time parameter is the service runtime of the target data center, and the switching parameters are the number of times the cooling system of the target data center is switched on and off within the service runtime.

[0143] The cooling system of the target computer room is controlled based on the cooling control strategy.

[0144] Optionally, obtaining N first user data points matching the target data center includes:

[0145] Obtain M second user data, each of which corresponds to one of M user accounts. The second user data is the internet behavior data generated by the corresponding user during broadband use based on the user account. M is greater than or equal to N, and M is a positive integer.

[0146] The M second user data are parsed to obtain M parsing results. The M parsing results correspond one-to-one with the M second user data. The parsing results are used to indicate whether the data center information corresponding to the second user data matches the data center information corresponding to the target data center.

[0147] Based on the M parsing results, N first user data are determined from the M second user data, and the data center information indicated by the parsing results corresponding to the first user data matches the data center information corresponding to the target data center.

[0148] Optionally, parsing the M pieces of second user data to obtain M parsing results includes:

[0149] The M sets of second user data are cleaned and parsed to obtain the M parsing results. Each of the M parsing results includes a first field and a second field. The first field indicates the user access line identifier information of the corresponding second user data, and the second field indicates the user access location information of the corresponding second user data. The user access line identifier information indicates the device identifier of the user accessing the network, and the user access location information indicates the geographical location of the user accessing the network. If the device identifier and the geographical location are the same as the data center information corresponding to the target data center, the data center information indicated by the parsing result matches the data center information corresponding to the target data center. If the device identifier and the geographical location are different from the data center information corresponding to the target data center, the data center information indicated by the parsing result does not match the data center information corresponding to the target data center.

[0150] The step of determining the N first user data from the M second user data based on the M parsing results includes:

[0151] For each piece of second user data, data center information corresponding to each piece of second user data is generated based on the first field and the second field corresponding to the second user data. The data center information includes the device identifier and the geographical location.

[0152] Based on the data center information corresponding to each of the second user data, the N first user data are determined from the M second user data.

[0153] Optionally, the step of calculating the load forecast value of the target data room from the N first user data includes:

[0154] Iterate through the N first user data sets to obtain the internet access information corresponding to each first user data set, including the user's internet access time period and the user's internet traffic.

[0155] Obtain the target time period and data center bandwidth capacity corresponding to the target data center, wherein the target time period is the service opening time period of the target data center, and the data center bandwidth capacity is the maximum bandwidth that the target data center can provide;

[0156] The bandwidth utilization rate of the target data center is obtained by calculating the bandwidth utilization rate of the target data center based on the N sets of internet access information, the target time period, and the data center bandwidth capacity.

[0157] The bandwidth utilization rate is determined as the predicted load value for the target data center.

[0158] Optionally, before inputting the load forecast value, the environmental variables corresponding to the target computer room, and other variables corresponding to the target computer room into the target deep learning model for prediction to obtain the cooling control strategy, the method further includes:

[0159] Obtain a training dataset, which includes historical load prediction values, historical environmental variables, historical running time parameters, historical switching parameters, and training labels corresponding to the target data center. The training labels are the historical cooling control strategies corresponding to the target data center. The historical load prediction values, historical environmental variables, historical running time parameters, and historical switching parameters are data within the same historical time period.

[0160] The deep learning model is trained based on the training dataset to obtain the target deep learning model.

[0161] Optionally, the refrigeration control strategy includes at least one of the following: refrigeration equipment supply / return air temperature regulation strategy, refrigeration equipment fan speed adjustment strategy, refrigeration equipment compressor speed adjustment strategy, refrigeration equipment on / off quantity control strategy, humidification / dehumidification control strategy, refrigerant switching strategy, fresh air switching strategy, dual cold source switching strategy, and compressor load setting strategy.

[0162] Optionally, the environmental variables include at least one of the following: Internet power, total power of the computer room, the ratio of energy consumed to energy consumed by the Internet load, supply air temperature of the refrigeration equipment, return air temperature of the refrigeration equipment, cold aisle temperature and humidity, hot aisle temperature and humidity, outdoor temperature and humidity, refrigeration equipment start / stop status, fan start / stop status, compressor start / stop status, fan speed, compressor speed, and outdoor temperature.

[0163] This application obtains N first user data matching the target computer room, calculates the load forecast value of the target computer room based on the N first user data, inputs the load forecast value, environmental variables, running time parameters and switching parameters into the target deep learning model for prediction, and obtains the cooling control strategy, thereby improving the control effect of the cooling system control strategy on the cooling system.

[0164] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, this computer program implements the various processes of the control method embodiment for the computer room cooling system described above, and achieves the same technical effect. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0165] This application also provides a computer program product, which is stored in a storage medium and executed by at least one processor to implement the various processes of the above-described control method embodiment for the computer room cooling system, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0166] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0167] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0168] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A control method for a computer room cooling system, characterized in that, The method includes: Obtain N first user data that match the target data center. The first user data is the internet access behavior data generated by the corresponding user based on the user account during broadband use. N is a positive integer. The load forecast value of the target computer room is obtained by calculating the N first user data, and the load forecast value is the cooling load required in the target computer room; The load forecast, the environmental variables corresponding to the target data center, the operating time parameters corresponding to the target data center, and the switching parameters corresponding to the target data center are input into the target deep learning model for prediction to obtain a cooling control strategy. The environmental variables include the temperature and humidity parameters corresponding to the target data center. The cooling control strategy is used to adjust the operating parameters corresponding to the cooling system of the target data center. The operating time parameter is the service runtime of the target data center, and the switching parameters are the number of times the cooling system of the target data center is switched on and off within the service runtime. The cooling system of the target computer room is controlled based on the cooling control strategy.

2. The method according to claim 1, characterized in that, The acquisition of N first user data points matching the target data center includes: Obtain M second user data, each of which corresponds to one of M user accounts. The second user data is the internet behavior data generated by the corresponding user during broadband use based on the user account. M is greater than or equal to N, and M is a positive integer. The M second user data are parsed to obtain M parsing results. The M parsing results correspond one-to-one with the M second user data. The parsing results are used to indicate whether the data center information corresponding to the second user data matches the data center information corresponding to the target data center. Based on the M parsing results, N first user data are determined from the M second user data, and the data center information indicated by the parsing results corresponding to the first user data matches the data center information corresponding to the target data center.

3. The method according to claim 2, characterized in that, The process of parsing the M pieces of second user data to obtain M parsing results includes: The M sets of second user data are cleaned and parsed to obtain the M parsing results. Each of the M parsing results includes a first field and a second field. The first field indicates the user access line identifier information of the corresponding second user data, and the second field indicates the user access location information of the corresponding second user data. The user access line identifier information indicates the device identifier of the user accessing the network, and the user access location information indicates the geographical location of the user accessing the network. If the device identifier and the geographical location are the same as the data center information corresponding to the target data center, the data center information indicated by the parsing result matches the data center information corresponding to the target data center. If the device identifier and the geographical location are different from the data center information corresponding to the target data center, the data center information indicated by the parsing result does not match the data center information corresponding to the target data center. The step of determining the N first user data from the M second user data based on the M parsing results includes: For each piece of second user data, data center information corresponding to each piece of second user data is generated based on the first field and the second field corresponding to the second user data. The data center information includes the device identifier and the geographical location. Based on the data center information corresponding to each of the second user data, the N first user data are determined from the M second user data.

4. The method according to claim 1, characterized in that, The calculation of the load forecast value of the target data room based on the N first user data includes: Iterate through the N first user data sets to obtain the internet access information corresponding to each first user data set, including the user's internet access time period and the user's internet traffic. Obtain the target time period and data center bandwidth capacity corresponding to the target data center, wherein the target time period is the service opening time period of the target data center, and the data center bandwidth capacity is the maximum bandwidth that the target data center can provide; The bandwidth utilization rate of the target data center is obtained by calculating the bandwidth utilization rate of the target data center based on the N sets of internet access information, the target time period, and the data center bandwidth capacity. The bandwidth utilization rate is determined as the predicted load value for the target data center.

5. The method according to claim 1, characterized in that, Before inputting the load forecast value, the environmental variables corresponding to the target computer room, and other variables corresponding to the target computer room into the target deep learning model for prediction to obtain the cooling control strategy, the method further includes: Obtain a training dataset, which includes historical load prediction values, historical environmental variables, historical running time parameters, historical switching parameters, and training labels corresponding to the target data center. The training labels are the historical cooling control strategies corresponding to the target data center. The historical load prediction values, historical environmental variables, historical running time parameters, and historical switching parameters are data within the same historical time period. The deep learning model is trained based on the training dataset to obtain the target deep learning model.

6. The method according to any one of claims 1-5, characterized in that, The refrigeration control strategy includes at least one of the following: supply / return air temperature regulation strategy for refrigeration equipment, fan speed adjustment strategy for refrigeration equipment, compressor speed adjustment strategy for refrigeration equipment, on / off quantity control strategy for refrigeration equipment, humidification / dehumidification control strategy, refrigerant switching strategy, fresh air switching strategy, dual cold source switching strategy, and compressor load setting strategy.

7. The method according to any one of claims 1-5, characterized in that, The environmental variables include at least one of the following: Internet power, total power of the computer room, the ratio of energy consumed to energy consumed by the Internet load, supply air temperature of the refrigeration equipment, return air temperature of the refrigeration equipment, cold aisle temperature and humidity, hot aisle temperature and humidity, outdoor temperature and humidity, refrigeration equipment start / stop status, fan start / stop status, compressor start / stop status, fan speed, compressor speed, and outdoor temperature.

8. A control device for a computer room cooling system, characterized in that, The device includes: The acquisition module is used to acquire N first user data that match the target computer room. The first user data is the internet access behavior data generated by the corresponding user during the broadband usage process based on the user account, and N is a positive integer. The calculation module is used to calculate the load forecast value of the target computer room based on the N first user data, and the load forecast value is the cooling load required in the target computer room; The prediction module is used to input the predicted load value, the environmental variables corresponding to the target data center, the runtime parameters corresponding to the target data center, and the switching parameters corresponding to the target data center into the target deep learning model for prediction, and obtain a cooling control strategy. The environmental variables include the temperature and humidity parameters corresponding to the target data center. The cooling control strategy is used to adjust the runtime parameters corresponding to the cooling system of the target data center. The runtime parameters are the service runtime of the target data center, and the switching parameters are the number of times the cooling system of the target data center is switched on and off within the service runtime. The control module is used to control the cooling system of the target computer room based on the cooling control strategy.

9. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1 to 7.

11. A computer program product, characterized in that, Includes computer instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1 to 7.

Images