A machine room energy-saving regulation and control program evaluation method, system, device and medium

Abstract

The application provides a machine room energy-saving regulation and control program evaluation method, system, device and medium, the method comprises the following steps: obtaining multiple simulation data of the machine room environment, the simulation data comprises first simulation data defined under a preset scene and second simulation data generated based on a pre-constructed environment simulation model; based on the energy-saving regulation and control program to be tested, multiple simulation data are processed respectively, and the regulation and control strategy corresponding to each simulation data is output; wherein the energy-saving regulation and control program is obtained by encapsulating a pre-constructed energy-saving regulation and control model; the expected regulation and control result corresponding to each simulation data is obtained, each regulation and control strategy is compared with the expected regulation and control result corresponding thereto, and finally, the evaluation result of the energy-saving regulation and control program is determined. The application can improve the evaluation accuracy of the energy-saving regulation and control program and find potential problems in the energy-saving regulation and control program in time.

Description

Technical Field

[0001] This invention relates to the field of big data AI energy-saving technology, and in particular to a method, system, equipment and medium for evaluating data center energy-saving control programs. Background Technology

[0002] In recent years, with increasing attention paid to energy conservation and emission reduction in data centers, and the development of big data and artificial intelligence technologies, more and more data centers are beginning to use AI-based energy-saving control programs to regulate their air conditioning equipment. These programs can automatically adjust the parameters of data center equipment based on real-time conditions (such as ambient temperature and humidity) to achieve energy conservation and emission reduction goals. However, current energy-saving control programs lack rigorous testing and evaluation, leading to the inability to identify potential problems during the development phase. This results in frequent failures when the data center energy-saving control programs are deployed in real-world environments. Furthermore, data center environmental conditions are diverse and complex; simple simulation tests under only a few conditions may still encounter environmental conditions that cause program interruptions and errors after the energy-saving control programs are deployed. This fails to improve the accuracy of energy-saving control, and cannot ensure the safety of the data center environment or the achievable energy-saving effects. Summary of the Invention

[0003] This invention provides a method, system, equipment, and medium for evaluating data center energy-saving control programs to address existing problems in related technologies. The technical solution is as follows:

[0004] In a first aspect, embodiments of the present invention provide a method for evaluating a data center energy-saving control program, including:

[0005] Acquire multiple simulation data of the data center environment. The simulation data includes first simulation data defined under a preset scenario and second simulation data generated based on a pre-built environment simulation model.

[0006] The energy-saving control program under test processes multiple simulation data and outputs the control strategy corresponding to each simulation data; the energy-saving control program is encapsulated from a pre-built energy-saving control model.

[0007] Obtain the expected control result corresponding to each simulation data, compare each control strategy with its corresponding expected control result, and finally determine the evaluation result of the energy-saving control program.

[0008] In one implementation, the desired control outcome includes:

[0009] The first expected control indicator is defined when the target data center encounters a preset scenario with a specified problem. This first expected control indicator includes the expected control operation and the expected response time; and...

[0010] The second expected indicator for the target data center in the intelligent scenario includes the number of alarms, alarm duration, fault severity, equipment control frequency, and energy saving effect.

[0011] In one implementation, each control strategy is compared with its corresponding desired control result, and the final evaluation result of the energy-saving control program under test is determined by:

[0012] The first control strategy is compared with the first expected control indicator, and the first evaluation result is output; the first control strategy is the result output based on the analysis of the first simulation data using the energy-saving control model; and...

[0013] The second control strategy is statistically analyzed to obtain the indicator results; the indicator results are compared with the second expected indicator to output the second evaluation result; wherein, the second control strategy is the result output by analyzing the second simulation data based on the energy-saving control model.

[0014] In one implementation, it further includes:

[0015] Acquire signal data from a specified data source, integrate the signal data, and obtain a first dataset containing device signals at all time steps;

[0016] Based on the first dataset, all device signals from the previous time step are used as input to the memory network model, and the device signals from the next time step are used as output to train the memory network model, thus obtaining the environment simulation model.

[0017] In one implementation, it further includes:

[0018] Obtain signal data from a specified data source, integrate the signal data, and obtain a second dataset;

[0019] Using the second dataset, the device signals and environmental conditions are used as inputs to the learning model, and the optimized control strategy is used as the output of the learning model. The learning model is trained and optimized through a preset reward function to obtain an energy-saving control model.

[0020] In one implementation, it further includes:

[0021] Based on the second dataset, multiple data subsets within different consecutive time periods are determined. These multiple data subsets are then processed to obtain initial state data used to define the initial state of the computer room environment.

[0022] The environmental simulation model is initialized based on the initial state data, so that the environmental simulation model can simulate the dynamic changes of the computer room environment from the initial state.

[0023] In one implementation, it further includes:

[0024] The simulated data is stored in a designated storage unit, and the stored simulated data is backed up periodically through the designated storage unit to obtain backup data, which is then stored in the designated storage unit.

[0025] Secondly, embodiments of the present invention provide a data center energy-saving control program evaluation system, which executes the data center energy-saving control program evaluation method described above.

[0026] Thirdly, embodiments of the present invention provide an electronic device comprising a memory and a processor. The memory and the processor communicate with each other via an internal connection path. The memory stores instructions, and the processor executes the instructions stored in the memory. When the processor executes the instructions stored in the memory, it causes the processor to perform the method described in any of the above embodiments.

[0027] Fourthly, embodiments of the present invention provide a computer-readable storage medium that stores a computer program, wherein when the computer program is run on a computer, the methods in any of the embodiments described above are executed.

[0028] The advantages or beneficial effects of the above technical solutions include at least the following:

[0029] This invention predefines corresponding first simulated data for preset scenarios where environmental or data problems may occur, and simultaneously collects second simulated data generated based on a pre-constructed environmental simulation model. An energy-saving control model is then used to generate corresponding control strategies for both the first and second simulated data, and these strategies are evaluated. This improves the accuracy of the energy-saving control program's evaluation, promptly identifies potential problems within the program, and facilitates subsequent optimization to further enhance its energy-saving control effect.

[0030] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the invention will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0031] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments disclosed in the invention and should not be construed as limiting the scope of the invention.

[0032] Figure 1 This is a flowchart illustrating the evaluation method for energy-saving control procedures in computer rooms according to the present invention.

[0033] Figure 2 This is a schematic diagram illustrating the execution steps of the data center energy-saving control program evaluation method of the present invention;

[0034] Figure 3 This is a schematic diagram illustrating the source of an example of the environment simulation program of this invention;

[0035] Figure 4 This is a schematic diagram of the unit connections of the data center energy-saving control program evaluation system of the present invention;

[0036] Figure 5 This is a schematic diagram illustrating the relationship between the environmental simulation program example, the energy-saving control program example, and the evaluation example of the present invention.

[0037] Figure 6 This is a schematic diagram of the interaction and logical time synchronization process of various program instances of the present invention;

[0038] Figure 7 This is a structural block diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0039] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0040] In recent years, with increased national attention to energy conservation and emission reduction in data centers, and the development of big data and artificial intelligence technologies, more and more data centers have begun to use AI-based energy-saving control programs to regulate air conditioning equipment. These programs can generate control strategies for air conditioning equipment based on real-time data center conditions (such as ambient temperature and humidity) and distribute them via gateway interfaces for remote control of air conditioning and other data center equipment. These programs need to maintain a safe data center environment, minimize unnecessary cooling redundancy, and achieve energy conservation and emission reduction goals. To ensure the environmental safety of the data center and the achievable energy-saving effects, rigorous testing and evaluation must be conducted before the model is deployed online to identify and address potential problems. Only after passing the evaluation can the data center energy-saving control program be deployed to the production environment.

[0041] Currently, the evaluation methods used in the industry for energy-saving control programs mainly include code unit testing, integration testing, and manual evaluation. While these methods can, to some extent, prevent problems such as program interruptions due to logical errors after the algorithm control program is deployed, they also have the following shortcomings:

[0042] (1) Limited evaluation conditions. Unit testing and integration testing can only be performed under a limited number of preset conditions. However, the environmental conditions in the computer room are diverse and complex. Even after the energy-saving control program is launched, it may still encounter environmental conditions that cause the program to interrupt or report errors.

[0043] (2) It is difficult to assess the safety and energy-saving effect of the control. Unit testing and integration testing can only ensure that the program does not interrupt or report errors under preset conditions, but cannot assess whether the control strategy generated by the program is reasonable. An unreasonable strategy may lead to excessively high or low temperature in the computer room environment, which will affect the safety and energy-saving effect of the computer room environment.

[0044] (3) There are potential safety hazards. When energy-saving control programs are run and evaluated in a real data center environment, the data center environment will be affected. If the data center is carrying important business, the impact of problems will be significant.

[0045] (4) Long evaluation cycle. Due to factors such as signal acquisition frequency, equipment control interval, and environmental lag, a manual evaluation in a real environment often requires waiting for several hours or even tens of hours, which seriously affects the efficiency of R&D iteration.

[0046] (5) It is difficult to achieve automated batch assessment. Manual assessment methods are limited by human resources and cannot achieve automation and batch assessment. When the energy-saving control program is upgraded and iterated frequently and the environmental conditions of the computer room to be assessed are diverse, the required labor cost is enormous.

[0047] When energy-saving control programs are frequently upgraded and iterated, and a large number of computer rooms are involved, the adverse effects of the above-mentioned problems will be particularly obvious.

[0048] To address the aforementioned problems, in a first aspect, embodiments of the present invention provide a method for evaluating data center energy-saving control programs, such as... Figure 1 , Figure 2 As shown, the specific method includes the following steps:

[0049] Step S1: Obtain multiple simulation data of the data center environment. The simulation data includes first simulation data defined under a preset scenario and second simulation data generated based on a pre-built environment simulation model.

[0050] Before evaluating the energy-saving control effect, various parameters need to be configured to specify the specific execution methods of various key operations in each execution unit. The configuration includes the specific parameters and corresponding values ​​of each execution unit, which includes signal acquisition units, data storage units, environmental simulation units, interface synchronization units, control strategy units, and result evaluation units. These configurations will be stored in the data storage unit for other units to read and use.

[0051] After parameter configuration is completed, signal data is collected from designated data sources such as the computer room environmental system via the signal acquisition unit and organized into relevant datasets. The resulting datasets will be stored in the data storage unit and can be read by other execution units such as the environmental simulation unit, control strategy unit, and result evaluation unit for training models and evaluating energy-saving control programs.

[0052] The datasets mainly include data center information datasets, equipment information datasets, signal information datasets, and historical signal datasets. Among them, the data center information, equipment information, and signal information datasets provide the identifiers, characteristics, and lists of related objects, and can be regarded as static data; while the historical signal dataset provides the values ​​of data center equipment signals at multiple points in time, and can be regarded as dynamic data.

[0053] In this embodiment, the collected signal data is integrated through feature engineering to obtain the first dataset required for training the environmental simulation model of the target computer room and the second dataset required for training the energy-saving control model of the target computer room, which are used to train the computer room environmental simulation model and the computer room energy-saving control model, respectively.

[0054] Specifically, the first dataset used to train the environmental simulation model includes static data of the computer room equipment and historical signal data (ambient temperature and humidity, air conditioning control parameters, electricity meter energy consumption, etc.). A Long Short-Term Memory (LSTM) network model architecture is used, where the model input for each time step is all the equipment signals at that time step, and the model output is the equipment signals for the next time step. The model is trained using historical signal data, allowing it to learn the dependencies between equipment signals and their evolution over time. After training, the resulting environmental simulation model can predict the values ​​of various equipment signals, such as temperature, humidity, and energy consumption, at every time interval (1 minute), forming a multivariate time series prediction task.

[0055] The second dataset used to train the energy-saving control model includes static data of the computer room equipment and historical signal data (ambient temperature and humidity, air conditioning control parameters, electricity meter energy consumption, etc.). A policy gradient method using a deep reinforcement learning architecture processes the input equipment signals and environmental conditions, outputting an optimized control policy. The control policy refers to the air conditioning control signals (such as temperature and humidity setpoints, on / off states, etc.). The model uses all current equipment states and environmental conditions as the state space and all controllable actions as the action space. A reward function related to energy-saving effects optimizes the model, rewarding low-energy-consumption and stable environmental parameter policies and penalizing unreasonable control behaviors, ultimately yielding the energy-saving control model. The energy-saving control model continuously interacts with the simulated environment, selecting actions and receiving feedback at each time step, gradually optimizing the control policy to maximize cumulative rewards.

[0056] For the energy-saving control model, the expression of the reward function used in this embodiment is as follows:

[0057] Reward=-(α×EnergyConsumption+b×EnvironmentDeviation);

[0058] Wherein, EnergyConsumption: represents the air conditioning energy consumption value at the current time step, with the goal of minimizing energy consumption; in this embodiment, the time step size used is 1 minute, so this variable represents the energy consumption value within 1 minute;

[0059] EnvironmentDeviation: Represents the deviation value of environmental parameters, measuring the deviation between parameters such as temperature and humidity and the target value. The goal is to minimize the deviation to maintain environmental stability. In this embodiment, 28 degrees Celsius is taken as the target value of the ambient temperature. If the actual temperature is 30 degrees Celsius, then the deviation is 2 degrees Celsius.

[0060] a, b: represent weighting coefficients used to balance the importance of energy saving effect and environmental stability; in this embodiment, a = 1 and b = 0.5.

[0061] After training the environmental simulation model and energy-saving control model, the environmental simulation model is packaged into an environmental simulation program and integrated into the environmental simulation unit, and the energy-saving control model is packaged into an energy-saving control program and integrated into the control strategy unit, in preparation for deployment in subsequent steps.

[0062] The environment simulation unit provides the control strategy unit with static data about the computer room, equipment, and signals, as well as dynamic data of the simulated environmental signals, through the interface synchronization unit. At the same time, it can also receive control signals from the control strategy unit and influence changes in the simulated environment accordingly.

[0063] The static data provided by the environmental simulation unit can be pre-set by the user, describing the basic characteristics of a specific computer room and its auxiliary equipment. Dynamic data, on the other hand, can be either pre-set by the user or generated through an environmental simulation model, depending on the chosen simulation method, describing the state changes of a given computer room and equipment. In addition to simulating data such as the computer room environment, the environmental simulation unit also supports simulating data-related issues, such as outliers, missing data, and data latency. By simulating one or more virtual computer rooms and problem conditions, it facilitates a comprehensive evaluation of subsequent energy-saving control procedures.

[0064] It should be noted that the present invention does not limit the specific algorithms and deep learning frameworks used within the environmental simulation model and energy-saving control model; in other embodiments, different algorithms and frameworks may be used for training.

[0065] In this embodiment, the environment simulation unit simulates the computer room environment in two ways: by using simulation data from two different scenarios as input to the energy-saving control program, it derives control strategies for each scenario. Specifically, the first type of simulation data is the first simulation data corresponding to the preset scenario, and the second type of simulation data is the second simulation data corresponding to the intelligent scenario.

[0066] Pre-defined scenario simulation involves envisioning typical problem scenarios (such as sudden increases in localized temperature in a data center, equipment failure, and data issues) based on business experience, and then manually constructing corresponding simulated data for evaluation. This method is ready to use as long as the simulated data is configured, and is suitable for evaluating the algorithm's response to known problems.

[0067] Intelligent scenario simulation primarily generates simulated data center environments through environmental simulation models, enabling the creation of more complex or unpredictable scenarios for evaluation. This approach requires deploying environmental simulation programs within the environmental simulation unit and is suitable for evaluating algorithm performance under unpredictable conditions.

[0068] Furthermore, before running the environment simulation program, it is necessary to select multiple data subsets from the second dataset within different consecutive time periods according to actual needs. These data subsets are then cleaned, formatted, or transformed to obtain initial state data that meets the input requirements of the environment simulation program. This initial state data defines the initial state of the data center environment, including setting key parameters such as initial temperature, humidity, and equipment status. The defined initial state data is then loaded into the environment simulation program, and the program is started. At this point, the environment simulation program will begin simulating the dynamic changes of the data center environment based on the initial state. Through these steps, the data center environment simulation program can be initialized, ensuring that the simulation environment starts from a data subset representing the actual operating state of the data center, thereby enabling effective simulation and evaluation.

[0069] It should be noted that, in addition to generating data center environment simulation data through an environmental simulation model, other models using machine learning or deep learning algorithms can also be used to generate data center environment simulation data. This invention does not restrict the specific machine learning or deep learning algorithms used in its models, but they must meet the following requirements to achieve interface synchronization and control strategy unit integration:

[0070] • It can be initialized using different data center status datasets, which mainly include signal values ​​of various devices in the data center;

[0071] • It can continuously predict and output various dynamic data composed of data center environmental signals to meet the needs of energy-saving control programs;

[0072] • It can receive control strategies from energy-saving control programs and change the direction of changes in the simulated environment accordingly;

[0073] Internal operations are implemented using logical time instead of physical time, and the logical time provided by the external module is used as the standard.

[0074] like Figure 3 As shown, multiple environmental simulation program instances can exist simultaneously for each environmental simulation method. An environmental simulation program instance refers to the process of a specific environmental simulation program initiated during the evaluation process. For preset scenario simulations, each environmental simulation program instance comes from a preset scenario dataset; for intelligent simulations, each environmental simulation program instance comes from a combination of a data center initial state dataset and an environmental simulation model.

[0075] Step S2: Based on the energy-saving control program to be tested, process multiple simulation data respectively and output the control strategy corresponding to each simulation data; wherein, the energy-saving control program is encapsulated from a pre-built energy-saving control model.

[0076] An energy-saving control model refers to a model that can generate energy-saving control strategies based on relevant input features. The energy-saving control model is simply encapsulated into an energy-saving control program integrated into the control strategy unit. This program, through the energy-saving control unit and interface unit, collects the necessary signal data from the environmental simulation unit and outputs the generated control strategy to the environmental simulation unit via the interface unit.

[0077] The energy-saving control model can be trained using data collected by the signal acquisition unit. It should be noted that this invention does not restrict the specific machine learning or deep learning algorithms used in the model, but it must meet the following requirements to achieve interface synchronization with the interface synchronization unit and the environment simulation unit:

[0078] • It can generate control strategies based on simulation data provided by the environmental simulation unit, and the control strategies can be received by the environmental simulation unit.

[0079] Internal operations are implemented using logical time instead of physical time, and the logical time provided by the external module is used as the standard.

[0080] During the assessment, multiple energy-saving control program instances may exist simultaneously. An energy-saving control program instance refers to the process of a specific energy-saving control program initiated during the assessment. Each energy-saving control program instance has one and only one corresponding environmental simulation program instance, and the two together constitute an assessment instance (e.g., ...). Figure 5 (As shown).

[0081] To accelerate the control process while ensuring consistency in the order and intervals of related events between the accelerated and environmental simulation units, the control strategy unit needs to obtain the logical time as a reference time for its internal program using the interface synchronization unit, thus achieving serial acceleration. Simultaneously, the control strategy unit can run multiple environmental simulation program instances concurrently, achieving parallel acceleration. These two acceleration methods reduce the time required for evaluation.

[0082] Step S3: Obtain the expected control result corresponding to each simulation data, compare each control strategy with its corresponding expected control result, and finally determine the evaluation result of the energy-saving control program.

[0083] Since this embodiment has two modes, preset scenario simulation and intelligent scenario simulation, the control strategies generated based on the simulation data under different scenarios are different, and correspondingly, the methods for evaluating the generated control strategies may also be different.

[0084] Specifically, for the simulation of a preset scenario, when defining the first simulation data corresponding to the preset scenario, since this simulation data is artificially constructed and is simulation data under a known problem, the expected control result under the preset scenario is also predictable. Similarly, the first expected control index corresponding to the preset scenario can be manually defined. The first expected control index includes the expected control operation and the expected response time, serving as the evaluation basis for the first control strategy generated under the preset scenario. That is, based on the energy-saving control model, the first simulation data is analyzed to output the first control strategy, and the first control strategy is compared with the first expected control index to output the first evaluation result.

[0085] For intelligent scenario simulation, a second expected indicator is preset for the target data center under the intelligent scenario. The second expected indicator includes alarm count, alarm duration, fault severity, equipment control frequency, and energy-saving effect. The control strategy generated under the intelligent scenario is evaluated based on the second expected indicator. That is, the second simulation data is analyzed based on the energy-saving control model to output the second control strategy, the second control strategy is statistically analyzed to obtain the indicator results, and the indicator results are compared with the second expected indicator to output the second evaluation result.

[0086] It should be noted that the method for evaluating the performance of the energy-saving control program in a given computer room simulation environment can be carried out in the evaluation unit. The results evaluation unit uses statistical analysis methods to evaluate the energy-saving control program from the following aspects:

[0087] • Robustness: This mainly refers to the ability of the energy-saving control program to maintain normal operation and smoothly execute its functions. For example, it includes situations where program warnings, errors, or interruptions occur during operation. Because the evaluation involves multiple simulation environments, the program will face more uncertain inputs; therefore, this type of evaluation can serve as an effective supplement to general unit testing and integration testing.

[0088] • Security: This mainly refers to the ability of the energy-saving control program to promptly identify and handle environmental anomalies and maintain various data center environmental indicators within safe ranges under various data center environmental conditions. For example, during the operation of the energy-saving control program, this includes the types of alarms (ambient temperature, humidity, etc.) that occur in the simulated data center environment, as well as their frequency, duration, and severity; and whether the energy-saving control program can make the expected response in specific scenarios, and the timeliness of that response. This type of assessment plays a crucial role in ensuring the security of the data center environment after the energy-saving control program goes live and cannot be replaced by general unit testing and integration testing.

[0089] • Energy efficiency: This mainly refers to the ability of energy-saving control programs to reduce the energy consumption of the computer room's air conditioning. For example, the amount of electricity saved and the energy saving rate compared to not using the program. This type of assessment is valuable in ensuring that the energy-saving control programs deliver the expected benefits after implementation.

[0090] It should be noted that the evaluation indicators that can be statistically analyzed are related to the simulation method in the environmental simulation unit. For example, indicators such as power saving and ambient temperature alarm duration, which involve long-term interaction between energy-saving control programs and the computer room simulation environment, can only be evaluated through intelligent scenario simulation. On the other hand, indicators such as expected operations and their response times, which require pre-specified conditions, need to be evaluated through preset scenario simulation. Indicators for robustness assessment (such as program error rate) are not limited by the simulation method.

[0091] After each assessment, the results will be tagged with information such as the energy-saving control program version and the computer room simulation environment, and stored in the data storage unit. Users can query the assessment results and make horizontal (same energy-saving control program version, different simulation environments) and vertical (same simulation environment, different energy-saving control program versions) comparisons.

[0092] Secondly, embodiments of the present invention provide a data center energy-saving control program evaluation system, such as... Figure 4 As shown, this system includes a parameter configuration unit, a signal acquisition unit, a data storage unit, an environmental simulation unit, a control strategy unit, an interface synchronization unit, and a result evaluation unit.

[0093] The parameter configuration unit is used to configure various parameters to specify the specific execution methods of various key operations in the device. The configuration includes multiple parameters and their corresponding values, involving the signal acquisition unit, data storage unit, environmental simulation unit, interface synchronization unit, control strategy unit, and result evaluation unit. These configurations will be stored in the data storage unit for other units to read and use.

[0094] The signal acquisition unit is used to collect signal data from data sources such as the computer room's environmental system and organize it into relevant datasets. The datasets obtained by the signal acquisition unit will be stored in the data storage unit and can be read by the environmental simulation unit, control strategy unit, and result evaluation unit for training models and evaluating energy-saving control programs.

[0095] The data storage unit is used to save and back up relevant data configured, collected, or generated by this device for use by other units. This device's data storage unit interacts with other units within the device that have data read / write requirements via an interface.

[0096] The data storage unit needs to store configuration parameters, signal data, evaluation data, log data, etc. In addition to data storage, this unit also has data backup capabilities, allowing for full backups, incremental backups, and other backup methods to prevent data loss due to system failures.

[0097] The environmental simulation unit is used to deploy environmental simulation and output computer room programs, and to collect environmental simulation data as input for energy-saving control programs. The environmental simulation unit provides the control strategy unit with static data about the computer room, equipment, and signals, as well as dynamic data of simulated environmental signals, through the interface synchronization unit; simultaneously, it can also receive control signals from the control strategy unit and influence changes in the simulated environment accordingly.

[0098] The control strategy unit is used to deploy the energy-saving control program to be tested and output the control strategy. The energy-saving control model is a model that can generate energy-saving control strategies based on relevant input features; the energy-saving control program is a simple encapsulation of the energy-saving control model, providing basic functions such as input / output processing and interface docking. The energy-saving control program, with the help of this unit and the interface docking unit, can collect the required signal data from the environmental simulation unit and output the generated control strategy to the environmental simulation unit via the interface docking unit.

[0099] Combination Figure 5As shown, the interface synchronization unit is used to connect and synchronize the computer room simulation environment with the energy-saving control program, and to capture relevant data as an evaluation basis. The interface synchronization unit establishes a one-to-one correspondence between multiple running instances in the environment simulation unit and the control strategy unit, forwards requests from different instances to the correct corresponding instances, and provides consistent logical timing for paired instances. At the same time, the interface synchronization unit can also save the captured data on the data storage unit for the result evaluation unit to read and use.

[0100] The functions provided by the interface synchronization unit include:

[0101] • Registration and association of assessment instances: When starting the assessment process, the interface synchronization unit needs to register the environment simulation program instance and the energy-saving control program instance, assign them unique identifiers, establish a one-to-one correspondence, and form an assessment instance.

[0102] • Define and synchronize logical time: For each evaluation instance, the interface synchronization unit needs to initialize and maintain a logical time so that the environmental simulation program instance and the energy-saving control program instance can run and complete the interaction according to the time. This function is mainly accomplished through the logical time management module and the time coordinator.

[0103] • Intercept and output evaluation data: While accepting and forwarding data from energy-saving control program instances and environmental simulation program instances, it outputs the data to the results evaluation unit to obtain evaluation results.

[0104] The logical time management module and the time coordinator are the main components for implementing the logical time acceleration described in this invention. The core responsibility of the logical time management module is to maintain and advance logical time; while the task of the time coordinator is to ensure data flow and interactive synchronization between programs at each logical time advancement step. This embodiment employs an event-driven logical time synchronization method, driving the advancement of logical time steps by simulating the interaction time between environmental program instances, energy-saving control program instances, and the time coordinator; each advancement advances the logical time by a preset step size.

[0105] In the above process, the speed of time step progression is limited only by the efficiency of the program's interactive operation and the preset time step size. Assuming that the average physical time consumed from the step of "requesting simulation environment data" to "forwarding control strategy" is 60 milliseconds, and the preset time step size is 60 seconds, it means that the time required to carry out the evaluation work can be reduced to one-thousandth of the original time.

[0106] The results evaluation unit is used to evaluate the performance of the energy-saving control program in a given computer room simulation environment. The data used in the evaluation includes simulated environment data output by the environmental simulation program instance and control strategy data output by the energy-saving control program instance. After statistical analysis of the data and obtaining the evaluation results, these results will be tagged and stored in the data storage unit for later use.

[0107] The following describes the energy-saving control system for the computer room in this embodiment, based on actual data:

[0108] The parameter configuration unit completes the configuration of various parameters involved in the operation of the device. For example, the configuration related to the signal acquisition unit is shown in Table 1:

[0109] Table 1: Example of parameter configuration for the implementation plan

[0110]

[0111]

[0112] The signal acquisition unit collects signal data from the environmental monitoring system and processes it to form a data center information dataset, a device information dataset, a signal information dataset, and a historical signal dataset, as shown in Tables 2, 3, 4, and 5 (anonymized). Since a data center typically contains multiple devices, and each device can have one or more signals, the device list field in the data center information dataset and the signal list field in the device information dataset store multiple corresponding device or signal identifiers in a list format. It should be noted that this embodiment uses only one data center as an example; therefore, the datasets only contain data related to a single data center. However, the system proposed in this embodiment supports simultaneous data collection from multiple data centers.

[0113] Table 2: Data Set of Computer Room Information

[0114]

[0115] Table 3: Equipment Information Dataset

[0116]

[0117]

[0118] Table 4: Signal Information Dataset

[0119]

[0120] Table 5: Historical Signal Dataset

[0121]

[0122]

[0123] This embodiment uses the relational database management system PostgreSQL to store data, and combines the data backup tool pgdump and the crontab scheduled task tool that comes with the Unix-like operating system to perform regular data backup.

[0124] It should be noted that the present invention is not limited to the relational database management system implementation used; in other embodiments, a relational database management system different from that in this embodiment may be used.

[0125] To implement the preset scenario simulation method, this embodiment constructs one or more simulated historical signal datasets based on business experience and the characteristics of the target data center, similar in format to Table 5. The simulated signal datasets need to reflect environmental or data problems that may occur during data center energy-saving control, such as abnormal temperature increases at a certain location in the data center or missing air conditioning operation status signals. After constructing the datasets, it is also necessary to define the expected control operations and response times of the data center energy-saving control program based on the problems reflected in the datasets, and complete the configuration and association with the preset scenarios through the parameter configuration unit. This embodiment constructs three preset scenarios, each lasting 24 hours. The problems reflected in each scenario and the corresponding expected operations and response times are described below (for simplicity, only the relevant equipment and signals closely related to the scenarios are described):

[0126] Preset scenario 1 is as follows:

[0127] • Evaluation objective: To evaluate whether the data center's energy-saving control program can promptly prevent the ambient temperature from rising continuously and avoid overheating.

[0128] • Scenario Summary: The temperature signal of the temperature sensing device "Computer Room Ambient Temperature -1" remained at 24.0 degrees Celsius from 00:00:00 to 23:00:00; starting from 23:00:00, the temperature of "Computer Room Ambient Temperature -1" increased at a rate of 0.1 degrees Celsius per minute, reaching 28.0 degrees Celsius at 23:40:00; the cooling device "Air Conditioner 1" closest to the "Computer Room Ambient Temperature -1" temperature sensor remained in the off state.

[0129] • Expected response: Turn on "Air Conditioner 1" and set the temperature setting of "Air Conditioner 1" to below 28.0 degrees Celsius.

[0130] • Expected response time: before 23:43:00.

[0131] Preset scenario 2 is as follows:

[0132] • Evaluation objective: To evaluate whether the data center energy-saving control program can promptly prevent the ambient temperature from continuously decreasing and reduce unnecessary energy consumption.

[0133] • Scenario Summary: The temperature signal of the temperature sensor "Computer Room Ambient Temperature -1" remained at 26.0 degrees Celsius from 00:00:00 to 23:00:00; starting from 23:00:00, it decreased at a rate of 0.1 degrees Celsius per minute, reaching 22.0 degrees Celsius at 23:40:00; the cooling device "Air Conditioner 1" closest to the "Computer Room Ambient Temperature -1" sensor was always on, with a temperature setpoint of 26.0 degrees Celsius; another cooling device "Air Conditioner 2" relatively close to the "Computer Room Ambient Temperature -1" sensor was also always on, with a temperature setpoint of 22.0 degrees Celsius.

[0134] • Expected response: Increase the temperature setting of "Air Conditioner 2" by 1-2 degrees Celsius.

[0135] • Expected response time: before 23:45:00.

[0136] Preset scenario 3 is as follows:

[0137] • Evaluation objective: To evaluate whether the data center energy-saving control program can handle signal loss issues in a timely and reasonable manner.

[0138] • Scenario summary: The refrigeration equipment "Air Conditioner 1" is always on from 00:00:00 to 23:00:00, and the return air temperature and its set value signal are always maintained at 25 degrees Celsius; the refrigeration equipment "Air Conditioner 2" which is close to "Air Conditioner 1" is always off; starting from 23:00:00, the return air temperature signal of "Air Conditioner 1" has been missing.

[0139] • Expected response operation: Turn on "Air Conditioner 2" to make the temperature setting of "Air Conditioner 2" close to the temperature setting of "Air Conditioner 1" (±1 degree Celsius); turn off "Air Conditioner 1".

[0140] • Expected response time: before 23:50:00.

[0141] To implement the intelligent scenario simulation method, in addition to training the data center environment simulation model and encapsulating it into a program according to the aforementioned steps, it is also necessary to specify the initial state of the target data center for the program to initialize the simulation program. Depending on the initial state, the simulation data output by the data center environment simulation program will also differ, which can increase the scope of the test scenarios to some extent. This initial state can be described by a subset of data within any consecutive time period in the historical signal dataset (Table 5). In this embodiment, data from three time periods were selected for the initialization of the data center environment simulation program. Furthermore, this embodiment also uses a parameter configuration unit to uniformly set the duration of the data center environment simulation to 24 hours, and also configures alarm thresholds for certain key signals (such as temperature) from equipment such as temperature sensors and air conditioners.

[0142] After completing the above configuration, the environmental simulation program instance, energy-saving control program instance, and evaluation instance can be started, and evaluation data collection can begin. After starting the environmental simulation program instance, the program begins generating various data center environmental signals (e.g., temperature sensor readings, air conditioning return air temperature, etc.). The energy-saving control program instance can request simulated signal data from the environmental simulation program instance through the interface synchronization unit, and based on this simulated signal data, generate appropriate control strategies (e.g., adjusting the temperature setpoint of an air conditioner, or switching the air conditioner on and off, etc.), and transmit the control strategy data to the environmental simulation unit through the interface synchronization unit, which in turn affects the environmental simulation program instance (see process details). Figure 6 (As shown). While forwarding environmental simulation signal data and control strategy data between the environmental simulation program instance and the energy-saving control program instance, the interface synchronization unit also forwards this data to the result evaluation unit for evaluation. Since this embodiment constructs three preset scenarios and three initial states of the computer room environment simulation, a total of six computer room environment simulation program instances are formed. These six environmental simulation program instances, together with six corresponding computer room energy-saving control program instances, will form evaluation instances.

[0143] In this embodiment, the duration of all six evaluation instances is set to 24 hours, with a time step of 1 minute. Both the data center environment simulation model and the data center energy-saving control model output predictions at a frequency of 1 minute per prediction (a total of 1440 predictions over 24 hours). This means that in a real environment, each evaluation would require 24 hours of physical time, resulting in a total physical time consumption of 144 hours. However, due to the adoption of a logical time-based acceleration method, the actual physical time consumed by each evaluation instance is significantly reduced, providing an acceleration of over 1000 times. Furthermore, since all instances are executed in parallel, the total physical time consumed is equal to the time consumed by the longest-running instance, rather than the sum of the times consumed by all six evaluation instances. The physical time consumed by running each of the aforementioned evaluation instances and the acceleration effect are shown in Table 6.

[0144] Table 6: Speed ​​Improvement Effect

[0145]

[0146] Furthermore, it should be noted that since the data center environment simulation model and the data center energy-saving control model in this embodiment use computationally intensive deep learning algorithms, they are the main determinants of runtime, and the runtime of the evaluation instance is approximately equal to the sum of the runtimes of the two. Because the preset scenario simulation method does not use the data center environment simulation model, the time consumed by environment simulation is approximately zero, thus the acceleration effect of the relevant evaluation instance is more significant. In other embodiments, the acceleration effect will vary depending on the complexity of the algorithm used and the hardware specifications employed.

[0147] The evaluation method and apparatus proposed in this embodiment support the evaluation of data center energy-saving control programs from major aspects such as robustness, security, and energy efficiency; however, depending on the environmental simulation method used in the evaluation examples, the evaluation may focus on different aspects:

[0148] • For the preset scenario simulation, this embodiment mainly evaluated whether the data center energy-saving control program can perform the expected control operation when encountering specific problem scenarios, and whether the response time is within the expected range. The main results are shown in Table 7.

[0149] For intelligent scenario simulation, this embodiment mainly evaluated the number, duration, and severity of data center environmental temperature alarms, equipment control frequency, energy-saving effect, and the number of program error interruptions during the evaluation period. The main evaluation results are shown in Table 8.

[0150] The results show that, in this embodiment, the data center energy-saving control program performed the expected operations within the expected timeframes in all three preset scenarios evaluated, thus passing the evaluation. However, in the three intelligent scenario simulations, program errors and excessively high data center temperatures were observed, indicating a problem with the internal logic of the energy-saving control program.

[0151] Table 7: Evaluation Results of Preset Scenario Simulation

[0152]

[0153]

[0154] Table 8: Intelligent Scenario Simulation Evaluation Results

[0155]

[0156]

[0157] Through the above steps, this embodiment has completed the evaluation of the data center energy-saving control program, and its beneficial effects are as follows:

[0158] 1. This embodiment proposes a relatively comprehensive evaluation system for data center energy-saving control programs, which can identify potential problems in the programs as much as possible and improve the safety and effectiveness of data center energy saving. It proposes to evaluate data center energy-saving control programs from three aspects: robustness, security, and energy efficiency. It also proposes to provide an operating environment for energy-saving control programs by combining preset scenario simulation and intelligent scenario simulation, so as to ensure that the energy-saving control programs can make the expected response in a timely manner under known conditions and still output reasonable control strategies in unknown environments.

[0159] 2. In this embodiment, the system can achieve large-scale, automated evaluation of data center energy-saving control programs through the cooperation of various units, which can significantly reduce the manual input of the research and development cycle.

[0160] 3. This embodiment uses logical time to replace physical time to realize the internal logic of the program, thereby accelerating the serial evaluation and reducing the time required for each evaluation. It also coordinates the interaction between the computer room environment simulation program and the computer room energy-saving control program by synchronizing the external standard logical time of the program, ensuring the correctness of the relevant results. Furthermore, it proposes to deploy environmental simulation program instances and energy-saving control program instances in a flexible and scalable manner to achieve parallel acceleration of the evaluation, further reducing the time required to conduct evaluations on multiple evaluation instances.

[0161] Thirdly, embodiments of the present invention provide an electronic device. Figure 7 A structural block diagram of an electronic device according to an embodiment of the present invention is shown. Figure 7As shown, the electronic device includes a memory 100 and a processor 200. The memory 100 stores a computer program that can run on the processor 200. When the processor 200 executes the computer program, it implements the data center energy-saving control program evaluation method in the above embodiment. The number of memories 100 and processors 200 can be one or more.

[0162] The electronic device also includes:

[0163] The communication interface 300 is used to communicate with external devices and perform data exchange and transmission.

[0164] If the memory 100, processor 200, and communication interface 300 are implemented independently, they can be interconnected via a bus to communicate with each other. This bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc.

[0165] Optionally, in a specific implementation, if the memory 100, processor 200, and communication interface 300 are integrated on a single chip, then the memory 100, processor 200, and communication interface 300 can communicate with each other through an internal interface.

[0166] This invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method provided in this invention.

[0167] This invention also provides a chip, which includes a processor for calling and executing instructions stored in a memory, causing a communication device on which the chip is installed to perform the method provided in this invention.

[0168] This invention also provides a chip, including: an input interface, an output interface, a processor, and a memory. The input interface, output interface, processor, and memory are connected through an internal connection path. The processor is used to execute code in the memory. When the code is executed, the processor is used to execute the method provided in this invention.

[0169] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. General-purpose processors can be microprocessors or any conventional processor. It is worth noting that the processor can be a processor supporting the Advanced Reduced Instruction Set Computing (RISC) machine (ARM) architecture.

[0170] Further, optionally, the aforementioned memory may include read-only memory and random access memory, and may also include non-volatile random access memory. The memory may be volatile or non-volatile, or may include both. Non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which serves as an external cache. Many forms of RAM are available by way of example, but not limitation. Examples include static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM).

[0171] In the above embodiments, implementation can be achieved, in whole or in part, by software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0172] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0173] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0174] 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 person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in the present invention, and these should all 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 method for evaluating energy-saving control programs in computer rooms, characterized in that, include: Multiple simulated data sets of a data center environment are acquired. These simulated data sets include first simulated data defined under a preset scenario and second simulated data generated based on a pre-built environment simulation model. The method for constructing the environment simulation model includes: acquiring signal data from a specified data source; integrating the signal data to obtain a first dataset containing device signals from all time steps; and training the memory network model based on the first dataset, using all device signals from the previous time step as input to the memory network model and the device signals from the next time step as output, to obtain the environment simulation model. The energy-saving control program to be tested processes multiple simulated data sets and outputs a control strategy corresponding to each simulated data set. The energy-saving control program is encapsulated from a pre-built energy-saving control model. The method for constructing the energy-saving control model includes: acquiring signal data from a specified data source; integrating the signal data to obtain a second dataset; training the learning model using equipment signals and environmental conditions as inputs and the optimized control strategy as outputs, based on the second dataset; and optimizing the learning model using a preset reward function to obtain the energy-saving control model. Obtain the expected control result corresponding to each of the simulated data, the expected control result including: a first expected control indicator defined when the target data center encounters the preset scenario with a specified problem, the first expected control indicator including expected control operation and expected response time; and a second expected indicator corresponding to the target data center in the intelligent scenario, the second expected indicator including alarm count, alarm duration, fault severity, equipment control frequency, and energy saving effect. Each of the aforementioned control strategies is compared with its corresponding desired control result to ultimately determine the evaluation result of the energy-saving control procedure.

2. The method for evaluating the energy-saving control program of a computer room according to claim 1, characterized in that, The step of comparing each of the control strategies with its corresponding expected control result to ultimately determine the evaluation result of the energy-saving control program includes: The first control strategy is compared with the first desired control index, and a first evaluation result is output; the first control strategy is the result output based on the analysis of the first simulation data by the energy-saving control model; and... The second control strategy is statistically analyzed to obtain the indicator results; the indicator results are compared with the second expected indicator to output the second evaluation result; wherein, the second control strategy is the result output by analyzing the second simulation data based on the energy-saving control model.

3. The method for evaluating the energy-saving control program of a computer room according to claim 1, characterized in that, Also includes: Based on the second dataset, multiple data subsets within different consecutive time periods are determined, and these multiple data subsets are processed to obtain initial state data for defining the initial state of the computer room environment. The environmental simulation model is initialized based on the initial state data, so that the environmental simulation model can simulate the dynamic changes of the computer room environment from the initial state.

4. The method for evaluating the energy-saving control program of a computer room according to claim 1, characterized in that, Also includes: The simulated data is stored in a designated storage unit, and the stored simulated data is backed up periodically through the designated storage unit to obtain backup data, which is then stored in the designated storage unit.

5. A data center energy-saving control program evaluation system, characterized in that, The evaluation method for the energy-saving control program of the computer room as described in any one of claims 1 to 4 shall be implemented.

6. An electronic device, characterized in that, include: A processor and a memory, wherein the memory stores instructions that are loaded and executed by the processor to implement the data center energy-saving control program evaluation method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the data center energy-saving control program evaluation method as described in any one of claims 1 to 4.

Images