Intelligent control method and system for cooling circulating water square bin of data center

Abstract

The invention relates to an artificial intelligence technology, and discloses an intelligent control method and system for a cooling circulating water square bin of a data center, and the method comprises the steps: constructing a digital twinborn model of a target cooling circulating water square bin, and obtaining a square bin twinborn model, predicting a future load curve based on the real-time environment data of the target cooling circulating water square bin and pre-acquired historical load data, and calculating the temperature of a water outlet of the target cooling circulating water square bin after a preset time interval by using the future load curve and a pre-acquired environment temperature change prediction curve to obtain a first predicted temperature, deducing the temperature of a water outlet of the target cooling circulating water square bin after a preset time interval according to the environment temperature change prediction curve and the future load curve by utilizing a square bin twinborn model to obtain a second prediction temperature, and generating a control instruction set based on a difference value between the first prediction temperature and the second prediction temperature, and adjusting the target cooling circulating water square bin based on the control instruction set. The operation efficiency of the cooling circulating water square bin can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to a wisdom control method and system for a cooling circulating water square tank of a data center. BACKGROUND

[0002] A data center is the core infrastructure of digital economy, and its energy consumption is huge, of which the energy consumption of the cooling system accounts for as high as 30% to 40%. Therefore, efficient and intelligent control of the cooling system of the data center is of great significance for reducing the overall energy consumption (PUE), operating cost and carbon emissions of the data center.

[0003] At present, the control of the cooling circulating water system of the data center generally adopts feedback control based on PID or simple start-stop rule strategy. These methods have inherent hysteresis and overshoot problems, and are difficult to cope with the large inertia characteristics of the system, resulting in slow response and high energy consumption. At the same time, the existing strategies are mostly single-point local optimization, lacking of collaborative control among components such as cooling towers, water pumps and chiller units, and cannot realize the global optimization of system energy efficiency. The core defect is that it completely depends on passive response and cannot use future load and environmental change prediction information for forward-looking regulation and control, which restricts the further breakthrough of the energy efficiency of the cooling system. SUMMARY

[0004] The present application provides a wisdom control method and system for a cooling circulating water square tank of a data center, which mainly aims to solve the problem of low efficiency and high power consumption of the existing cooling circulating water square tank control method.

[0005] To achieve the above purpose, the wisdom control method for a cooling circulating water square tank of a data center provided by the present application comprises: constructing a digital twin model of a target cooling circulating water square tank to obtain a square tank twin model; obtaining real-time load data of a preset data center and real-time environmental data of the target cooling circulating water square tank; performing future load prediction based on the real-time load data and pre-acquired historical load data to obtain a future load curve; calculating the outlet temperature of the target cooling circulating water square tank after a preset time interval by using the future load curve and a pre-acquired environmental temperature change prediction curve to obtain a first predicted temperature; deriving the outlet temperature of the target cooling circulating water square tank after a preset time interval according to the environmental temperature change prediction curve and the future load curve by using the square tank twin model to obtain a second predicted temperature; generating a control instruction set based on the difference between the first predicted temperature and the second predicted temperature, and adjusting the target cooling circulating water square tank based on the control instruction set.

[0006] Optionally, the digital twin model of the target cooling circulating water square tank is constructed, comprising: acquiring three-dimensional design data and real-time sensing data of the target cooling circulating water square tank; constructing a three-dimensional model according to the three-dimensional design data to obtain a square tank three-dimensional model; physically modeling according to the three-dimensional design data to obtain a hydraulic model, a thermodynamic model, and a device performance model; constructing a three-dimensional physical mechanism model according to the square tank three-dimensional model, the hydraulic model, the thermodynamic model, and the device performance model; compensating and calibrating the three-dimensional physical mechanism model using the real-time sensing data to obtain an error compensation model; constructing a digital twin of the target cooling circulating water square tank according to the error compensation model and the three-dimensional physical mechanism model to obtain a square tank twin model.

[0007] Optionally, the future load prediction is based on the real-time load data and the pre-acquired historical load data to obtain a future load curve, comprising: acquiring load data in the same time period as the real-time load data in the historical load data to obtain a comparative load data set; extracting a curve feature of the real-time load data to obtain a real-time curve feature, and extracting a curve feature of each comparative load data in the comparative load data set to obtain a comparative curve feature set; calculating a feature similarity between the real-time curve feature and each comparative curve feature in the comparative curve feature set to obtain a similarity set; judging whether the maximum similarity in the similarity set is greater than a preset similarity threshold; If greater, extract a periodic feature of the historical load data, and perform future load prediction based on the periodic feature and the real-time curve feature to obtain a future load curve; If less than or equal to, perform future load prediction according to pre-acquired business activity metadata and the real-time load data to obtain a future load curve.

[0008] Optionally, the future load prediction is based on the periodic feature and the real-time curve feature to obtain a future load curve, comprising: acquiring a timestamp of the maximum similarity corresponding feature curve and the real-time curve feature to obtain a first timestamp and a second timestamp; acquiring a plurality of periodic curve features between the first timestamp and the second timestamp in the periodic feature; Calculate the similarity of the real-time curve feature and each periodic curve feature respectively, and generate a fusion weight of each periodic curve feature according to the similarity of each periodic curve feature; Weight and fuse the periodic curve features between the first timestamp and the second timestamp in the periodic features based on the fusion weight of each periodic curve feature, to obtain a future load curve.

[0009] Optionally, the future load prediction according to the pre-acquired business activity metadata and the real-time load data to obtain a future load curve comprises: Obtaining a business activity sequence in a future preset time period according to the business activity metadata; Locating the timestamps of corresponding business activities in the historical load data according to the business activity sequence to obtain a set of activity timestamps; Extracting the load data of a preset time length before and after each timestamp in the set of activity timestamps in the historical load data to obtain a set of activity load data; Converting the load data in the set of activity load data into curve data to obtain a set of activity curves; Splicing the data curves in the set of activity curves according to the business activity sequence based on the real-time load data to obtain a spliced curve; Performing smoothing processing on the spliced curve to obtain a future load curve.

[0010] Optionally, the first predicted temperature is obtained by calculating the outlet temperature of the target cooling circulating water tank after a preset time interval using the future load curve and a pre-acquired environmental temperature change prediction curve, comprising: Obtaining historical outlet data and historical environmental temperature data of the target cooling circulating water tank; Constructing a dynamic parameter mapping table according to the historical outlet data and the historical environmental temperature data; Performing bilinear interpolation according to the dynamic parameter mapping table and the environmental temperature change prediction curve to obtain a dynamic parameter sequence; Performing temperature prediction iterative calculation based on a preset dynamic transfer function model according to the dynamic parameter sequence to obtain the first predicted temperature.

[0011] Optionally, the second predicted temperature is obtained by deducing the outlet temperature of the target cooling circulating water tank after a preset time interval using the environmental temperature change prediction curve and the future load curve according to the tank twin model, comprising: Performing state initialization on the tank twin model based on the real-time environmental data to obtain an initialized model; Using the initialization model, high-fidelity dynamic simulation is performed based on the environmental temperature change prediction curve, the future load curve, and the preset time step to obtain a complete state snapshot of all variables within each time step, thus obtaining a state data sequence. Extract the outlet temperature sequence from the state data sequence; The second predicted temperature is obtained based on the outlet temperature sequence.

[0012] Optionally, generating a control instruction set based on the difference between the first predicted temperature and the second predicted temperature includes: Calculate the difference between the first predicted temperature and the second predicted temperature to obtain the temperature difference. Determine whether the temperature difference is greater than a preset difference threshold; If it is greater than the first predicted temperature, then the maximum value between the first predicted temperature and the second predicted temperature is obtained to get the target temperature value; If it is less than or equal to, then the average of the first predicted temperature and the second predicted temperature is calculated to obtain the target temperature value; A set of control instructions is generated based on the target temperature value.

[0013] Optionally, generating a control instruction set based on the target temperature value includes: Based on the target temperature value, feedforward control calculations are performed to obtain feedforward control commands; Obtain the real-time outlet temperature of the target cooling circulating water container, and calculate the outlet temperature difference between the real-time outlet temperature and the target temperature value. Calculate the correction control quantity based on the temperature difference at the water outlet; An initial control command is generated based on the corrected control quantity and the feedforward control command; The initial control commands are subjected to multivariate coordination and constraint optimization to obtain a control command set.

[0014] To address the aforementioned problems, the present invention also provides a smart control system for a data center cooling circulating water tank, the system comprising: The model building module is used to construct a digital twin model of the target cooling circulating water container, thus obtaining the container twin model; The data acquisition module is used to acquire real-time load data of the preset data center and real-time environmental data of the target cooling circulating water container. The load prediction module is used to predict the future load based on the real-time load data and the pre-acquired historical load data, and obtain the future load curve. The temperature prediction module is used to calculate the outlet temperature of the target cooling circulating water tank after a preset time interval using the future load curve and the pre-acquired ambient temperature change prediction curve to obtain a first predicted temperature, and to use the tank twin model to extrapolate the outlet temperature of the target cooling circulating water tank after a preset time interval based on the ambient temperature change prediction curve and the future load curve to obtain a second predicted temperature. The instruction control module is used to generate a control instruction set based on the difference between the first predicted temperature and the second predicted temperature, and to adjust the target cooling circulating water silo based on the control instruction set.

[0015] This invention constructs a digital twin model of a target cooling water circulation tank, obtaining a tank twin model. It acquires real-time load data from a preset data center and real-time environmental data of the target cooling water circulation tank. Based on the real-time load data and pre-acquired historical load data, it predicts future load to obtain a future load curve. Using the future load curve and a pre-acquired ambient temperature change prediction curve, it calculates the outlet temperature of the target cooling water circulation tank after a preset time interval to obtain a first predicted temperature. Using the tank twin model, it extrapolates the outlet temperature of the target cooling water circulation tank after the preset time interval based on the ambient temperature change prediction curve and the future load curve to obtain a second predicted temperature. Based on the difference between the first and second predicted temperatures, it generates a control command set and adjusts the target cooling water circulation tank based on the control command set. Therefore, the intelligent control method and system for data center cooling water circulation tanks proposed in this invention can solve the problems of low efficiency and high power consumption in existing cooling water circulation tank control methods. Attached Figure Description

[0016] Figure 1 A flowchart illustrating a smart control method for a data center cooling circulating water tank according to an embodiment of the present invention; Figure 2 This is a functional block diagram of an intelligent control system for a data center cooling circulating water silo provided in an embodiment of the present invention.

[0017] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] This application provides a smart control method for a data center cooling water circulation tank. The executing entity of the smart control method for the data center cooling water circulation tank includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the smart control method for the data center cooling water circulation tank can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0020] Reference Figure 1 The diagram shown is a flowchart illustrating a smart control method for a data center cooling water circulation tank according to an embodiment of the present invention. In this embodiment, the smart control method for the data center cooling water circulation tank includes: S1. Construct a digital twin model of the target cooling circulating water container to obtain the container twin model.

[0021] In this embodiment of the invention, the digital twin model refers to a dynamic, high-fidelity virtual model constructed in virtual space that completely corresponds to a physical entity (such as a cooling silo) using sensors, the Internet of Things, and data modeling technology. It achieves real-time monitoring, performance simulation, fault prediction, and optimized control of the physical entity's operation process by mapping the physical entity's state (such as temperature, pressure, and equipment operating parameters) in real time and integrating physical mechanisms with data-driven algorithms, thereby forming an intelligent system that interacts and evolves in parallel with the physical entity.

[0022] In this embodiment of the invention, constructing a digital twin model of the target cooling circulating water container includes: Acquire the three-dimensional design data and real-time sensor data of the target cooling circulating water container; A 3D model is constructed based on the 3D design data to obtain the 3D model of the warehouse. Physical modeling is performed based on the three-dimensional design data to obtain hydraulic model, thermodynamic model, and equipment performance model; A three-dimensional physical mechanism model is constructed based on the three-dimensional model of the container, the hydraulic model, the thermodynamic model, and the equipment performance model. The real-time sensing data is used to compensate and calibrate the three-dimensional physical mechanism model to obtain an error compensation model. A digital twin of the target cooling circulating water container is constructed based on the error compensation model and the three-dimensional physical mechanism model, resulting in a container twin model.

[0023] In detail, the three-dimensional design data may include design (CAD) drawings, piping and instrumentation diagrams, equipment lists and their specifications.

[0024] In detail, the real-time sensing data may include water pump frequency, valve opening degree, cooling tower fan speed, temperature at each level, pressure, flow rate, and power consumption.

[0025] In detail, the physical modeling based on the three-dimensional design data involves constructing a mathematical model that describes the system's intrinsic working mechanism using the laws of physics. The hydraulic model calculates the pressure distribution and flow rate allocation in the pipeline network based on fluid dynamics equations (such as Bernoulli's equation and Darcy's formula); the thermodynamic model calculates temperature changes and heat exchange efficiency in components such as heat exchangers and cooling towers based on heat transfer equations; and the equipment performance model fits the performance curves of pumps and fans, as well as the COP characteristics of chillers, into mathematical functions.

[0026] In detail, the construction of a three-dimensional physical mechanism model based on the three-dimensional model of the container, the hydraulic model, the thermodynamic model, and the equipment performance model involves the deep integration of these four models. The core of this process lies in establishing a one-to-one mapping relationship between spatial geometric entities and physical calculation units. For example, a section of pipe in the three-dimensional model is associated with a hydraulic calculation unit, using its length and diameter as input parameters for the hydraulic model; a three-dimensional model of a heat exchanger is associated with a thermodynamic calculation unit. In this way, a computable, high-fidelity simulation model that follows physical laws is created in a visualized three-dimensional space.

[0027] In detail, the compensation and calibration of the three-dimensional physical mechanism model using the real-time sensing data is achieved by introducing a data-driven method to continuously compare the calculation results of the three-dimensional physical mechanism model with the measured values ​​of the real-time sensing data to calculate the error. Then, a machine learning algorithm is used to train an error compensation model that can learn the complex nonlinear relationship between these residuals and the current operating conditions (such as load and ambient temperature).

[0028] In detail, the construction of the digital twin of the target cooling circulating water container based on the error compensation model and the three-dimensional physical mechanism model refers to the organic integration of the error compensation model and the three-dimensional physical mechanism model. Its output can be expressed as: digital twin prediction value = mechanism model prediction value + error compensation model prediction value.

[0029] S2. Obtain real-time load data of the preset data center and real-time environmental data of the target cooling circulating water container.

[0030] In this embodiment of the invention, the acquisition of real-time load data of the preset data center and real-time environmental data of the target cooling circulating water tank can be achieved by using a monitoring agent deployed inside the data center to collect load indicators such as power consumption, CPU utilization, and network traffic of its IT equipment in real time, while simultaneously collecting environmental parameters such as temperature, humidity, and wind speed through a sensor network installed inside and around the cooling tank, thereby providing accurate, synchronous, and real-time multi-dimensional data input for subsequent intelligent control decisions.

[0031] S3. Based on the real-time load data and the previously acquired historical load data, predict the future load to obtain the future load curve.

[0032] In this embodiment of the invention, the step of predicting future load based on the real-time load data and pre-acquired historical load data to obtain a future load curve includes: Obtain load data from the historical load data that is in the same time period as the real-time load data to obtain a set of comparison load data; Extract the curve features of the real-time load data to obtain real-time curve features, and extract the curve features of each comparative load data in the comparative load data set to obtain a comparative curve feature set; Calculate the feature similarity between the real-time curve feature and each comparative curve feature in the comparative curve feature set to obtain a similarity set; Determine whether the maximum similarity in the similarity set is greater than a preset similarity threshold; If it is greater than the value, then the periodic characteristics of the historical load data are extracted, and the future load is predicted based on the periodic characteristics and the real-time curve characteristics to obtain the future load curve. If the load is less than or equal to the load, then the future load is predicted based on the pre-acquired business activity metadata and the real-time load data to obtain the future load curve.

[0033] In detail, obtaining the load data from the historical load data that falls within the same time period as the real-time load data to obtain the comparison load data set involves locating historical segments within the historical load data that perfectly correspond in time to the current real-time load data. For example, if the current real-time data is a load sequence from 10:00 AM to 11:00 AM this Wednesday, the system will extract all historical load data segments from 10:00 AM to 11:00 AM on Wednesdays to form the "comparison load data." This is done to eliminate differences caused by inherent factors such as date type and hourly cycle.

[0034] In detail, the process involves extracting curve features from the real-time load data to obtain real-time curve features, and extracting curve features from each comparative load data point in the comparative load data set to obtain a comparative curve feature set. This aims to transform the raw time-series data into quantitative features that better reflect its inherent patterns. The system employs signal processing or time-series analysis techniques to extract a series of key features from the "real-time load data" and "comparative load data." These features may include, but are not limited to: the mean (representing the average level), variance or standard deviation (representing the degree of fluctuation), slope (representing an upward or downward trend), periodicity, and peaks or troughs at specific time points. By extracting these features, the complex problem of curve comparison is transformed into a more operational problem of feature vector comparison.

[0035] In detail, the calculation of the feature similarity between the real-time curve features and each comparative curve feature in the comparative curve feature set can be performed by calculating cosine similarity (which measures the difference in the direction of feature vectors), Euclidean distance (which measures the absolute distance in the feature vector space), or dynamic time warping (DTW) distance (which is specifically used to measure the similarity of the shapes of time series that may have different lengths).

[0036] Specifically, determining whether the maximum similarity in the similarity set is greater than a preset similarity threshold is a decision branch point. The maximum similarity value is compared with a threshold pre-set through experience or experimentation. This threshold is a key parameter that defines how the system distinguishes between "normal mode" and "abnormal mode." If the similarity is greater than the threshold, it is determined that the current load is operating according to historical patterns; conversely, it is determined that the current load has deviated abnormally, possibly affected by unknown or sudden factors. This judgment logic is the foundation for the entire scheme to achieve intelligent decision-making.

[0037] In this embodiment of the invention, the step of predicting future load based on the periodic characteristics and the real-time curve characteristics to obtain the future load curve includes: Obtain the feature curve corresponding to the maximum similarity and the timestamp of the real-time curve feature to obtain the first timestamp and the second timestamp; Obtain multiple periodic curve features between the first and second timestamps in the periodic features; The similarity between the real-time curve feature and each periodic curve feature is calculated respectively, and the fusion weight of each periodic curve feature is generated based on the similarity of each periodic curve feature. The periodic curve features between the first and second timestamps in the periodic features are weighted and fused based on the fusion weight of each periodic curve feature to obtain the future load curve.

[0038] In detail, obtaining multiple periodic curve features between the first and second timestamps in the periodic features involves extracting multiple complete periodic data within a specific time period from the periodic features based on the first and second timestamps. For example, if there are 5 days between the first and second timestamps, and the main cycle is 1 day, then the system will extract 5 daily periodic curve features corresponding to these 5 days.

[0039] In detail, the process of generating the fusion weight of each periodic curve feature based on the similarity of each periodic curve feature means that the more similar a historical period is to the current real-time pattern, the higher its weight; the lower the similarity, the lower the weight, and it can even be ignored.

[0040] In this embodiment of the invention, the step of predicting future load based on pre-acquired business activity metadata and the real-time load data to obtain a future load curve includes: Obtain the sequence of business activities within a future preset time period based on the business activity metadata; Based on the business activity sequence, locate the timestamp of the corresponding business activity in the historical load data to obtain the activity timestamp set; Extract load data from the historical load data for each timestamp in the activity timestamp set, within a preset time length before and after each timestamp, to obtain the activity load data set. The load data in the active load data set is converted into curve data to obtain an active curve set; Based on the real-time load data, the data curves in the activity curve set are spliced ​​together according to the business activity sequence to obtain the spliced ​​curve; The spliced ​​curve is smoothed to obtain the future load curve.

[0041] In detail, the business activity metadata is a planned time with clear business meaning. It is different from a continuous physical time stream. Instead, it is a series of discrete time points or time periods defined by the enterprise's work plan. Its core value lies in describing when and what type of business activity (such as data backup, promotional activities, system release, etc.) will occur. Different business activities will have different impacts on the data center load.

[0042] In detail, the step of locating the timestamp of the corresponding business activity in the historical load data according to the business activity sequence to obtain the activity timestamp set refers to locating the timestamps in the historical load data that have had the same business activity as in the business activity sequence.

[0043] In detail, the extraction of load data for each time stamp in the activity timestamp set within the historical load data for a preset time length before and after the timestamp means extracting a preset time length (e.g., from 1 hour before the start of the activity to 2 hours after the end of the activity) of historical load data centered on each timestamp.

[0044] In detail, the process of using the real-time load data as a benchmark and stitching together the data curves in the activity curve set according to the business activity sequence to obtain a stitched curve uses the latest real-time load data value as the starting point for prediction, ensuring the continuity between the prediction and the current system state. Then, according to the time arrangement of the business activity sequence, corresponding curve templates are selected from the activity curve set, and these templates are stitched together sequentially.

[0045] In detail, the smoothing process of the spliced ​​curve to obtain the future load curve can be achieved by using data smoothing algorithms (such as moving average filtering, Gaussian filtering, or spline interpolation) to process these transition intervals of the spliced ​​curve, eliminating unnatural abrupt changes and making the entire prediction curve smoother and more reasonable.

[0046] S4. Calculate the outlet temperature of the target cooling circulating water container after a preset time interval using the future load curve and the pre-acquired ambient temperature change prediction curve to obtain the first predicted temperature.

[0047] In this embodiment of the invention, the step of calculating the outlet temperature of the target cooling circulating water tank after a preset time interval using the future load curve and the pre-acquired ambient temperature change prediction curve to obtain the first predicted temperature includes: Obtain historical outlet data and historical ambient temperature data of the target cooling circulating water container; A dynamic parameter mapping table is constructed based on the historical outlet data and the historical ambient temperature data. Based on the dynamic parameter mapping table and the environmental temperature change prediction curve, bilinear interpolation is performed to obtain the dynamic parameter sequence; Based on a preset dynamic transfer function model, temperature prediction iterative calculations are performed according to the dynamic parameter sequence to obtain the first predicted temperature.

[0048] In detail, the construction of the dynamic parameter mapping table based on the historical outlet data and the historical ambient temperature data can be achieved using machine learning algorithms (such as recursive least squares) to analyze the dynamic relationship between historical outlet temperature and ambient temperature, and to train a mathematical model based on this analysis. The parameters of this model will change with operating conditions (such as different ambient temperature ranges). Finally, the system stores all operating conditions and their corresponding optimal model parameters in a two-dimensional lookup table, resulting in the dynamic parameter mapping table.

[0049] In detail, the bilinear interpolation based on the dynamic parameter mapping table and the environmental temperature change prediction curve refers to the system performing bilinear interpolation calculations on the dynamic parameter mapping table for each future time point on the prediction curve, based on the predicted environmental temperature value at that point. Specifically, this involves calculating a set of optimal model parameters for each future time point, thereby transforming the originally discrete parameter mapping table into a continuous, smooth dynamic parameter sequence that perfectly matches future environmental changes.

[0050] In detail, the preset dynamic transfer function model performs temperature prediction iterative calculations based on the dynamic parameter sequence to obtain the first predicted temperature. Starting from the currently known outlet temperature, the model iteratively calculates step by step according to the time sequence, and finally deduces the outlet temperature at each moment in the future period, thereby obtaining the first predicted temperature.

[0051] S5. Using the twin model of the container, the outlet temperature of the target cooling circulating water container after a preset time interval is deduced based on the environmental temperature change prediction curve and the future load curve, and a second predicted temperature is obtained.

[0052] In this embodiment of the invention, the environmental temperature change prediction curve is mainly obtained by deeply integrating the future wet-bulb temperature forecast provided by the professional meteorological data service API interface with the real-time micro-environmental data (such as temperature, humidity, and wind speed) collected by the micro-meteorological station deployed in the data center, and then dynamically calibrating and locally correcting it through a machine learning model.

[0053] In this embodiment of the invention, the step of using the container twin model to extrapolate the outlet temperature of the target cooling circulating water container after a preset time interval based on the environmental temperature change prediction curve and the future load curve, and obtaining the second predicted temperature, includes: Based on the real-time environmental data, the state of the Fangcang twin model is initialized to obtain the initialized model; Using the initialization model, high-fidelity dynamic simulation is performed based on the environmental temperature change prediction curve, the future load curve, and the preset time step to obtain a complete state snapshot of all variables within each time step, thus obtaining a state data sequence. Extract the outlet temperature sequence from the state data sequence; The second predicted temperature is obtained based on the outlet temperature sequence.

[0054] In detail, the state initialization of the warehouse twin model based on the real-time environmental data refers to modulating the variables in the warehouse twin model according to the real-time environmental data to complete the initialization.

[0055] In detail, the high-fidelity dynamic simulation using the initialization model based on the predicted environmental temperature change curve, the future load curve, and a preset time step involves using the predicted environmental temperature change curve and the future load curve as external inputs to drive the model's operation. The simulation process advances the virtual time step by step according to the preset time step (e.g., one step every minute). Within each time step, the model comprehensively calculates the dynamic response of the system under all input influences based on its inherent physical mechanisms (hydraulic and thermodynamic equations) and data-driven rules, thereby outputting a complete snapshot of the state of all variables at that moment.

[0056] In detail, obtaining the second predicted temperature based on the outlet temperature sequence means obtaining temperature data at a specific time point from the outlet temperature sequence.

[0057] S6. Generate a control instruction set based on the difference between the first predicted temperature and the second predicted temperature, and adjust the target cooling circulating water silo based on the control instruction set.

[0058] In this embodiment of the invention, the step of generating a control instruction set based on the difference between the first predicted temperature and the second predicted temperature includes: Calculate the difference between the first predicted temperature and the second predicted temperature to obtain the temperature difference. Determine whether the temperature difference is greater than a preset difference threshold; If it is greater than the first predicted temperature, then the maximum value between the first predicted temperature and the second predicted temperature is obtained to get the target temperature value; If it is less than or equal to, then the average of the first predicted temperature and the second predicted temperature is calculated to obtain the target temperature value; A set of control instructions is generated based on the target temperature value.

[0059] In detail, the calculation of the difference between the first predicted temperature and the second predicted temperature intuitively reflects the degree of discrepancy between theoretical calculations based on historical data and virtual models based on real-time simulation in judging the future state of the system. It is a core indicator for assessing the uncertainty and potential risks of system prediction.

[0060] In this embodiment of the invention, generating a control instruction set based on the target temperature value includes: Based on the target temperature value, feedforward control calculations are performed to obtain feedforward control commands; Obtain the real-time outlet temperature of the target cooling circulating water container, and calculate the outlet temperature difference between the real-time outlet temperature and the target temperature value. Calculate the correction control quantity based on the temperature difference at the water outlet; An initial control command is generated based on the corrected control quantity and the feedforward control command; The initial control commands are subjected to multivariate coordination and constraint optimization to obtain a control command set.

[0061] In detail, the feedforward control calculation based on the target temperature value to obtain the feedforward control command involves performing forward calculations based on known and impending disturbances (such as future load curves and predicted ambient temperature change curves) and a pre-set thermodynamic model. The purpose is to pre-calculate the theoretically required operating state of the cooling system (such as water pumps and cooling tower fans) to counteract the effects of these future disturbances and stabilize the outlet temperature at the target value. The resulting feedforward control command is a basic open-loop control quantity that provides the system with preliminary, forward-looking adjustment actions, significantly reducing control lag and improving system response speed.

[0062] In detail, the step of calculating the corrected control quantity based on the outlet temperature difference is based on a preset PID controller algorithm to calculate a control quantity that eliminates the deviation that the feedforward control could not fully compensate for, according to the outlet temperature difference and the trend of the difference.

[0063] In detail, the step of generating an initial control command based on the corrected control quantity and the feedforward control command is to generate an initial control command by adding the feedforward control command and the corrected control quantity.

[0064] In detail, the multi-variable collaborative and constraint optimization of the initial control commands to obtain the control command set represents a system-level global optimization. The initial control commands were calculated independently for each individual actuator, without fully considering the coupling relationships between multiple variables within the system (such as pump frequency, fan speed, and valve opening) and physical constraints (such as equipment safe operating range and minimum flow limits). This step treats all these initial commands as a whole, using the lowest total system energy consumption or highest overall efficiency as the optimization objective, and rapidly solves and optimizes them within a mathematical model that considers all constraints. The final output control command set is a coordinated and feasible set of optimal commands, ensuring that the entire cooling system operates safely, stably, and efficiently as an organic whole, thus achieving a leap from "single-point control" to "global optimization."

[0065] like Figure 2 The diagram shown is a functional block diagram of an intelligent control system for a data center cooling circulating water silo provided in an embodiment of the present invention.

[0066] The intelligent control system 100 for the data center cooling water circulation tank described in this invention can be installed in an electronic device. Depending on the functions implemented, the intelligent control system 100 for the data center cooling water circulation tank may include a model building module 101, a data acquisition module 102, a load prediction module 103, a temperature prediction module 104, and an instruction control module 105. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and which are stored in the memory of the electronic device.

[0067] In this embodiment, the functions of each module / unit are as follows: The model building module 101 is used to build a digital twin model of the target cooling circulating water container to obtain the container twin model; The data acquisition module 102 is used to acquire real-time load data of the preset data center and real-time environmental data of the target cooling circulating water container. The load prediction module 103 is used to predict the future load based on the real-time load data and the pre-acquired historical load data, and obtain the future load curve. The temperature prediction module 104 is used to calculate the outlet temperature of the target cooling circulating water tank after a preset time interval using the future load curve and the pre-acquired ambient temperature change prediction curve to obtain a first predicted temperature, and to use the tank twin model to deduce the outlet temperature of the target cooling circulating water tank after a preset time interval based on the ambient temperature change prediction curve and the future load curve to obtain a second predicted temperature. The instruction control module 105 is used to generate a control instruction set based on the difference between the first predicted temperature and the second predicted temperature, and to adjust the target cooling circulating water silo based on the control instruction set.

[0068] In detail, the modules in the intelligent control system 100 for the data center cooling circulating water tank described in this embodiment of the invention employ the same methods as described above during use. Figure 1 The intelligent control method for the data center cooling circulating water silo described herein uses the same technical means and can produce the same technical effect, so it will not be elaborated here.

[0069] In the embodiments provided by this invention, it should be understood that the disclosed devices, systems, and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0070] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0071] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0072] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0073] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0074] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0075] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or systems stated in a system claim may also be implemented by a single unit or system through software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any specific order.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent control of a data center cooling circulating water tank, characterized in that, The method comprises: constructing a digital twin model of the target cooling circulating water square tank to obtain a square tank twin model; obtaining real-time load data of a preset data center and real-time environmental data of the target cooling circulating water square tank; based on the real-time load data and the pre-acquired historical load data, predicting future load to obtain a future load curve; using the future load curve and the pre-acquired environmental temperature change prediction curve to calculate the outlet temperature of the target cooling circulating water square tank after a preset time interval to obtain a first predicted temperature; using the square tank twin model to deduce the outlet temperature of the target cooling circulating water square tank after a preset time interval according to the environmental temperature change prediction curve and the future load curve to obtain a second predicted temperature; based on the difference between the first predicted temperature and the second predicted temperature, generating a control instruction set, and adjusting the target cooling circulating water square tank based on the control instruction set. 2.The intelligent control method of the data center cooling circulating water square tank according to claim 1, wherein, The method comprises: acquiring three-dimensional design data and real-time sensing data of the target cooling circulating water square tank; constructing a three-dimensional model according to the three-dimensional design data to obtain a square tank three-dimensional model; performing physical modeling according to the three-dimensional design data to obtain a hydraulic model, a thermodynamic model, and a device performance model; constructing a three-dimensional physical mechanism model according to the square tank three-dimensional model, the hydraulic model, the thermodynamic model, and the device performance model; compensating and calibrating the three-dimensional physical mechanism model using the real-time sensing data to obtain an error compensation model; constructing a digital twin of the target cooling circulating water square tank according to the error compensation model and the three-dimensional physical mechanism model to obtain a square tank twin model. 3.The intelligent control method of the data center cooling circulating water tank according to claim 1, wherein, The method comprises: acquiring load data in the same time period as the real-time load data from the historical load data to obtain a comparison load data set; extracting the curve features of the real-time load data to obtain real-time curve features, and extracting the curve features of each comparison load data in the comparison load data set to obtain a comparison curve feature set; calculating the feature similarity of the real-time curve features and each comparison curve feature in the comparison curve feature set to obtain a similarity set; determining whether the maximum similarity in the similarity set is greater than a preset similarity threshold; if greater, extracting the periodic features of the historical load data, and predicting future load based on the periodic features and the real-time curve features to obtain a future load curve; if less than or equal to, predicting future load based on pre-acquired business activity metadata and the real-time load data to obtain a future load curve. 4.The intelligent control method of the data center cooling circulating water tank according to claim 3, wherein, The method comprises: acquiring the time stamps of the maximum similarity corresponding feature curve and the real-time curve features to obtain a first time stamp and a second time stamp; acquire a plurality of periodic curve features between the first timestamp and the second timestamp in the periodic feature; respectively calculate a similarity of the real-time curve feature and each periodic curve feature, and generate a fusion weight of each periodic curve feature according to the similarity of each periodic curve feature; perform weighted fusion on the periodic curve features between the first timestamp and the second timestamp in the periodic feature based on the fusion weight of each periodic curve feature, to obtain a future load curve. 5.The intelligent control method of the data center cooling circulating water square tank according to claim 3, wherein, The future load prediction according to the pre-acquired business activity metadata and the real-time load data to obtain a future load curve comprises: acquiring a business activity sequence in a future preset time period according to the business activity metadata; locating a timestamp of corresponding business activity in the historical load data according to the business activity sequence to obtain an activity timestamp set; extracting load data of a preset time length before and after each timestamp in the activity timestamp set in the historical load data to obtain an activity load data set; converting the load data in the activity load data set into curve data to obtain an activity curve set; splicing data curves in the activity curve set according to the business activity sequence with the real-time load data as a reference to obtain a spliced curve; performing smoothing processing on the spliced curve to obtain a future load curve. 6.The intelligent control method of the data center cooling circulating water tank of claim 1, wherein, The first prediction temperature is obtained by calculating the outlet temperature of the target cooling circulating water tank after a preset time interval using the future load curve and a pre-acquired environmental temperature change prediction curve, comprising: acquiring historical outlet data and historical environmental temperature data of the target cooling circulating water tank; constructing a dynamic parameter mapping table according to the historical outlet data and the historical environmental temperature data; performing bilinear interpolation according to the dynamic parameter mapping table and the environmental temperature change prediction curve to obtain a dynamic parameter sequence; performing temperature prediction iterative calculation based on a preset dynamic transfer function model according to the dynamic parameter sequence to obtain the first prediction temperature. 7.The intelligent control method of the data center cooling circulating water tank according to claim 6, wherein, The second prediction temperature is obtained by deducing the outlet temperature of the target cooling circulating water tank after a preset time interval using the environmental temperature change prediction curve and the future load curve according to the tank twin model, comprising: performing state initialization on the tank twin model based on the real-time environmental data to obtain an initialized model; performing high-fidelity dynamic simulation deduction according to the environmental temperature change prediction curve, the future load curve and a preset time step using the initialized model to obtain a complete state snapshot of all variables in each time step, to obtain a state data sequence; extracting an outlet temperature sequence from the state data sequence; obtaining the second prediction temperature according to the outlet temperature sequence. 8.The intelligent control method of the data center cooling circulating water tank of claim 1, wherein, The control instruction set is generated based on the difference between the first prediction temperature and the second prediction temperature, comprising: calculating the difference between the first prediction temperature and the second prediction temperature to obtain a temperature difference; determining whether the temperature difference is greater than a preset difference threshold; If greater, a maximum value between the first predicted temperature and the second predicted temperature is obtained to obtain a target temperature value; If less than or equal to, a mean value of the first predicted temperature and the second predicted temperature is calculated to obtain a target temperature value; A control instruction set is generated according to the target temperature value. 9.The intelligent control method of the data center cooling circulating water tank according to claim 8, wherein, The generating of the control instruction set according to the target temperature value comprises: Feedforward control calculation is performed according to the target temperature value to obtain a feedforward control instruction; A real-time outlet temperature of the target cooling circulating water tank is obtained, and an outlet temperature difference value between the real-time outlet temperature and the target temperature value is calculated; A correction control amount is calculated according to the outlet temperature difference value; An initial control instruction is generated according to the correction control amount and the feedforward control instruction; The initial control instruction is subjected to multivariable coordination and constraint optimization to obtain a control instruction set.

10. An intelligent control system for a data center cooling circulating water square tank, characterized in that, The system comprises: A model construction module is configured to construct a digital twin model of a target cooling circulating water tank to obtain a tank twin model; A data acquisition module is configured to acquire real-time load data of a preset data center and real-time environmental data of the target cooling circulating water tank; A load prediction module is configured to predict future load based on the real-time load data and historical load data acquired in advance to obtain a future load curve; A temperature prediction module is configured to calculate an outlet temperature of the target cooling circulating water tank after a preset time interval by using the future load curve and an environmental temperature change prediction curve acquired in advance to obtain a first predicted temperature, and to deduce the outlet temperature of the target cooling circulating water tank after the preset time interval by using the tank twin model according to the environmental temperature change prediction curve and the future load curve to obtain a second predicted temperature; An instruction control module is configured to generate a control instruction set based on a difference between the first predicted temperature and the second predicted temperature, and to adjust the target cooling circulating water tank based on the control instruction set.

Images