Machine room cold channel closed management method and device and terminal equipment

Abstract

The invention relates to the technical field of data center machine room energy conservation, and discloses a machine room cold channel closed management method and device and terminal device.The method comprises the following steps that the state of a cold channel is obtained by deploying environment monitoring nodes, a heat-flow coupling model and a stability function are constructed, a feedback controller is designed to generate an air door and refrigeration control instruction, and a heat-flow coupling model is established; linkage adjustment of the air velocity and the temperature is achieved, closed-loop control is constructed, and efficient and energy-saving management of a cold channel is achieved; the device comprises an environment monitoring module, a state modeling module, a stability evaluation module, a state feedback control module and an execution linkage module. Through state fusion updating, a component control structure, a stability function driving strategy and a linkage execution mechanism, fine adjustment and efficient cooperative control of the temperature and airflow of the cold channel are achieved, the adjustment stability and response efficiency are improved, energy consumption is remarkably reduced, and the problems that an existing system is poor in control precision, poor in linkage performance and the like are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of energy conservation in data center computer rooms, and in particular to a method, device and terminal equipment for enclosed cold channel management in a computer room. Background Art

[0002] In the current context of green data center construction, energy efficiency has become a core consideration. Cold aisle containment systems are widely deployed to reduce energy waste caused by the mixing of hot and cold airflows. However, traditional cold aisle control methods mostly rely on fixed physical enclosures, such as simple partitions using fences, doors, or covers. While these solutions achieve initial isolation between hot and cold, they lack inherent sensing capabilities and are unable to respond to real-time changes in hot and cold conditions.

[0003] Some systems currently attempt to incorporate basic sensor monitoring methods, such as temperature probes and infrared detection. However, these devices often have limited layouts, limited feedback, and fail to form an effective feedback control loop. Common practices still rely on manual inspections or timed adjustments to dampers and air conditioning parameters. The entire system operates essentially in a "passive response" state, with coarse adjustment granularity and long adjustment cycles, making it difficult to adapt to the rapidly changing load environment of the computer room. In particular, the system often fails to detect even the slightest changes in airtightness, resulting in unnoticed waste of cooling capacity.

[0004] Furthermore, existing control strategies generally lack global optimization capabilities based on thermal-fluid coupling. Control commands are often issued point-to-point, with no coordinated logic between dampers and refrigeration equipment. Adjustment directions are unclear, and the execution sequence can be inconsistent. If heat accumulates in a specific area, the system may simply increase cooling power while ignoring the reconfiguration of airflow paths, resulting in frequent instances of "cooling without effect." The resulting decrease in cooling efficiency and increase in energy consumption are difficult to fundamentally address with traditional methods.

[0005] Therefore, the present invention proposes a method, device and terminal equipment for closed management of cold aisles in a computer room to address the deficiencies of the prior art. Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the present invention provides a method, device and terminal equipment for closed management of cold channels in a computer room, which solves the problems of the existing system in achieving high-efficiency control due to incomplete state perception, delayed adjustment response and poor equipment linkage.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for managing the closed cold aisle of a computer room, comprising the following steps: S1. Deploy environmental monitoring nodes at key locations within the cold aisle area to collect real-time air temperature and air velocity at key locations, and fuse the environmental parameters of each key location to construct a cold aisle state vector. S2. Input the state vector into a thermal-fluid coupling state-space modeling module to establish a state-space model of the time-varying temperature and airflow velocity inside the cold aisle. The model is used to dynamically describe the state evolution process of the system under multi-source disturbance conditions. S3. Based on the state-space model, a stability function is constructed to evaluate the stability of the cold aisle thermal-flow field. The stability function is represented by a quadratic function consisting of the state vector and a symmetric positive definite matrix, and is used to monitor the degree of stability deviation of the system under the influence of the current disturbance in real time. S4. Based on the evolution trend of the stability function, a state feedback controller is designed to calculate the control instructions of the cold aisle execution structure using the current state vector. The control instructions include the damper angle control signal and the refrigeration unit power control signal. S5. Synchronously output the control command to the damper actuator and cooling equipment in the cold channel, jointly adjust the air flow rate and temperature distribution inside the cold channel, and update the state vector to enter the next cycle, so as to form a closed-loop control process for continuous monitoring and adjustment.

[0008] Preferably, the key positions include air passages at the cold channel air inlet, the front side of the server rack, the rear side of the server rack, and the cold channel outlet.

[0009] Preferably, the environmental monitoring node includes: Temperature sensors, used to detect air temperature at key locations; Wind speed sensor, used to detect air flow speed at key locations; Humidity sensor, used to monitor air humidity; The data acquisition module is used to integrate the measurement data of various sensors and upload them to the host computer.

[0010] Preferably, the state space model is established in the following form: x(t+1)=Ax(t)+Bu(t)+w(t); Among them, x(t) represents the state vector, u(t) represents the control instruction vector, w(t) represents the disturbance term, and A and B are system matrices.

[0011] Preferably, the stability function is a quadratic function of the following form: V(x)=x T Px; Among them, x is the current state vector, x T is the transpose of x, and P is a symmetric positive definite matrix used to measure the system energy level and stability deviation.

[0012] Preferably, the state feedback controller constructs the following optimization objective by minimizing the system deviation based on the real-time change trend of the stability function: Where u(t) is the control input vector at time t, x(t+1) represents the system state vector at time t+1, P is a symmetric positive definite matrix, and x(t+1) T Px(t+1) represents the stability function, which is a quadratic function, and V(x(t+1)) represents the stability function value of the system at the next moment; And according to the system model x(t+1)=Ax(t)+Bu(t), the optimal control input u(t) is calculated, and the control input includes: The damper angle adjustment signal based on the disturbance response is used to correct the cold air flow path; Cooling power adjustment signal based on heat load fluctuation to combat temperature anomalies; The feedback controller is implemented using a linear quadratic regulator method or an improved algorithm thereof, and has the ability to adaptively adjust gain parameters to meet the control accuracy requirements under dynamic changes in different server loads.

[0013] Preferably, the damper angle control signal drives an adjustable guide plate arranged at the top or bottom of the cold channel to adjust the spatial distribution of the airflow direction and flow rate by linking multiple air guide actuators, wherein: When the monitoring node detects that the temperature at the front end of a rack exceeds the set threshold, the controller will prioritize adjusting the opening angle of the air guide plate above that area to increase the supply of cold air to that area. The opening adjustment of multiple air guide plates is linked and optimized based on the changing trend of the local temperature gradient in the state vector, thereby forming a directional cooling path facing the hot spot area.

[0014] Preferably, the refrigeration unit power control signal is used to dynamically control the refrigeration capacity of the refrigeration module in the cooling system, and its control logic includes: Predicting temperature trends over multiple future cycles using the state-space model allows for pre-adjustment of cooling unit output power to avoid hysteresis effects. In non-uniform heat load scenarios, cooling priority levels are divided according to the distribution characteristics of the state vector, and a hierarchical control strategy is adopted to apply different target powers to each cooling module to achieve a balance between local enhanced cooling and global energy efficiency optimization; The control signal also includes a compound control variable for the compressor frequency, cooling water flow or air supply speed, which is used to improve the overall cooling response bandwidth.

[0015] A closed management device for cold aisles in a computer room, comprising: Environmental monitoring module, used to collect air temperature, air velocity and humidity information at multiple key locations in the cold aisle and generate state vectors; A state modeling module is used to construct a state space model of the cold channel heat-flow field based on the state vector; A stability evaluation module is used to calculate the stability function and evaluate the stability deviation degree of the cold channel; A state feedback control module is used to calculate the damper angle and the refrigeration unit power control signal according to the stability function; An execution linkage module is used to output control signals to the damper actuator and cooling equipment to adjust the airflow and temperature distribution in the cold channel. A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for closed management of the cold channel in the computer room is implemented.

[0016] The present invention provides a method, device, and terminal device for managing the sealed cold aisle in a computer room. It has the following beneficial effects: 1. This invention achieves dynamic correction and filtering of state feedback information by constructing a state update mechanism that integrates measured data with model predictions, effectively improving system control stability and response continuity. Compared with prior art methods that rely on direct drive control based on single-point real-time measurements, this invention reduces feedback jitter caused by environmental fluctuations and sampling errors, addressing the prior art issue of state acquisition errors that can easily lead to system overregulation.

[0017] 2. This invention utilizes a component control structure that differentiates damper angle and cooling power, enabling independent control of different physical zones with high adjustment precision and flexibility. Compared to existing technologies that evenly distribute cooling capacity and impose coarse control granularity, this invention finely adjusts airflow and cooling intensity within the cold aisle based on local temperature conditions, significantly reducing system energy consumption and addressing issues of low cooling efficiency and resource waste.

[0018] 3. By introducing a stability function-driven control strategy generation mechanism, this invention eliminates the need for system regulation to rely solely on temperature threshold triggers. Instead, it performs global judgment and control based on the dynamic stability of the entire thermal-fluid field. Compared to existing approaches that simply reduce the temperature based on temperature anomalies, this invention offers greater foresight and proactive intervention capabilities, resolving the issues of delayed response and unclear control direction in existing systems.

[0019] 4. This invention achieves coordinated and linked regulation between damper actuators and refrigeration equipment by implementing a control signal mapping table and priority scheduling logic within the execution linkage module. Compared to existing control methods where each device operates independently and lacks a coordination mechanism, this invention improves overall system regulation efficiency and resolves issues such as inconsistent linkage responses and low execution efficiency between devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flow chart of the method of the present invention; Figure 2 This is a diagram of the device architecture of the present invention; Figure 3 It is a schematic diagram of the terminal device structure of the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0022] See also Figure 1 The embodiment of the present invention provides a method for managing a closed cold aisle in a computer room, comprising the following steps: S1. Deploy environmental monitoring nodes at key locations within the cold aisle area to collect real-time air temperature and air velocity at key locations, and fuse the environmental parameters of each key location to construct a cold aisle state vector. In the process of implementing the closed cold aisle management method of a computer room of the present invention, in order to accurately grasp the typical airflow paths and heat distribution conditions in the cold aisle, thereby providing basic support for subsequent modeling and feedback control, it is necessary to preferentially deploy environmental monitoring nodes in key locations within the cold aisle area. This step S1 not only constitutes the perception input layer of this system, but also serves as the data basis for the subsequent state space model construction S2, stability function calculation S3, and controller execution S4. The quality of data collection will directly affect the convergence efficiency and dynamic response capability of the entire system adjustment strategy.

[0023] In this embodiment, monitoring nodes are deployed at the following key locations within the cold aisle to build a data perception network with high temporal and spatial resolution: Cold aisle air inlet: Usually the initial section of the air supply path, where the air temperature is the lowest and the flow rate is faster; The front side of each server rack: the inlet side before the cold air enters the equipment, used to monitor the effectiveness of the air supply; The rear side of each server rack is the equipment exhaust vent, with the highest temperature and significant airflow disturbance. Cold channel outlet: connected to the hot air return channel, used to determine the local return air short circuit and airflow distribution balance.

[0024] Generally, at least one monitoring node is deployed in each of the above areas. As an option, for areas with a rack density greater than 20 units, the deployment density can be increased by configuring one node for every two racks to strengthen the distribution.

[0025] In one possible implementation, the environmental monitoring node includes the following core modules: Temperature sensor: Use digital or analog type sensor to measure the ambient temperature value T in degrees Celsius (℃). The typical measurement range is -10℃ to 85℃ with a resolution of 0.1℃. It is used to monitor heat flow accumulation. Wind speed sensor: measures the local airflow linear velocity V in meters per second (m / s), with a typical measurement range of 0 to 10 m / s. It is used to determine whether the air volume effectively penetrates the rack area. Humidity sensor: used to collect relative humidity H in percentage (%RH), with a typical range of 0-100%. Data acquisition module: completes the synchronous collection, time-series packaging, encoding, and uploading of temperature, wind speed, and humidity measurement data. Its communication method can be wired (RS485, CAN) or wireless (Wi-Fi, ZigBee, LoRa, etc.) communication protocol. In some embodiments, the data acquisition module integrates a real-time clock (RTC) and a cache chip to support breakpoint resumption and time series alignment to avoid sampling misalignment.

[0026] The monitoring node data will be used as the input variable of the system state space model. In each sampling period, the system constructs a set of state vectors. Specifically: x(t)=[T1(t),V1(t),H1(t),T2(t),V2(t),H2(t),…,T n (t),V n (t),H n (t)] T ; Where x(t) represents the state vector of the system at time t, with a dimension of 3n×1; T i (t) represents the temperature value collected by the i-th node, in °C; V i (t) represents the wind speed value collected by the i-th node, in m / s; H i (t) represents the relative humidity value collected by the i-th node, in %RH; n is the number of monitoring nodes; T represents vector transposition, and the result is a column vector; In an actual system, if the humidity change has little impact on the modeling, the humidity dimension can be ignored. However, in the complete solution of the present invention, in order to maintain the uniformity of the input vector structure, it is recommended to retain all dimensions.

[0027] Before data is passed into the modeling system, it is necessary to normalize each dimension to improve the numerical stability of the model. The normalization process is as follows: in: is the normalized i-th state variable; x i (t) is the original state variable; μ i is the sample mean value of the i-th variable; σ i is the sample standard deviation of the i-th variable; This normalization method (Z-score) is robust and suitable for dynamic systems with non-mean stability.

[0028] Take a typical medium-sized computer room as an example, which contains four rows of cold aisles, each row of which has 12 server racks. According to the recommended strategy of this embodiment, each row should be equipped with at least: 1 air inlet node; 12 front-of-rack nodes; 12 rack-mounted nodes; 1 exit node.

[0029] Each cold aisle column has 26 nodes, and the four columns have a total of 104 nodes. The system can automatically construct the state vector position mapping matrix based on the node number, which facilitates the subsequent calculation of the state space matrices A and B.

[0030] In a possible expansion scenario, the environmental monitoring node can also introduce infrared thermal imaging or TOF-based spatial ranging functions to assist in determining airflow turbulence areas and reconstructing streamline fields, thereby supporting high-order state estimation processes.

[0031] S2. Inputting the state vector into a thermal-fluid coupling state space modeling module to establish a state space model of the temperature and airflow velocity inside the cold aisle over time, wherein the model is used to dynamically describe the state evolution process of the system under multi-source disturbance conditions; After collecting real-time temperature and air velocity data from key cold aisle areas and constructing state vectors, the state vector x(t) constructed in step S1 is input into the thermal-fluid coupling state modeling module to further describe the dynamic evolution of the cold aisle's internal environmental state over time. This module's core function is to establish a mathematical model describing the system's state evolution based on state-space theory, thereby supporting subsequent stability assessment and feedback control calculations. The model should effectively reflect the temperature-airflow coupling effect and the relationship between the control variable and disturbance in the system's evolution.

[0032] In this embodiment, a discrete time state space model is used to model the cold channel system. The specific model form is as follows: x(t+1)=Ax(t)+Bu(t)+w(t); Where: x(t) represents the state vector of the system at time t, with dimension n x ×1, where n x =3n, n is the number of monitoring nodes, each node corresponds to the temperature T i , wind speed V i 、Humidity H i Three components; x(t+1) is the predicted state vector of the system at time t+1, which has the same meaning as x(t); u(t) is the control instruction input vector, with dimension n u ×1, including the control quantity of each execution module (such as damper opening, cooling unit power, etc.) at time t; w(t) is the external disturbance vector with dimension n x ×1, used to describe factors that cannot be directly controlled, such as random load changes and nonlinear heat source fluctuations; A is the system state transfer matrix, with a dimension of n x ×n x , describes the impact of the current state on the next state; B is the control input influence matrix, dimension is n x ×n u , describing the regulatory effect of each control input on different state dimensions.

[0033] In one possible implementation, the construction of matrix A needs to consider the spatial adjacency and physical interactions between state variables and should have a sparse structure to improve computational efficiency. For example: The temperature of a node at the next moment is mainly affected by its current state and the wind speed and temperature of surrounding nodes; Air flow velocity is affected by upstream damper adjustments and adjacent heat sources.

[0034] Therefore, the non-zero elements of matrix A are only distributed in the diagonal and its adjacent strip areas, forming a strip-shaped sparse matrix structure.

[0035] In general, matrix B is also sparse. Its nonzero elements are located at the mapping locations between specific states and their controlled inputs. For example, damper adjustment only affects the wind speed component of a specific regional node, while cooling power adjustment primarily affects the temperature dimension of the corresponding region.

[0036] In some embodiments, the matrix A may be identified in the following data-driven manner: Where: T represents the length of the training sequence; λ represents the sparse constraint strength coefficient; ||A||1 represents the L1 norm of the matrix elements, which is used to control overfitting and enhance interpretability.

[0037] In this embodiment, the disturbance vector w(t) can be modeled as a Gaussian process with time correlation to reflect the persistence and regional coupling of uncontrollable fluctuations in the real system. It is defined as follows: Where: μ w is the perturbation mean vector, usually zero vector; Σ w is the perturbation covariance matrix, dimension is n x ×n e , whose non-zero structure is similar to that of matrix A, to characterize the propagation effect of disturbances between adjacent state variables.

[0038] Specifically, in some instances, the covariance matrix Σ w It can be set as a block diagonal structure, with each block corresponding to a different cold channel sub-area; or it can be set as an exponential decay spatial correlation model: Where: 2 is the disturbance intensity; d i,j is the physical distance between node i and node j; θ is the correlation attenuation factor, which controls the spatial propagation rate of disturbance.

[0039] In order to improve the generalization ability of the model in dynamic scenarios, the system supports the function of updating model parameters online. Taking the state matrix A as an example, the update method is as follows: Where: η is the update step size (learning rate), which is generally set at 10 -3 to 10 -2 within the scope; Represents the current predicted value; the update strategy is fine-tuned based on the back propagation of the prediction error.

[0040] In a specific implementation, the above update process is triggered once every several sampling periods to avoid model oscillation.

[0041] In the modeling process, in order to ensure that the predicted state variables meet physical rationality, it is also necessary to set boundary values ​​for each component in x(t). For example: Temperature: T i (t)∈[15,45](unit: °C); Wind speed: V i (t)∈[0.2,7.5](unit: m / s); Humidity: H i(t)∈[20,90](unit: %RH); If the prediction result is out of bounds, it can be corrected by truncation or reset strategy to avoid abnormal control instruction generation.

[0042] S3. Constructing a stability function for evaluating the cold channel thermal-flow field stability based on the state-space model, wherein the stability function is represented by a quadratic function consisting of a state vector and a symmetric positive definite matrix, and is used to monitor in real time the degree of stability deviation of the system under the influence of the current disturbance; After completing the thermal-fluid coupled state-space modeling, the system requires a mechanism to monitor in real time whether the current state deviates from the thermodynamic steady state, thereby providing a basis for the controller's decision-making. Therefore, in step S3, a mathematical function, the "stability function," is constructed based on the state vector to measure the degree to which the system as a whole deviates from the steady-state center.

[0043] The core idea of ​​the stability function is to perform a weighted evaluation of the current system state x(t) and extract the global characteristics of the state change energy, thereby compressing the high-dimensional multi-source state into a scalar indicator and simplifying monitoring and judgment.

[0044] In this embodiment, the stability function is constructed using the following standard quadratic function: V(x(t))=x(t) T Px(t); Where: V(x(t)) represents the system stability function value at time t, which is a dimensionless scalar and is used to measure the degree to which the system currently deviates from the steady state; x(t) represents the state vector of the system at time t, with a dimension of n x ×1, where n x =3n, n is the total number of deployed monitoring nodes; x(t) T is the transposed vector of x(t), with dimension 1×n x ; P is the weight matrix in the stability function, dimension is n x ×n x , satisfying P = P T , and P is a symmetric positive definite matrix, which is used to assign corresponding weights to different state variables while ensuring the non-negativity and monotonicity of the function value.

[0045] The positive definiteness of a symmetric positive definite matrix P satisfies the following conditions: This condition ensures that any non-zero state offset will cause a positive increase in V(x), thus having the properties of an "energy function" in the physical sense.

[0046] In one possible implementation, the matrix P can be constructed in one of the following three ways: Method 1: Manually set diagonal weights and set off-diagonal elements to 0, reflecting only the independence of each state dimension. Method 2: Construct a symmetric banded sparse matrix based on the spatial adjacency relationship between state variables; Method 3: Based on the system closed-loop performance indicators, obtain from the following optimization problem: Where: T is the length of the sampling sequence; λ is the regularization parameter that controls the sparsity of the weights; ||P|| F Represents the Frobenius norm, which is used to constrain the matrix size; Indicates that P is a positive definite matrix.

[0047] In general, the diagonal elements of the matrix P are used to regulate the relative importance of different state dimensions (temperature, wind speed, humidity); the non-diagonal elements can be used to reflect the state linkage relationship between nodes.

[0048] In some embodiments, the state variables can be divided into multiple regions, such as the inlet air region, the server front region, the server rear region, the return air region, etc., and a sub-matrix P is defined for each region. k , the construction structure is as follows: Where: K is the total number of regions; x k is the state subvector of the kth region; P k is the local positive definite weight matrix of the corresponding region; R ij is the coupling impact weight matrix between different regions.

[0049] Such partitioned stability functions can enhance the system's response to "local anomalies" or "thermal islanding" phenomena.

[0050] In each sampling period, the system calculates the stability function V(x(t)) in real time based on the collected state vector x(t) and compares it with the reference steady-state function value V ref For comparison, the following deviation is defined: ΔV(t)=V(x(t))-V ref ; Where: V ref is the reference stable state function value, which can be the historical average value during the system training period or obtained based on steady-state simulation calculation; ΔV(t) is the stability offset value of the current state, reflecting the degree of deviation of the current state compared to the steady-state center.

[0051] If ΔV(t)>ε, the system will be judged as "deviating from steady state", where ε is an artificially set margin threshold.

[0052] In some embodiments, the system uses a sliding window approach to perform multi-cycle detection to avoid misjudgment of transient disturbances. The following sliding average is defined: Where: N is the sliding window length, usually set to 3-10; is the average stability function value of the last N cycles; like Continuously higher than V ref +ε, the controller is triggered to start executing the compensation instruction.

[0053] In order to further enhance the response capability to strong disturbances, in this embodiment, a disturbance margin optimization function is introduced, which is defined as follows: Where: M w (t) is the disturbance response margin index; ||w(t)||2 is the L2 norm of the current disturbance vector, indicating the magnitude of the disturbance energy; if M w (t) increases significantly, indicating that the system is extremely sensitive to disturbances, and there may be model mismatch or insufficient adjustment. The modeling parameters or weight matrix should be adjusted in a timely manner.

[0054] S4. Based on the evolution of the stability function, a state feedback controller is designed. The current state vector is used to calculate the control instructions for the cold aisle execution structure. These instructions include the damper angle control signal and the cooling unit power control signal. After completing system state modeling S2 and stability function construction S3, to ensure that the cold aisle environment can maintain a stable operating state under the combined effects of dynamic loads and external disturbances, it is necessary to design a controller structure with state feedback capabilities to achieve an effective closed-loop response from system state to execution instructions. The controller's task is to output a set of control inputs u(t) based on the current state vector x(t) and the system state space model. These control inputs are used to adjust the damper angle and cooling unit power, thereby guiding the system toward a low-stability deviation state.

[0055] In this example, the controller design is based on a strategy to minimize the future evolution trend of the stability function. Specifically, the control input u(t) is solved to minimize the stability function V(x(t+1)) at the next moment, using the system state model as a constraint. The optimization objective function is defined as follows: Combined with the aforementioned state evolution model: x(t+1)=Ax(t)+Bu(t); Substituting in the optimization objective expansion, we can get: Expanding it further: Where: x(t) is the current system state vector, dimension is n x ×1; u(t) is the control input vector, dimension is n u ×1; P is the symmetric positive definite matrix in the stability function, dimension n x ×n x ; A is the system state transfer matrix, dimension n x ×n x ; B is the control input influence matrix, dimension n x ×n u ; P is a positive definite weight matrix (or Lyapunov matrix), dimension n×n, used to measure the cost of state deviation; A T is the transpose of matrix A; B T is the transpose of matrix B; x(t) T is the transpose of vector x(t), dimension 1×n; u(t) T is the transpose of vector u(t), dimension 1×m.

[0056] This optimization problem is a standard convex quadratic programming problem. The objective function is a quadratic function with respect to u(t) and is easy to solve.

[0057] Find the gradient of the above objective function with respect to u(t) and set the derivative to zero, we have: The optimal control law is solved as: u * (t)=-(B T PB) -1 B T PAx(t); Where: u * (t) is the optimal control input solution at the current moment; (B T PB) -1 is the matrix B T The inverse of PB must be a positive definite matrix (i.e. B T PB>0) to ensure the existence of the control solution; The overall control law can be expressed as a linear feedback structure, namely: u(t)=Kx(t), where K=-(B T PB) -1 B T PA; Matrix K is the feedback gain matrix, dimension is n u ×n x , directly mapping the current state to the control instructions.

[0058] In the present invention, the control input u(t) includes the following two types of physical signals: Air damper adjustment signal: used to control the electric air damper actuator at each key air duct position. The signal form is PWM duty cycle command or 0-10V analog quantity. The typical control frequency is 1Hz, and the adjustment angle range is 0°-90°. Refrigeration power control signal: used to adjust the air conditioner compressor frequency, fan speed or chilled water valve opening. The signal can be transmitted via RS485, Modbus and other protocols, with a response time of less than 2s.

[0059] In some embodiments, the above two control variables can be further divided into multiple sub-channels to achieve regional independent adjustment.

[0060] In one possible implementation, in order to enhance the system's tolerance to external disturbances w(t), the disturbance term can be directly introduced into the controller optimization structure to construct a robust control strategy. The system state model is expanded to: x(t+1)=Ax(t)+Bu(t)+w(t); The control objective becomes: That is, the expected stability function value under the influence of disturbance is minimized, where the disturbance satisfies This method introduces the disturbance covariance matrix Σ under the premise that the control solution structure remains unchanged. w , used to adjust the amplitude of the control gain and enhance the adaptability to thermal burst interference.

[0061] To ensure that the system operates within the physical constraints of the actuator, the control output must also meet the following constraints: min ≤u(t)≤u max ; Where: u min is the lower limit of the control input, including the minimum opening of the damper and the minimum cooling power; u max To control the input upper limit, it is limited according to the physical specifications of the equipment; If the optimal solution u * (t) exceeds the above range and can be corrected to the nearest feasible value by projection. The specific method is: clip (t)=min(max(u * (t),u min ),u max ); Among them, u clip (t) is the actual control command value after limiting sent to the actuator, which has been constrained within the allowable range.

[0062] In addition, in some special cases (such as multiple devices sharing control bandwidth), soft constraints or penalty functions can be added to expand the problem into a constrained optimization problem.

[0063] S5. Synchronously output the control command to the damper actuator and cooling equipment in the cold aisle, jointly adjust the air flow rate and temperature distribution inside the cold aisle, and update the state vector to enter the next cycle, so as to form a closed-loop control process for continuous monitoring and adjustment; Based on the state evaluation and control command calculation process completed by the aforementioned state modeling (S2), stability function construction (S3), and feedback controller design (S4), the control commands generated by the controller are applied to the actual execution object in step S5 to achieve physical regulation of the heat and flow state of the cold aisle. To establish a continuous, highly responsive, and low-latency control closed loop, the system state must be re-collected after execution to complete the state vector update and enter the next control cycle, thus forming a complete closed-loop control process.

[0064] In this embodiment, the control input vector u(t) output by the controller includes two types: damper execution signal and cooling control signal. Its structure is defined as: Where: u(t) is the total control input vector, dimension is n u ×1;u f (t) is the damper angle adjustment signal vector, with dimension n f ×1, corresponding to each guide plate; u c (t) is the cooling power adjustment signal vector, with dimension n c ×1, corresponding to each cooling module; n u =n f +n c represents the sum of the control input dimensions.

[0065] Generally, the control signal is sent to the execution port number through serial communication (such as RS485), industrial Ethernet or wireless protocol; the instruction structure includes the device address, adjustment target value, execution limit parameters, receipt confirmation bit and other contents.

[0066] In one possible implementation, the controller maintains a set of executor mapping tables: Where: addr i is the physical address of the i-th executor; type i ∈{throttle,cooling} represents the actuator type; Indicates its control range.

[0067] This mapping structure is used to quickly route and distribute control signals, improving the real-time performance of the transmission.

[0068] The damper control signal u of the present invention f(t) The guide plate structure acts on the top or bottom of the cold channel of the computer room, and its function is to adjust the air flow direction and local air volume distribution.

[0069] Specifically, in this embodiment: When the temperature T at the monitoring node i i (t) Exceeding the preset cooling target When , the controller determines that the location is a local thermal anomaly; The controller will identify the air deflector j above the corresponding area and adjust the air door angle to: Where: θ j (t) is the opening angle of the jth guide plate at time t; δ θ To control the incremental step size; sign(·) represents the sign function, which is used to determine the adjustment direction.

[0070] In order to avoid disturbance transfer caused by single point regulation, the system supports local temperature gradient vector Realize coordinated adjustment of air doors: The controller constructs a guide path based on the local high temperature gradient area and links multiple sets of air door angles to form a directional cooling air corridor.

[0071] The cooling power command u output by the controller c (t) can be further decomposed into the following three control channels: Where: f c (t) is the compressor frequency adjustment instruction, unit Hz; q c (t) is the cooling water flow rate, unit is L / min; v c (t) is the fan supply air speed adjustment value, unit is m / s.

[0072] As an option, to improve the response feedforward, the system uses a predicted temperature series Adjust the cooling power curve in advance: Among them, P i (t) represents the control priority cycle index value of the i-th monitoring area at the current time t; Indicates the variable k corresponding to the maximum value, that is, the k value at which the objective function achieves the maximum value; represents the air temperature value of the i-th area predicted at time t+k; N is the length of the prediction time window (number of prediction steps).

[0073] And set the target value for the current cooling power as: in: is the target output power of the i-th refrigeration unit at time t; η is the refrigeration adjustment coefficient; k * is the period corresponding to the maximum predicted temperature rise; P base Provides static power maintenance for the device.

[0074] The power of multiple cooling modules will be sorted by the size of regional temperature offsets, and graded targets will be applied to achieve both energy saving and improved cooling capacity in key areas.

[0075] After the execution is completed, the system will immediately re-collect the temperature, wind speed and humidity data of all key locations and construct a new state vector: x(t+1)=[T1(t+1),V1(t+1),H1(t+1),…,T n (t+1),V n (t+1),H n (t+1)] T ; Among them, x(t+1) represents the state vector of the cold channel system at time t+1, with a dimension of 3n×1, where n represents the total number of monitoring nodes; T i (t+1) represents the air temperature at the i-th monitoring location at time t+1, in degrees Celsius; V i (t+1) represents the airflow velocity at the i-th monitoring position at time t+1, in meters per second (m / s); H i (t+1) represents the relative humidity of the i-th monitoring position at time t+1, in percentage (%); i∈{1,2,…,n} represents the monitoring position number, which is usually the physical measuring point or virtual measuring point index arranged in the cold channel.

[0076] To avoid state noise amplification caused by execution delays or data residuals, the system introduces a fusion state update mechanism, which is defined as follows: in: It is the fused state input for the next cycle calculation; Indicates the system prediction state; α is the weighting coefficient between real-time observation and prediction, with a typical value of 0.6-0.9; represents the predicted state vector.

[0077] This fusion approach can take into account both the practicality of execution feedback and the consistency of the modeling structure, improving the system's tolerance to sudden errors and data delays.

[0078] The device for enclosing a cold aisle in a computer room described below and the method for enclosing a cold aisle in a computer room described above can refer to each other.

[0079] See also Figure 2 , a closed management device for cold aisles in a computer room, comprising: The environmental monitoring module is used to deploy monitoring nodes at multiple key locations within the cold aisle to collect real-time air temperature, air velocity, and humidity information at each location. The collected data is standardized and structured through a unified data acquisition mechanism. This in turn constructs a state vector covering the entire cold aisle area, providing a comprehensive data foundation for subsequent state modeling and control.

[0080] The state modeling module is used to receive the state vector generated by the environmental monitoring module and establish a state space model of the dynamic evolution of the heat-flow field based on this. This model can describe the coupling laws of heat propagation and airflow between each monitoring location. It supports the modeling expression of the system response trend and provides a basis for stability assessment and control strategy reasoning.

[0081] The stability assessment module is used to calculate the stability index of the current cold channel heat-flow system in real time based on the current system state output by the state modeling module; by comparing the state change trend with the set steady-state benchmark, it is used to determine whether the system has deviation, fluctuation or local instability; the assessment result is used as the judgment signal for triggering the control strategy and is fed back to the control module for control signal generation.

[0082] The state feedback control module is used to generate a corresponding control strategy based on the stability index and current state vector output by the stability assessment module. The control strategy includes angle adjustment signals for multiple dampers and power output adjustment signals for multiple refrigeration devices. It has linkage logic and priority mechanisms to support adaptive adjustment of local thermal anomalies, changes in overall airflow distribution, and other situations.

[0083] The execution linkage module is used to output the damper and refrigeration control signals generated by the control module to each actuator respectively; the actuator includes an electric guide plate, a compressor, a chilled water pump, an air supply fan, etc.; by executing the adjustment instructions, the air volume distribution and the refrigeration intensity in the local area of ​​the cold channel are changed to achieve the linkage optimization control of the air flow rate and the temperature field; at the same time, the status update and data feedback are completed to form a closed-loop adjustment mechanism.

[0084] The system of this embodiment can be used to execute the above method embodiments, and its principles and technical effects are similar, so they will not be repeated here.

[0085] Please see the attached Figure 3The present invention also provides a terminal device, comprising: a processor and a memory, wherein the memory stores a computer program executable by the processor, and when the computer program is executed by the processor, the above method is performed.

[0086] The present invention also provides a storage medium on which a computer program is stored. When the computer program is run by a processor, the above method is executed.

[0087] Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0088] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for managing the closed cold aisle of a computer room, characterized in that: The following steps are involved: S1. Deploy environmental monitoring nodes at key locations within the cold aisle area to collect real-time air temperature and air velocity at key locations, and fuse the environmental parameters of each key location to construct a cold aisle state vector. S2. Inputting the state vector into a thermal-fluid coupling state space modeling module to establish a state space model of the temperature and airflow velocity inside the cold aisle over time, wherein the model is used to dynamically describe the state evolution process of the system under multi-source disturbance conditions; S3. Constructing a stability function for evaluating the cold channel thermal-flow field stability based on the state-space model, wherein the stability function is represented by a quadratic function consisting of a state vector and a symmetric positive definite matrix, and is used to monitor in real time the degree of stability deviation of the system under the influence of the current disturbance; S4. Based on the evolution trend of the stability function, a state feedback controller is designed to calculate the control instructions of the cold aisle execution structure using the current state vector. The control instructions include the damper angle control signal and the refrigeration unit power control signal. S5. Synchronously output the control command to the damper actuator and cooling equipment in the cold channel, jointly adjust the air flow rate and temperature distribution inside the cold channel, and update the state vector to enter the next cycle, so as to form a closed-loop control process for continuous monitoring and adjustment.

2. A method for managing a closed cold aisle in a computer room according to claim 1, characterized in that: The key locations include the air passages at the cold aisle air inlet, the front of the server rack, the rear of the server rack, and the cold aisle outlet.

3. A method for managing a closed cold aisle in a computer room according to claim 1, characterized in that: The environmental monitoring node includes: Temperature sensors, used to detect air temperature at key locations; Wind speed sensor, used to detect air flow speed at key locations; Humidity sensor, used to monitor air humidity; The data acquisition module is used to integrate the measurement data of various sensors and upload them to the host computer.

4. A method for managing cold aisles in a computer room according to claim 1, characterized in that: The state space model is established in the following form: x(t+1)=Ax(t)+Bu(t)+w(t); Among them, x(t) represents the state vector, u(t) represents the control instruction vector, w(t) represents the disturbance term, and A and B are system matrices.

5. A method for managing a closed cold aisle in a computer room according to claim 1, characterized in that: The stability function is a quadratic function of the following form: V(x)=x T Px; Among them, x is the current state vector, x T is the transpose of x, and P is a symmetric positive definite matrix used to measure the system energy level and stability deviation.

6. A method for managing a closed cold aisle in a computer room according to claim 1, characterized in that: The state feedback controller constructs the following optimization objective by minimizing the system deviation based on the real-time change trend of the stability function: Where u(t) is the control input vector at time t, x(t+1) represents the system state vector at time t+1, P is a symmetric positive definite matrix, and x(t+1) T Px(t+1) represents the stability function, which is a quadratic function, and V(x(t+1)) represents the stability function value of the system at the next moment; And according to the system model x(t+1)=Ax(t)+Bu(t), the optimal control input u(t) is calculated, and the control input includes: The damper angle adjustment signal based on the disturbance response is used to correct the cold air flow path; Cooling power adjustment signal based on heat load fluctuation to combat temperature anomalies; The feedback controller is implemented using a linear quadratic regulator method or an improved algorithm thereof, and has the ability to adaptively adjust gain parameters to meet the control accuracy requirements under dynamic changes in different server loads.

7. A method for managing cold aisles in a computer room according to claim 1, characterized in that: The damper angle control signal drives the adjustable guide plate arranged at the top or bottom of the cold channel to adjust the spatial distribution of the airflow direction and flow rate by linking multiple air guide actuators, wherein: When the monitoring node detects that the temperature at the front end of a rack exceeds the set threshold, the controller will prioritize adjusting the opening angle of the air guide plate above that area to increase the supply of cold air to that area. The opening adjustment of multiple air guide plates is linked and optimized based on the changing trend of the local temperature gradient in the state vector, thereby forming a directional cooling path facing the hot spot area.

8. A method for managing a closed cold aisle in a computer room according to claim 1, characterized in that: The refrigeration unit power control signal is used to dynamically control the refrigeration capacity of the refrigeration module in the cooling system, and its control logic includes: Predicting temperature trends over multiple future cycles using the state-space model allows for pre-adjustment of cooling unit output power to avoid hysteresis effects. In non-uniform heat load scenarios, cooling priority levels are divided according to the distribution characteristics of the state vector, and a hierarchical control strategy is adopted to apply different target powers to each cooling module to achieve a balance between local enhanced cooling and global energy efficiency optimization; The control signal also includes a compound control variable for the compressor frequency, cooling water flow or air supply speed, which is used to improve the overall cooling response bandwidth.

9. A device for sealing cold aisles in a computer room, applied to a method for sealing cold aisles in a computer room according to any one of claims 1 to 8, characterized in that: include: Environmental monitoring module, used to collect air temperature, air velocity and humidity information at multiple key locations in the cold aisle and generate state vectors; A state modeling module is used to construct a state space model of the cold channel heat-flow field based on the state vector; A stability evaluation module is used to calculate the stability function and evaluate the stability deviation degree of the cold channel; A state feedback control module is used to calculate the damper angle and the refrigeration unit power control signal according to the stability function; The execution linkage module is used to output control signals to the damper actuator and cooling equipment to adjust the airflow and temperature distribution in the cold aisle.

10. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, it implements the method for closed cold aisle management in a computer room according to any one of claims 1 to 8.

Images