Multi-coroutine control method for precise air conditioning system of data center

Abstract

The present disclosure provides a multi-coroutine control method for a precision air conditioning system in a data center, which comprises constructing a continuously changing characteristic quantity reflecting the migration trend of cold and heat loads between partitions based on monitoring data of each partition in the data center; based on the continuously changing characteristic quantity, dynamically constructing and determining a mapping stable interval for each coroutine to determine whether to maintain the current mapping or allow reconstruction; if it is determined to maintain the current mapping, progressively adjusting the temperature and air volume set values of the air conditioner controlled by the corresponding coroutine; if it is determined to allow reconstruction, performing controlled coroutine mapping relationship adjustment with the goal of minimizing system disturbance; executing the control instructions obtained by the adjustment and verifying and closing the loop based on the system feedback state after the instruction execution, which can adapt to the load rapid migration scene, suppress control jitter under the premise of ensuring control responsiveness, and realize smooth and stable temperature control and air volume regulation.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of data center thermal management, in particular to a multi-coroutine control method for a data center precision air conditioning system. BACKGROUND

[0002] A large number of IT devices such as servers are deployed inside a data center, which continuously generate heat during operation and need to rely on a precision air conditioning system to maintain a suitable temperature environment. In order to improve control efficiency and response speed, a multi-coroutine control method is often used in the prior art, in which the data center space is divided into multiple sub-regions, and each coroutine is responsible for the control task of one or a group of precision air conditioners, thereby realizing parallel local regulation. However, in business scenarios such as cloud computing and artificial intelligence training, the thermal load inside the data center often presents the characteristics of rapid and large fluctuations, and the hot spot region may migrate within minutes or even seconds, resulting in continuous changes in cold and hot distribution. In such a scenario of frequent and rapid switching of cold and hot loads, the existing multi-coroutine control method relies on a fixed mapping relationship between the coroutine and the partition, and once the load hot spot migrates, the reconstruction and remapping of the coroutine are triggered. This process involves the redistribution of control tasks and the jump adjustment of air volume and temperature settings, which, at the system execution level, is manifested as control command oscillation caused by frequent switching of the mapping relationship, i.e., control jitter. Jitter not only causes temperature and air volume to be unstable, affecting refrigeration efficiency and equipment life, but also may cause local overheating risk due to air flow organization disorder.

[0003] It can be seen that the existing method only responds to the instantaneous load state, does not continuously distinguish the spatial migration trend of the load, and does not set a tolerance interval in the control logic to maintain a stable mapping, so it cannot distinguish between short-term fluctuations and sustained migration, resulting in frequent switching of the coroutine and control instability.

[0004] Therefore, there is an urgent need for a multi-coroutine control method for a data center precision air conditioning system that can adapt to the scenario of rapid migration of loads, so as to suppress control jitter and achieve smooth and stable temperature control and air volume regulation while ensuring control responsiveness. SUMMARY

[0005] Therefore, in order to solve the problems brought by the prior art, the present application provides a multi-coroutine control method for a data center precision air conditioning system.

[0006] In a first aspect, the present disclosure provides a multi-coroutine control method for a data center precision air conditioning system, the method comprising:

[0007] S1, based on the monitoring data of each partition of the data center, constructing a continuous change representation reflecting the migration trend of the cold and hot load between the partitions;

[0008] S2, dynamically constructing and determining a mapping stable interval for each coroutine based on the continuous change characteristic quantity, to determine whether to maintain the current mapping or allow reconstruction;

[0009] S3, if it is determined to maintain the current mapping, performing progressive collaborative adjustment on temperature and air volume setting values of the air conditioner controlled by the corresponding coroutine;

[0010] S4, if it is determined to allow reconstruction, performing controlled coroutine mapping relationship adjustment with the goal of minimizing system disturbance;

[0011] S5, executing the control instructions obtained by the above S3 or S4 adjustment, and verifying and closing loop based on the system feedback state after the instruction execution.

[0012] Optionally, the S1 comprises:

[0013] Based on the return air temperature, supply air temperature and heat load data of each partition of the data center, the instantaneous refrigeration pressure representing the comprehensive refrigeration demand of the partition is calculated;

[0014] The change rate of the instantaneous refrigeration pressure is calculated to obtain a partition migration intensity sequence reflecting the speed and direction of load change;

[0015] Based on the partition migration intensity sequence, statistical analysis is performed within a sliding time window to generate a continuous change characteristic quantity containing migration direction, average cumulative intensity and duration.

[0016] Optionally, the S2 comprises:

[0017] The continuous change characteristic quantity of each partition under the jurisdiction of the coroutine is summarized, and a coroutine mapping offset quantity reflecting the degree of impact on the mapping relationship is calculated;

[0018] The upper boundary of the mapping stable interval at the current time is dynamically determined according to the historical sequence of the coroutine mapping offset quantity;

[0019] The reconstruction gating quantity is calculated by comprehensively considering the intensity and duration of the real-time offset quantity exceeding the upper boundary;

[0020] Whether to maintain the current mapping or allow reconstruction is determined by comparing the reconstruction gating quantity with a preset threshold.

[0021] Optionally, the determination of whether to maintain the current mapping or allow reconstruction by comparing the reconstruction gating quantity with a preset threshold comprises:

[0022] Only when the reconstruction gating quantity continuously exceeds the preset threshold for a preset confirmation duration, it is determined to allow reconstruction; otherwise, it is determined to maintain the current mapping.

[0023] Optionally, the S3 comprises:

[0024] calculating a coroutine-level adjustment target direction quantity based on the instantaneous refrigeration pressure of the partition under the jurisdiction of the coroutine;

[0025] calculating a temperature set value progressive correction quantity according to the deviation between the adjustment target direction quantity and the stable reference pressure, and applying the temperature set value progressive correction quantity to the current temperature set value to obtain an updated temperature set value;

[0026] calculating a fan speed progressive correction quantity based on the deviation between the average return air temperature of the partition under the jurisdiction of the coroutine and the updated temperature set value, and applying the fan speed progressive correction quantity to the current fan speed set value;

[0027] performing consistency constraint on the temperature set value progressive correction quantity and the fan speed progressive correction quantity to ensure adjustment direction coordination;

[0028] outputting the updated temperature set value and fan speed set value as corresponding control instructions.

[0029] Optionally, the S4 comprises:

[0030] For the coroutine triggering reconstruction, a migration priority is calculated based on the activity level of the partition migration, and a candidate migration partition is screened out;

[0031] An optimal target coroutine is determined for the candidate migration partition based on the principle of minimizing the misfit cost of the target coroutine, and the mapping relationship adjustment is completed;

[0032] A control quantity is calculated based on the adjusted mapping relationship, and a new control quantity is generated by weighted fusion of the newly calculated control quantity and the baseline control quantity frozen before reconstruction through a smoothing fusion algorithm.

[0033] Optionally, the calculation of the migration priority comprises:

[0034] The migration priority is calculated according to the average cumulative intensity, the instantaneous migration intensity and the duration of the partition, wherein the greater the average cumulative intensity and the instantaneous migration intensity and the shorter the duration, the higher the migration priority.

[0035] Optionally, the S5 comprises:

[0036] The control instructions are sent to each precise air conditioning actuator;

[0037] The return air temperature, supply air temperature and thermal load feedback data of each partition after the execution of the instructions are collected;

[0038] The control effect is verified based on the feedback data, and the feedback data is used as the input of the continuous change representation quantity in the next control cycle to form a closed-loop control.

[0039] In a second aspect, the present disclosure provides an electronic device, comprising a memory and at least one processor, wherein the memory stores a computer program, and the processor is configured to execute the computer program to implement the method of the first aspect.

[0040] In a third aspect, the present disclosure provides a computer storage medium, which stores a computer program, and the computer program is executed to implement the method of the first aspect.

[0041] The present disclosure has the following advantages compared with the prior art:

[0042] 1) By constructing a continuous change characterization of the cold and hot load space distribution, the basis for control decision is improved from passive response to instantaneous load state to active perception and quantification of load migration trend, effectively overcoming the misjudgment problem caused by the inability to distinguish short-term fluctuations from sustained migration in the prior art, providing a stable and forward-looking judgment basis for the entire control system, and reducing unnecessary control actions triggered by noise or instantaneous spikes from the source.

[0043] 2) A coroutine mapping maintenance judgment mechanism based on dynamic stable interval and reconstruction gating quantity is proposed. By calculating the adaptive boundary of each coroutine that can tolerate load fluctuations in real time, and introducing a gating quantity that integrates instability strength and duration for judgment, the system can intelligently distinguish between normal fluctuations and threatening migration. It solves the key defect of the prior art that fixed mapping relationship inevitably causes frequent coroutine switching and control command oscillation when the hot spot migrates, and realizes the stability and decisiveness of the control logic.

[0044] 3) A dual-mode smooth execution strategy under mapping stability and failure is designed. In the stable interval, gradual control adjustment within the coroutine is adopted to realize the mild and continuous change of control quantity while maintaining the architecture unchanged, and directly generate smooth evolving control instructions; when the stable interval is confirmed to be invalid, the controlled minimal coroutine reconstruction is started, and based on the reconstruction result, the seamless transition control instruction is generated through the smooth fusion algorithm. This strategy ensures that the system can output continuous, stable and coordinated control instructions under any load change scenario, whether it is responding to short-term fluctuations or adapting to sustained migration, eliminating control jitter and achieving efficient and reliable temperature control and air volume regulation of the data center precision air conditioning system. BRIEF DESCRIPTION OF DRAWINGS

[0045] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the present disclosure and serve to explain the principles of the present disclosure together with the specification.

[0046] Figure 1A flow chart of a multi-coroutine control method of a data center precision air conditioning system is shown.

[0047] Figure 2 A flow chart of a coroutine mapping stability determination and dual-mode control is shown.

[0048] The explicit embodiments of the present disclosure have been shown by the above-described drawings, and will be described in more detail hereinafter. These drawings and textual descriptions are not intended to limit the scope of the concept of the present disclosure by any means, but to illustrate the concept of the present disclosure to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0049] The present disclosure will be further described below with reference to the drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present disclosure, and cannot be used to limit the protection scope of the present disclosure.

[0050] The components of the embodiments of the present application generally described and illustrated in the accompanying drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of the application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0051] Hereinafter, the terms "include", "have", and their conjugates used in the various embodiments of the present application are only intended to denote that specific features, numbers, steps, operations, elements, components, or combinations thereof are present, and should not be understood as excluding the presence or addition of one or more other features, numbers, steps, operations, elements, components, or combinations thereof in advance.

[0052] Unless otherwise defined, all terms used herein (including technical terms and scientific terms) have the same meaning as commonly understood by those skilled in the art to which the various embodiments of the present application belong. Terms such as those defined in generally used dictionaries will be interpreted as having the same meaning as the contextual meaning in the relevant technical field and will not be interpreted as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of the present application.

[0053] The existing method is frequently triggered when facing the frequent and rapid hot and cold load migration in business scenarios such as cloud computing and artificial intelligence training, and the system cannot distinguish between short-term fluctuations and continuous migration due to the lack of continuous perception and quantification of load space migration trends and the dependence on fixed coroutine-partition mapping relationship, frequent triggering of coroutine reconstruction and accompanying jump control adjustment, and then causing control instruction oscillation, that is, control jitter. The jitter not only makes it difficult to stabilize the temperature control and air volume, affecting the refrigeration efficiency and equipment life, but also may cause local overheating risk due to airflow organization disorder. In view of this, the embodiment of the present disclosure provides a multi-coroutine control method of a data center precision air conditioning system, which aims to distinguish fluctuations and migration from the decision source, realizes continuous smooth output of control instructions by constructing a dynamic stable interval and a dual-mode control strategy, and thereby suppresses jitter. The implementation process of the method will be described in detail through several embodiments combined with the drawings.

[0054] Figure 1 The multi-coroutine control method of the data center precision air conditioning system provided by the embodiment of the present disclosure is shown in the flow chart as Figure 1 , which can include the following steps:

[0055] S1: Based on the monitoring data of each partition of the data center, a continuous change representation quantity reflecting the migration trend of the hot and cold load between the partitions is constructed.

[0056] During the operation of the data center, in order to accurately capture and quantify the migration trend of the hot and cold load in space, and avoid the system jitter caused by the control decision made only according to the instantaneous load state, a set of representation quantities capable of continuously describing the characteristics of the change direction, intensity and duration of the load are constructed. This process converts the physical monitoring values into continuous indexes with clear trend significance through data acquisition, calculation and fusion of the system. The following sub-steps are implemented.

[0057] S1.1: The data center is divided into multiple independent partitions according to the service range of the precision air conditioner, and the return air temperature, supply air temperature and real-time heat load basic data of each partition are synchronously collected at a uniform sampling period.

[0058] Firstly, the data center plane is divided into several partitions according to the service range of the precision air conditioner, each partition is assigned a unique serial number, denoted as i. For the ith partition, three basic quantities are obtained in each sampling period: return air temperature , supply air temperature and partition heat load , and form discrete sequences under a uniform sampling period: return air temperature sequence , supply air temperature sequence and heat load sequence . The sampling period is 0.5 seconds to 10 seconds, preferably 1 second to 2 seconds, which is used to cover the rapid fluctuations of cloud computing and AI training tasks.

[0059] The original data collected is continuous. For ease of calculation and processing, it needs to be discretized. At each sampling time, the system records the return air temperature value, the supply air temperature value and the partition heat load value of each partition, and forms three discrete sequences in time sequence: return air temperature discrete sequence, supply air temperature discrete sequence and partition heat load discrete sequence. Among them, the return air temperature and the supply air temperature are directly measured by high-precision temperature sensors deployed on the corresponding partition return air channel and supply air channel.

[0060] The partition heat load is obtained by aggregating the real-time power consumption data of all IT cabinets in the partition, or estimated according to the heat exchange power on the air conditioner side. Through this step, a time-aligned and source-unified basic data sequence is established for all subsequent calculations.

[0061] S1.2: Based on the collected basic data, the instantaneous refrigeration pressure of each partition at each sampling time is calculated based on the temperature difference and load factor, and a refrigeration pressure sequence representing the comprehensive refrigeration demand of the partition is formed.

[0062] Based on the obtained partition basic data sequence, a single indicator that can comprehensively reflect the partition refrigeration demand, namely the instantaneous refrigeration pressure, needs to be further calculated. This pressure value integrates both temperature difference driving and load driving factors, and is used to represent the comprehensive demand tension of the partition for refrigeration capacity at a specific time.

[0063] For the i-th partition, the instantaneous refrigeration pressure at discrete time t is calculated as follows:

[0064]

[0065] wherein, is the instantaneous refrigeration pressure of the i-th partition at time t, which is a dimensionless number; is the return air temperature of partition i collected at time t, with the unit of Celsius degree; is the supply air temperature of partition i collected at time t, with the unit of Celsius degree; is the heat load of partition i at time t, with the unit of kilowatt. is a load constant for normalization, with a value range of 5 kilowatts to 50 kilowatts, which can be set according to the typical cabinet power consumption of the data center. a and b are the temperature difference weight coefficient and the load weight coefficient, respectively, used to adjust the contribution proportion of the temperature difference term and the load term in the total pressure, wherein the coefficient a has a unit of 1 / ℃, used to normalize the temperature difference dimension; the coefficient b is a dimensionless number, and the value range of both is 0.2 to 2.0, and the specific value can be determined through actual system calibration.

[0066] ​The above calculation is repeated for each sampling time of each partition, and the discrete sequence of refrigeration pressure of the partition {P(1), P(2),..., P(t),...} is obtained. The sequence maps the originally multi-dimensional, physically different basic monitoring quantity into a single pressure quantity with consistent physical meaning and evolving over time, laying a computational foundation for subsequent quantification of the direction and strength of load migration.

[0067] S1.3: According to the refrigeration pressure sequence, the change amount per unit time is calculated to obtain the partition migration strength sequence representing the speed and direction of load change.

[0068] The instantaneous refrigeration pressure reflects the static demand at a certain time, and to capture the migration behavior of the load, the key lies in its change trend. Therefore, the change rate of the refrigeration pressure sequence needs to be calculated based on the refrigeration pressure sequence, and the change rate is defined as the migration strength of the partition, which is used to describe the speed and direction of the cold and hot load in the time dimension.

[0069] The change rate of the refrigeration pressure of the i-th partition in the adjacent sampling interval is calculated as the migration strength of the partition. The migration strength is calculated as follows:

[0070] ,

[0071] wherein, is the migration strength of the i-th partition at time t, with the unit of per second; is the refrigeration pressure of the partition i at the current time; is the refrigeration pressure of the partition i at the last sampling time; is the uniform sampling period set in step S1.1, with the unit of seconds.

[0072] When the calculation result of is positive, it indicates that the refrigeration demand of the partition is increasing, i.e., the thermal pressure is in an upward trend; when the result is negative, it indicates that the refrigeration demand of the partition is decreasing, i.e., the thermal pressure is in a downward trend. By sequentially calculating at each sampling time, the discrete sequence of migration strength of each partition i is obtained .

[0073] Optionally, to enhance the robustness to instantaneous abnormal readings of the sensor, a simple reasonableness check and smoothing can be performed on the refrigeration pressure sequence before calculating the migration strength . For example, if it is detected that the jump amplitude of the current time relative to the last time exceeds the threshold value physically possible, such as a temperature jump exceeding 5°C, the pressure value of the last moment or the sliding average value of the pressure of the previous moments can be used as The alternative values participate in the calculation, thus avoiding the dramatic interference of single-point abnormal data on the migration intensity sequence. This optional data preprocessing link can improve the robustness of subsequent trend judgment.

[0074] This step converts the basis of control decision from instantaneous point value to continuous trend, reducing false judgment and system jitter caused by short-time load spikes or measurement noise from the source.

[0075] S1.4: Perform statistical analysis on the migration intensity sequence within the sliding time window, calculate and output the continuous change characterization quantity set containing migration direction, average cumulative intensity and duration.

[0076] Although the migration intensity sequence can reflect the change trend, it still contains random fluctuations. To robustly judge whether the load has undergone real and continuous migration, statistical analysis of the migration intensity in a longer time window is needed to build the final continuous change characterization quantity. This characterization quantity is a comprehensive index set containing migration direction, intensity and duration.

[0077] First, to smooth short-term fluctuations, calculate the average cumulative intensity S(t) of each partition within the sliding time window. This value reflects the average intensity level of load migration activity in the recent period of time, and the calculation formula is:

[0078] ,

[0079] Among them, is the average cumulative intensity of the i-th partition within the sliding window at time t, with the unit of per second; W is the length of the sliding window, i.e. the number of sampling points contained, with a value range of 3 to 60; is the migration intensity of the k-th sampling point of the partition within the sliding window (k=1 corresponds to the earliest time within the window, and k=W corresponds to the current time t); the summation symbol indicates the summation of the absolute values of the migration intensities of all W sampling points within the window.

[0080] Secondly, define the duration quantity to quantify how long the current migration trend has been continuously maintained. This quantity is calculated by counting the length of time when the absolute value of the migration intensity continuously exceeds a certain threshold:

[0081] ,

[0082] Among them, is the duration quantity of the i-th partition, with the unit of seconds; N is the number of sampling points of the partition that continuously satisfy ; is the set intensity threshold, with a value range of 0.01 to 1.0 per second; is the sampling period.

[0083] Finally, the migration direction is determined. This is based on the consistency of the sign of the migration intensity within the sliding time window. Specifically, the rule is: among all sampling points within the statistical window... The percentage of points is considered. If this percentage exceeds a preset threshold, such as 70%, the migration direction is determined to be positive, indicating that the load is converging towards that zone or that the thermal pressure is showing a continuous upward trend. If the percentage of points exceeds 70%, the migration direction is determined to be negative, indicating that the load is migrating out of that zone or that the thermal pressure is showing a continuous downward trend; if neither of these conditions is met, the migration direction is determined to be zero, indicating that there is currently no clear and consistent migration direction. This judgment is based solely on... The sign of is independent of the magnitude of its absolute value.

[0084] Based on the above calculations, the system ultimately outputs a set of continuously changing characteristic quantities for each partition, including migration direction and migration intensity. and duration This set of representations, from the perspective of trends rather than instantaneous values, clearly answers four key questions: whether the load is migrating, in which direction it is migrating, the strength of the migration, and how long this migration has lasted. This provides a direct and reliable quantitative criterion for establishing the stable interval of the coroutine mapping in subsequent steps.

[0085] In the technical solution of this disclosure, multi-dimensional instantaneous monitoring data such as return air temperature, supply air temperature, and zoned heat load are fused and calculated into a unified instantaneous cooling pressure. The rate of change is then calculated to obtain the migration intensity. Finally, within a sliding time window, continuous change parameters characterizing the migration direction, average cumulative intensity, and duration are statistically obtained. This step elevates the decision-making basis of the control system from discrete instantaneous point values ​​easily affected by noise to continuous quantitative indicators capable of depicting the spatial migration trend of the load. This fundamentally changes the perception dimension of the control system, enabling it to distinguish between short-term random fluctuations and continuous real load migration, laying a reliable data foundation for subsequent stable and jitter-free control decisions.

[0086] S2: Based on the continuously changing representation quantity, dynamically construct and determine the stable mapping interval for each coroutine to decide whether to maintain the current mapping or allow reconstruction.

[0087] After obtaining the continuous variation characterization quantity capable of describing the load space migration trend of the cold and hot loads, in order to cope with the control jitter problem caused by the rapid switching of the load hot spot, a corresponding mapping stable interval needs to be constructed for each established coroutine control task. The interval limits the coroutine in the space range it is responsible for, and can tolerate how much degree of load migration without changing its control mapping relationship. When the monitored load change is still within the stable interval, the system will determine that the current coroutine mapping relationship is still effective, and only allow fine tuning of the air volume, temperature and other control quantities within the original coroutine framework, thereby avoiding frequent and unnecessary coroutine reconstruction. The overall determination and control process is shown in Figure 2 The specific implementation is realized through the following sub-steps.

[0088] S2.1: solidify the mapping relationship of each coroutine control task and its current covered partition set, and aggregate the continuous variation characterization quantity of the governed partitions to the coroutine level.

[0089] The system first needs to determine all the running coroutine control tasks in the current data center, and bind them with specific physical space ranges. Each coroutine is usually responsible for controlling a group of precision air conditioners or a logical sub-region. The system reads the defined coroutine number set, and solidifies the partition set covered by each coroutine. The covered partition set is composed of a series of partition serial numbers, which clearly indicates the space range managed and responded by the coroutine at present.

[0090] Subsequently, the system aggregates the continuous variation characterization quantity calculated for each partition in step S1. For each coroutine, the system extracts the migration direction quantity, average cumulative intensity quantity and duration quantity from all the partitions in its covered partition set. After aggregating these characterization quantities, the basic index set at the coroutine level is formed. This set reflects the macroscopic situation of the load migration trend in the overall region governed by the coroutine. This step establishes a clear and stable analysis object for subsequent stability calculation, that is, the trend information originally scattered in hundreds of partitions is aggregated to dozens of coroutine dimensions according to the management responsibility, avoiding the uncertainty introduced by the analysis range drift in the calculation process.

[0091] S2.2: for each coroutine, statistics the proportion of partitions satisfying the strong migration condition in its covered partitions and the average duration, and calculate the coroutine mapping offset quantity quantifying the impact degree of the mapping relationship.

[0092] After determining the management range and internal trend of each coroutine, the next step is to quantify the impact degree of the load hot spot migration on the existing coroutine mapping relationship. For this purpose, a index called coroutine mapping offset quantity is defined. This index aims to comprehensively reflect two key information: one is the proportion of the partitions in the region governed by the coroutine that are undergoing significant migration; the other is how persistent these migration behaviors are.

[0093] The calculation process is as follows: for the state of any coroutine at a discrete time, the system traverses all partitions within its set of covered partitions. For each partition, it is determined whether it meets the strong migration condition. The strong migration condition is a double criterion that requires the average cumulative intensity measure of the partition to be no lower than the preset intensity threshold , and its duration amount to be no lower than the preset duration threshold , wherein The value range of is 0.02 to 0.8 per second, The value range of is 3 seconds to 120 seconds. These two thresholds are usually calibrated according to actual working conditions during the system trial run stage, or set to conservative values that are not easily triggered by short-term fluctuations by operation and maintenance experience.

[0094] The number of partitions that meet the strong migration condition is counted and denoted as . At the same time, the average value of the duration amounts of these partitions is calculated and denoted as . The coroutine mapping offset E is calculated by the following formula:

[0095] ,

[0096] wherein is the total number of partitions within the set of covered partitions of the coroutine, and its typical value range is 4 to 80; is a duration normalization constant, and its value range is 10 seconds to 300 seconds. The offset E is a dimensionless number, and the larger its value, the more severe the challenge of the current load migration to the existing mapping relationship of the coroutine. The system calculates this value for each coroutine at each sampling time, thereby forming a discrete sequence of coroutine mapping offset that changes over time. This step converts the abstract problem of whether load migration has affected the effectiveness of the coroutine into a quantifiable index that can be continuously monitored and calculated.

[0097] S2.3: Based on the historical sequence of the coroutine mapping offset, the recent average value and fluctuation amplitude are dynamically calculated, and then the upper boundary of the mapping stable interval of the coroutine at the current time is determined.

[0098] After obtaining the sequence of the coroutine mapping offset, the system does not simply set a fixed threshold for it, but dynamically constructs an adaptive mapping stable interval for each coroutine based on the historical fluctuation law of the offset. The advantage of this approach is that the stable interval can adjust to the overall level of the field load fluctuation, thereby more intelligently distinguishing between normal short-term fluctuations and real trend migration.

[0099] The stable interval is constructed in the form of a central value plus a tolerance band. First, the system calculates the average value of the offset within a sliding historical time window The mean value reflects the normal level of the recent offset. The calculation formula is:

[0100] ,

[0101] where W is the length of the sliding window, which can be 5 to 120 sampling points; E(k) is the offset of the kth sampling point in the sliding window.

[0102] Then, the average deviation amplitude V of the offset relative to its mean value in the same sliding window is calculated, which represents the normal fluctuation range of the offset:

[0103] ,

[0104] Based on the mean value and the fluctuation amplitude V, the current mapping stable upper boundary U of the coroutine can be defined:

[0105] ,

[0106] where c is the tolerance coefficient, which can be adaptively fine-tuned according to the historical fluctuation level of the system as a preferred implementation. For example, when the recent average deviation amplitude V is continuously at a low level, the value of c can be appropriately reduced to increase the sensitivity to migration; when V is continuously at a high level, the value of c can be appropriately increased to enhance the tolerance to normal fluctuations, so that the stable interval can better fit the actual running conditions of the system. When the real-time calculated coroutine mapping offset E does not exceed the stable upper boundary U, the system considers that the current load migration is still within the acceptable stable interval of the coroutine. The establishment of this dynamic boundary enables the system to effectively distinguish between short-term fluctuations around the normal level and abnormal migrations that continuously break through the historical fluctuation range.

[0107] S2.4: Calculate the reconstruction gating quantity of the comprehensive out-of-bound strength and duration by comparing the real-time offset with the stable interval upper boundary, and determine whether to allow coroutine reconstruction based on it.

[0108] After defining the dynamic stable interval, the system needs a set of fine judgment mechanism to decide when to maintain the mapping and when to allow reconstruction. The strategy of directly triggering reconstruction with instantaneous out-of-bound is too sensitive and easy to cause jitter. Therefore, the persistence gating logic is introduced to upgrade the judgment from one-time comparison to comprehensive consideration of the out-of-bound strength and out-of-bound duration.

[0109] First, define the out-of-bound strength indicator B to quantify the degree of the current offset exceeding the stable upper boundary:

[0110] ,

[0111] where The function ensures that only when E is greater than U, a positive value is taken, otherwise it is zero; exp is the natural exponential function; r is the scale factor, the value range is 0.05 to 0.5. When E does not exceed U, B is zero; when E exceeds U and exceeds more, B value is closer to 1, indicating that the risk of instability is higher.

[0112] Then, the system accumulates the out-of-bound strength indicator B in the entire continuous time period from the current time to the first time that does not satisfy the out-of-bound condition, to form a reconstructed gating quantity R:

[0113] ,

[0114] Wherein, M represents a set of all continuous out-of-bound sampling points from the first out-of-bound to the current time; B(m) is the out-of-bound strength indicator of the mth sampling point during continuous out-of-bound; is the sampling period; the summation range is all continuous out-of-bound sampling points from the first out-of-bound to the current time. The unit of the reconstructed gating quantity R is second, which comprehensively represents the intensity and duration of the instability risk.

[0115] Finally, the reconstructed gating quantity R is compared with a preset gating threshold . The value range of the gating threshold is 2 seconds to 180 seconds, preferably 10 seconds to 60 seconds. If R is less than , the system determines that the mapping remains unchanged, and the mapping relationship of the coroutine remains unchanged, and subsequent gradual control adjustment will be performed in step S3. If R is not less than , it is determined that the reconstruction is allowed, which means that the load migration has been continuous and strong enough to threaten the effectiveness of the current mapping relationship, and the system will trigger the subsequent controlled coroutine reconstruction process. This step fundamentally suppresses frequent reconstruction caused by load fluctuation around the stability boundary, and provides a key pre-decision condition for stable operation of the control system.

[0116] In the technical scheme of the embodiment of the present disclosure, the continuous change representation quantity generated in step S1 constructs a dynamic adaptive mapping stability interval for each coroutine, and defines a reconstructed gating quantity that comprehensively represents the out-of-bound degree and duration as the core of the decision. This step makes the system no longer respond to load changes immediately and rigidly, but can tolerate normal fluctuations within a certain range and time. Only when the load migration activity is continuous and strong, causing the reconstructed gating quantity to accumulate beyond the threshold, the system determines that the current mapping relationship is invalid. This mechanism effectively suppresses frequent coroutine reconstruction and instruction oscillation caused by load fluctuation around the critical state from the source of decision logic, and establishes a stable core of the control system.

[0117] S3: If it is determined to maintain the current mapping, the temperature and air volume set value of the air conditioner controlled by the corresponding coroutine is gradually adjusted in the corresponding coroutine.

[0118] When the system determines that a coroutine is still in its mapped stable interval, i.e., the load migration has not reached the threshold triggering reconstruction, the control quantity is adjusted smoothly and continuously in a progressive manner only according to the load change of the region covered by the coroutine, and control instructions that can be used for delivery are directly generated. In this way, the air volume and temperature setting of the air conditioning system can gently follow the evolution of the load migration, avoiding the generation of a large control instruction jump due to short-term fluctuations or initial migration of hot spots, thereby realizing stable and jitter-free control transition. The following sub-steps are used to achieve this.

[0119] S3.1: According to the reconstruction gating quantity, confirm the progressive adjustment enable state in the coroutine, and calculate the coroutine-level adjustment target direction quantity based on the refrigeration pressure of the partition under jurisdiction.

[0120] The system first needs to confirm whether it is currently allowed to enter the progressive adjustment phase. If it is still in the mapping maintenance phase, the system determines that progressive adjustment in the coroutine is allowed, and sets a progressive adjustment enable flag to true. If the load migration has continued to the extent that reconstruction is needed, the system sets the progressive adjustment enable flag to false at this time, and the subsequent sub-steps will be skipped. adjustment calculation, directly freeze the current control quantity, and prepare for possible mapping reconstruction.

[0121] For the coroutine allowed to perform progressive adjustment, a unified adjustment target direction needs to be calculated for it. From step S1, the refrigeration pressure of each partition in the coroutine coverage partition set is obtained . The coroutine-level adjustment target quantity is obtained by calculating the weighted mean of the refrigeration pressures of these partitions, and the calculation formula is as follows:

[0122] ,

[0123] wherein, is the coroutine-level adjustment target quantity, which is a dimensionless number representing the required refrigeration pressure adjustment direction of the coroutine as a whole; is the refrigeration pressure of the i-th partition in the coroutine coverage partition set at the current time ; is the weight value corresponding to the partition, which is in the range of 0.5 to 3.0. The setting of the weight value may be based on various strategies, for example, the higher the cabinet power consumption proportion of the partition, the larger the weight value, or the higher the frequency of hot spots in the history of the partition, the larger the weight value. If the power consumption proportion strategy is used, the weight value may be set as a linear amplification value of the ratio of the power consumption of the partition to the total power consumption of the region under the jurisdiction of the coroutine. The calculated As a continuously changing reference quantity, it provides clear and consistent guidance for the subsequent coordinated progressive adjustment of temperature and air volume.

[0124] S3.2: According to the deviation of the target direction quantity and the stable reference pressure of the coroutine level, calculate the temperature setting value progressive correction quantity after limiting processing, and update the temperature control instruction.

[0125] After the adjustment enable state and the coroutine level target are determined, the temperature setting value of the precision air conditioner is first progressively corrected. This is the most direct link in the control response. The system obtains the current temperature setting value of the precision air conditioner managed by the coroutine .

[0126] Then, the target deviation quantity of the temperature setting is calculated , which reflects the gap between the current coroutine overall refrigeration pressure and an ideal stable state:

[0127] ,

[0128] Among them, is the coroutine level adjustment target quantity calculated in the last step; is the stable reference pressure of the coroutine, which is a dimensionless constant determined according to the recent historical running state of the coroutine, and the value range is usually between 0.2 and 2.0. For example, The average refrigeration pressure value of the coroutine in the past one hour can be taken, which represents a balanced state when the system is stably running.

[0129] Based on the target deviation quantity , the system calculates the progressive correction quantity of the temperature setting in this period . In order to ensure the smoothness of the adjustment and prevent the adjustment from being too large at a time, the correction quantity needs to be limited:

[0130] ,

[0131] Among them, is the temperature correction coefficient, which determines the mapping ratio of the deviation quantity to the actual temperature change quantity, and the value range is 0.1 to 2.0, unit: ℃ / unit pressure; is the maximum temperature correction amplitude allowed in a single sampling period, and the value range is 0.05 to 0.8 degrees Celsius; clamp is a limiting function, which limits the calculation result to between .

[0132] Subsequently, the temperature setting value is updated:

[0133] ,

[0134] The new set value obtained Also need to be constrained in the safe operation range allowed by the precision air conditioning equipment, for example, between 18 degrees Celsius and 28 degrees Celsius, if it exceeds, it will be automatically truncated to the boundary value. If the progressive adjustment enable flag of the current coroutine is false, it will directly make , The temperature setting remains unchanged. This step converts the traditional step temperature adjustment into a small, limited amplitude progressive correction per cycle, making the temperature control instruction present a continuous and smooth change curve, effectively suppressing the system disturbance caused by frequent and large changes in the set value.

[0135] S3.3: According to the deviation of the coroutine average return air temperature and the new temperature set value, the fan speed progressive correction amount after amplitude limiting processing is calculated, and the fan control instruction is updated.

[0136] The adjustment of the temperature set value must be coordinated with the change of the fan air volume, in order to achieve efficient and stable refrigeration effect. After completing the temperature set update, the fan speed needs to be adjusted progressively accordingly, to ensure that the air volume change and the temperature change direction are coordinated, to avoid the situation that the temperature has been adjusted low but the air volume has not followed up in time, so as to prevent the local return air temperature from appearing short-term abnormality.

[0137] The system first calculates the return air temperature deviation of the coroutine level. To ensure that only the changes of the covered partitions are used within the coroutine, the average return air temperature of the partition is the weighted average of the return air temperature of each partition in the coroutine covered partition set at the current time , Wherein, The temperature deviation is calculated as follows:

[0138] ,

[0139] Wherein, The return air temperature deviation is in Celsius.

[0140] Next, the temperature deviation is mapped to the correction amount of the fan speed. Similarly, to maintain smoothness, the correction amount is limited in amplitude:

[0141] ,

[0142] Wherein, The air volume correction coefficient determines how many percentage points of fan speed adjustment correspond to each degree of temperature deviation, with a value range of 1 to 12 percentage points per degree Celsius; The maximum fan speed correction amplitude per cycle is 1 to 10 percentage points.

[0143] Finally, the fan speed set value is updated: ​

[0144] ,

[0145] where F is the current fan speed percentage; and are the minimum and maximum fan speed limits allowed, usually 30% to 50%, 90% to 100%. This step ensures that the air volume adjustment follows the temperature change in the same gradual and limited manner, avoiding airflow instability caused by the fan speed oscillating between high and low gears when the load switches rapidly.

[0146] S3.4: Apply consistency constraints to the calculated temperature and air volume correction amounts to ensure that the adjustment directions are coordinated, and output the local control amount set and state flags for this coroutine.

[0147] To ensure that the temperature setting and fan speed are logically consistent in the adjustment direction and coordinate with each other rather than canceling each other out, consistency checks need to be performed on the calculated temperature and air volume correction amounts. For example, when the temperature setting is reduced to enhance refrigeration, the fan speed should theoretically increase to deliver more cold air, and vice versa.

[0148] For this purpose, define a consistency constraint indicator Q:

[0149] ,

[0150] where is the temperature setting gradual correction amount, is the fan speed gradual correction amount. Its physical meaning is: when is negative, indicating that the temperature setting is reduced, at this time is positive, indicating that the air volume is increased, and their product Q is positive; when is positive, indicating that the temperature setting is increased, at this time is negative, indicating that the air volume is reduced, and the product Q is also positive. Therefore, Q is positive, indicating that the adjustment direction of temperature and air volume is coordinated; if Q is negative, it indicates that the adjustment direction of the two is conflicting, which may be caused by the calculation timing or weight allocation in complex load changes.

[0151] When it is detected that Q is negative, the system will reduce the fan correction amount to weaken the impact of its conflicting direction. The fan correction amount after reduction is: where is the reduction coefficient, with a value range of 0.2 to 0.8. Then, the system uses to recalculate the fan speed update result .

[0152] Finally, the system outputs the set of local control variables of the coroutine in this cycle. The set includes: the updated temperature set value , the updated fan speed , and an enable flag indicating whether the current cycle is in the mapping maintenance phase. The updated and , that is, constitute the final control instructions directly issued to the precision air conditioner controlled by the coroutine in the mapping maintenance state. This step ensures that even after complex internal calculations, the control instructions generated within the coroutine are self-consistent and logically unified, so that in the stage when the hot spot migration is not stable, the smooth and coordinated control under the premise of not changing the space mapping relationship is perfectly realized, and a stable and reliable current state is provided for the entire system to transition to step S4 controlled reconstruction when necessary.

[0153] In the technical solution of the embodiments of the present disclosure, the system is started when it is determined to be in the mapping stable interval. The core is to make gradual and limited amplitude smooth adjustment of the control variables of the precision air conditioner under the premise of keeping the mapping relationship of the coroutine and the partition absolutely unchanged. This step calculates the coroutine-level adjustment target, and performs coordinated and limited-amplitude gradual correction of the temperature set and the fan speed, while introducing consistency constraints to ensure that the adjustment direction is self-consistent. This enables the output of the air conditioning system to gently and continuously follow the load changes, achieving disturbance-free dynamic adjustment without changing the system control structure, thereby completely avoiding control instruction jumps and system jitter when dealing with general load fluctuations.

[0154] S4: If it is determined that reconstruction is allowed, perform controlled coroutine mapping relationship adjustment with the goal of minimizing system disturbance.

[0155] When the system determines that the reconstruction gating quantity of a coroutine continuously exceeds the threshold, that is, the mapping stable interval has been confirmed to be invalid, perform a controlled and minimized coroutine mapping relationship reconstruction, and generate the final control instructions for smooth transition on this basis. Unlike frequent and blind global reconstruction, this step emphasizes triggering reconstruction only when necessary, and the reconstruction process itself is constrained and gradual, aiming to completely solve the control instability problem caused by the continuous migration of hot spots, while avoiding introducing new system jitter or performance cliffs in the reconstruction operation itself. The following sub-steps are implemented.

[0156] S4.1: Trigger controlled reconstruction determination according to the state flags of each coroutine, and immediately freeze the current control variables of the coroutine to be reconstructed as the reference for the reconstruction operation.

[0157] The system first makes a final decision based on the set of local control variables of each coroutine outputted in step S3. This set contains a critical map-keep-enabled flag, which is directly derived from the reconfiguration gating variable decision result in step S2.4. If the flag is reconfiguration-allowed, the system freezes the current control variables of the coroutine as the baseline and triggers the subsequent reconfiguration and instruction recalculation process; if the flag is map-maintained, the system directly adopts the control instructions generated in S3.

[0158] For any coroutine, the system checks its map-keep-enabled flag. If the flag is in the allowed state, it indicates that this coroutine is still within the stable interval, and the system determines that this time does not trigger the reconfiguration process for it. At the same time, to ensure control continuity, the system adopts the latest control variables calculated in step S3, i.e., the updated temperature set value and fan speed as the instructions to be executed in this period, and marks this state as the control variables being frozen. Freezing means that in the subsequent reconfiguration decision period, these control variables will remain unchanged and no longer accept gradual adjustments.

[0159] Conversely, if the map-keep-enabled flag of the coroutine is in the disallowed state, the system determines to trigger controlled reconfiguration for the coroutine. Triggering reconfiguration is a critical decision point, which means that the system confirms that the current load migration trend has been sustained and strong, and the original spatial mapping relationship is no longer optimal, and needs to be adjusted. At the same time of triggering reconfiguration, the system also immediately freezes the current control variables of the coroutine, i.e., adopts the and outputted in step S3 as the baseline value before the start of the reconfiguration operation, and marks them as and . The freezing operation has a dual significance: first, it provides a stable starting point for subsequent reconfiguration calculations; second, it prevents unpredictable fluctuations in control instructions due to continuous fine-tuning during the sensitive period when the mapping relationship is about to change, thereby avoiding secondary oscillation in the reconfiguration process.

[0160] Further, to prevent repeated reconfiguration when the load state changes rapidly but not persistently, the system can set a temporary reconfiguration lock period for each coroutine that has just completed reconfiguration, for example, ranging from 30 seconds to 300 seconds. During the lock period, even if the reconfiguration gating variable R calculated for the coroutine exceeds the threshold value again, the system does not trigger a new reconfiguration process, but continues to use the current mapping relationship and performs gradual adjustments. After the lock period ends, the normal stable interval decision and reconfiguration triggering logic are restored. This mechanism ensures that the system has enough time to stabilize under the new mapping relationship after each reconfiguration, effectively avoiding control structure oscillation caused by too short evaluation periods.

[0161] S4.2: For the coroutine that triggers the reconstruction, calculate the migration priority of its covered partition, filter and sort to obtain the candidate migration partition set for this reconstruction cycle.

[0162] For coroutines that have been triggered for refactoring, the system does not reshuffle all the partitions under their jurisdiction. Instead, it first accurately identifies which partitions are the key sources of mapping instability, i.e., the partitions with the most significant load migration behavior and the most urgent need for attribution adjustment. This minimizes the scope of the refactoring.

[0163] The system focuses on the set of partitions currently covered by the coroutine and calculates a migration priority for each partition i. This priority is a comprehensive indicator designed to identify partitions that are undergoing strong, active migration. Its calculation formula is as follows:

[0164] ,

[0165] in, Let i be the migration priority of the i-th partition; The average cumulative intensity of this region is derived from step S1; The migration strength of this partition is derived from step S1; The duration of this partition is derived from step S1. , and The corresponding normalization constants range from 0.02 to 0.8 seconds, 0.01 to 1.0 seconds, and 10 to 300 seconds, respectively. The design logic of this formula is that when both the average intensity S and the instantaneous intensity |G| are relatively large, and the duration D is not yet too long, the priority H reaches a higher value. This means that the partition is in the process of migration and in its active phase, and is the object that most needs to be monitored and adjusted.

[0166] The system calculates the migration priority for all partitions. Then, set a priority threshold. Filter out all Value not less than The partitions constitute the candidate partition set for this reconstruction. Subsequently, based on... The partitions within the set are sorted from largest to smallest value to obtain a clear migration priority sequence. For coroutines that have not triggered refactoring, their candidate partition set is empty. This step greatly reduces the complexity of refactoring and the impact on the overall system stability.

[0167] S4.3: With the goal of minimizing overall system misfit, find the optimal target goroutine for candidate migration partitions, perform minimal mapping relationship adjustments, and output the new mapping relationship.

[0168] After obtaining the list of partitions that need to be migrated, the system performs the actual mapping relationship adjustment to adjust a small number of the most critical partitions to a more suitable coroutine instead of starting over.

[0169] First, the system determines the number of partitions to be migrated in this reconstruction cycle plan . To maintain control continuity, a small value is usually set, for example, 1 to 5. The system selects the top priority partitions from the migration priority sequence as the set of partitions to be actually migrated in this cycle.

[0170] For each partition to be migrated, find the best migration target coroutine. The system will traverse all other coroutines in the current data center as potential targets. As a more preferred implementation, to improve the rationality of airflow organization after reconstruction, the system can preferentially consider coroutines that are adjacent to or have a high degree of association with the partition to be migrated in physical location. Specifically, a physical association factor can be preset for each pair of “partition-coroutine” to represent the closeness of the two in spatial layout or air duct system. First, filter out a candidate coroutine subset with a physical association factor higher than a preset threshold, and then only in this subset, perform subsequent adaptation cost calculation and comparison.

[0171] For each potential target coroutine, the system calculates the adaptation cost C after migrating the partition. The adaptation cost evaluates the discomfort that the migration operation may cause to the target coroutine, and the calculation formula is:

[0172]

[0173] wherein is the new coroutine level adjustment target amount of the target coroutine assuming that the partition is migrated; is the stable reference pressure of the target coroutine itself; is the total heat load of the region governed by the target coroutine after migration; is the total load normalization constant of the coroutine, which ranges from 50 kW to 500 kW, and can be set according to the typical management scale of a single coroutine in a data center. and are the pressure deviation weight and the size penalty weight, respectively, both of which are dimensionless numbers and have a value range of 0.5 to 5.0, which can be determined by system calibration to balance the influence of the two terms in the formula. The meaning of this formula is that the adaptation cost consists of two parts: one is the degree of damage to the original pressure balance of the target coroutine after migration; the other is the possible excessive size burden of the target coroutine due to receiving a new region. The system calculates the C value of all potential targets for the partition to be migrated, and selects the coroutine with the smallest C value as the final migration target.

[0174] By calculating the adaptation cost C of all ​The system completes the mapping adjustment of this reconstruction cycle by repeating the above calculation cost and selecting the optimal target for each partition to be migrated. Finally, the new coroutine mapping relationship set after reconstruction is output. This step reflects that each change in mapping relationship is local, evaluated, and targeted at minimizing the overall system discomfort, ensuring that the state of the reconstructed system is not only new, but also more optimal and stable.

[0175] S4.4: Calculate the control quantity based on the new coroutine mapping relationship, and fuse it with the frozen reference control quantity through a smoothing fusion algorithm to generate and issue the final system-level control instruction to each actuator.

[0176] After determining the new coroutine mapping relationship, the system needs to generate the final control instruction based on this relationship.

[0177] The system first calculates the temperature set value for each coroutine based on the completely new coroutine coverage range and fan speed . The calculation principle and steps are consistent with those described in step S3, but based on the reconstructed partition set.

[0178] However, if the newly calculated and are directly issued for execution, there may be a large difference between the values and the frozen instructions , before reconstruction, causing the device to execute the action abruptly. To this end, the system introduces a smoothing fusion mechanism. The temperature set instruction issued to the precision air conditioning execution unit is generated by the following formula:

[0179] ,

[0180] Similarly, the fan speed instruction is:

[0181] ,

[0182] where is the smoothing factor, with a value range of 0.3 to 0.9. When is larger, the final instruction is more biased towards the frozen value before reconstruction, with a gentle change; when is smaller, it transitions to the new calculated value more quickly. By adjusting , the speed of the system state migration after reconstruction can be controlled, achieving a soft landing.

[0183] The and generated through this smoothing fusion mechanism, i.e., constitute the final control instruction issued to the precision air conditioner under the allowed reconstruction state.

[0184] In the technical solution of the embodiments of the present disclosure, when the stable interval is confirmed to be invalid, a controlled and minimized coroutine reconstruction is performed. This step accurately identifies a small number of partitions that most urgently need to be adjusted by migrating priority, and performs mapping optimization with the principle of minimizing the inadaptation cost of target coroutines. When generating new control instructions, a smooth fusion mechanism is used to weight and fuse the frozen instructions before reconstruction and the calculation instructions after reconstruction, to ensure the continuity of the output instructions. The entire reconstruction process is local, evaluated and smoothly transitioned, so that in the scenario where the control architecture must be changed to cope with continuous and severe load migration, the control instability problem can be completely solved, and the secondary performance impact and instruction jitter caused by the reconstruction operation itself are eliminated.

[0185] S5: Execute the control instructions adjusted by S3 or S4 above, and verify and close loop with the system feedback state after instruction execution.

[0186] After completing the gradual adjustment under mapping maintenance in step S3 or the controlled adjustment under reconstruction permission in step S4, the system generates corresponding control instructions. These instructions are put into execution, and the system state after execution is monitored and verified, and the key feedback information is used for new round of control decision, so as to form an adaptive and closed loop control loop. The following sub-steps are implemented.

[0187] S5.1: Package and issue system-level control instructions to each execution mechanism.

[0188] The system collects all the control instructions generated by the coroutines in this period. For the coroutines in the mapping maintenance state, the instructions are the updated temperature set values and fan speeds output in step S3; for the coroutines that have undergone reconstruction adjustment, the instructions are the temperature set values and fan speeds generated by step S4 through smooth fusion. The system packages these instructions according to the correspondence between the coroutines and physical devices, forms a unified system-level control instruction set, and reliably issues it to the local controller of each precision air conditioner for execution through the control network. This step ensures that the control decision is accurately and synchronously converted into the action of the field device.

[0189] S5.2: Monitor the instruction execution state and collect key feedback data of each partition.

[0190] After the control instructions are issued and executed, the system starts a monitoring window to collect key feedback data reflecting the effect of instruction execution. This includes: the return air temperature , the supply air temperature , and the actual running speed and real-time thermal loads of the zones obtained from the power monitoring system All data are collected and aligned with the same uniform sampling period as step S1.1, ensuring the consistency of data timing and laying the foundation for subsequent comparative analysis.

[0191] S5.3: Verify the control effect based on the feedback data and calculate the system performance indicators.

[0192] Based on the collected feedback data, the system calculates the core performance indicators for verifying the control effect. First, the temperature control deviation of each zone is calculated wherein is the target temperature of the zone (derived from the current temperature setting instruction of its corresponding coroutine). Second, the overall energy efficiency trend of the system is evaluated, for example, the ratio change of total refrigeration output to total power consumption is calculated. Finally, the air flow organization stability is analyzed by comparing whether the difference of return air temperature of adjacent zones exceeds the safety threshold to determine whether there is local air flow short circuit or hot spot residue. This step quantifies the physical response of the system into assessable performance indicators.

[0193] S5.4: Integrate the feedback information and link it to the next control cycle to form a closed loop.

[0194] The system integrates the results generated in the monitoring and verification link of this cycle to form a closed loop feedback information package. Specifically, the , and data collected in S5.2 are used as input data sources for building the continuous change representation of cold and heat load space distribution in step S1 of the next control cycle. At the same time, the performance indicators calculated in S5.3 are compared with the preset target. If the indicators continue to deteriorate (such as long-term excessive temperature deviation), it can be used as an abnormal signal to trigger higher-level system diagnosis or adjust the control strategy parameters (such as fine-tuning the stable interval tolerance coefficient). This step realizes the complete closed loop of control decision, execution, verification, feedback, and re-decision, enabling the entire control system to have the ability of adaptive optimization and long-term stable operation.

[0195] In the technical solution of the embodiments of the present disclosure, by adding an independent control instruction execution and state verification step, a complete closed loop of multi-coroutine control process is realized. This step not only ensures the reliable execution of control instructions, but more importantly, through real-time monitoring and effect quantification of the system state after execution, it provides real and timely feedback data for the load trend analysis of the next cycle. This makes the entire control method no longer a one-time open-loop response, but a dynamic adaptive system that can continuously learn and optimize based on historical action effects, thereby fundamentally improving the long-term control precision, energy efficiency and stability of the data center precision air conditioning system when facing complex and variable loads.

[0196] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, which can include a processor, a communications interface, a memory and a communications bus, wherein the processor, the communications interface and the memory complete communications with each other through the communications bus. The processor can invoke a logical instruction in the memory to execute the method provided by each of the above embodiments.

[0197] In addition, the logical instruction in the memory described above can be implemented in the form of a software functional unit and sold or used as an independent product, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present disclosure, in essence or the part that contributes to the prior art, or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0198] On the other hand, the present disclosure also provides a non-transitory computer readable storage medium having a computer program stored thereon, which is executed by a processor to implement the method provided by each of the above embodiments.

[0199] The device embodiments described above are only schematic, wherein the units shown as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e. they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the present embodiment according to actual needs. Those skilled in the art can understand and implement it without creative labor.

[0200] Those skilled in the art can clearly understand the implementation of the various embodiments by means of software and necessary general hardware platforms through the description of the above embodiments, and of course, the various embodiments can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product. The computer software product can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0201] It should be understood that the above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than limit them; although the present disclosure has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. A multi-process control method for a precision air conditioning system in a data center, characterized in that, The method includes: S1. Based on the monitoring data of each partition of the data center, construct a continuous change characterization quantity that reflects the migration trend of hot and cold loads between partitions; S2. Based on the continuously changing representation quantity, dynamically construct and determine the stable mapping interval for each coroutine to decide whether to maintain the current mapping or allow reconstruction. S3. If it is determined that the current mapping should be maintained, the temperature and air volume setpoints of the air conditioner controlled by the corresponding coroutine will be gradually and collaboratively adjusted. S4. If it is determined that reconfiguration is allowed, then perform controlled coroutine mapping relationship adjustment with the goal of minimizing system disturbance; S5. Execute the control command obtained by adjusting S3 or S4 above, and verify and connect the closed loop based on the system feedback status after the command is executed. S1 includes: Based on the return air temperature, supply air temperature and heat load data of each partition of the data center, the instantaneous cooling pressure that characterizes the overall cooling demand of the partition is calculated. Calculate the rate of change of the instantaneous cooling pressure to obtain a zonal migration intensity sequence that reflects the speed and direction of load changes; Based on the migration intensity sequence of the partition, statistical analysis is performed within a sliding time window to generate a continuous variation characterization quantity including migration direction, average cumulative intensity and duration. S3 includes: Calculate the target direction of the coroutine-level adjustment based on the instantaneous cooling pressure of the partition under the coroutine's jurisdiction; Based on the deviation between the target adjustment amount and the stable reference pressure, the gradual correction amount of the temperature setpoint is calculated, and the gradual correction amount of the temperature setpoint is applied to the current temperature setpoint to obtain the updated temperature setpoint. Based on the deviation between the average return air temperature of the partition under the jurisdiction of the coroutine and the updated temperature setpoint, the fan speed progressive correction amount is calculated, and the fan speed progressive correction amount is applied to the current fan speed setpoint. Consistency constraints are applied to the gradual correction amount of the temperature setpoint and the gradual correction amount of the fan speed to ensure coordinated adjustment directions; The updated temperature setpoint and fan speed setpoint are output as corresponding control commands; S4 includes: For the coroutine that triggers the refactoring, the migration priority is calculated based on the activity level of its partition migration, and candidate migration partitions are selected. Based on the principle of minimizing the misfit cost of the target goroutine, the optimal target goroutine is determined for the candidate migration partition, and the mapping relationship is adjusted. The control quantity is calculated based on the adjusted mapping relationship. The newly calculated control quantity is then weighted and fused with the baseline control quantity frozen before reconstruction using a smooth fusion algorithm to generate the corresponding control command.

2. The multi-process control method for a data center precision air conditioning system according to claim 1, characterized in that, S2 includes: Summarize the continuously changing characteristics of each partition under the jurisdiction of the coroutine, and calculate the coroutine mapping offset that reflects the degree of impact on the mapping relationship; The upper boundary of the stable mapping interval at the current moment is dynamically determined based on the historical sequence of the coroutine mapping offset. The reconstruction gating amount is calculated by combining the intensity and duration of the real-time offset exceeding the upper boundary. The current mapping is maintained or reconstruction is allowed by comparing the reconstruction gating value with a preset threshold.

3. The multi-process control method for a data center precision air conditioning system according to claim 2, characterized in that, The step of determining whether to maintain the current mapping or allow reconstruction by comparing the reconstruction gate with a preset threshold specifically includes: Reconstruction is only allowed after the reconstruction gate value continuously exceeds the preset threshold for a preset confirmation time; otherwise, the current mapping is maintained.

4. The multi-process control method for a data center precision air conditioning system according to claim 1, characterized in that, The calculation of migration priority is specifically as follows: The migration priority is calculated by comprehensively considering the average cumulative intensity, instantaneous migration intensity, and duration of the partition. The higher the average cumulative intensity and instantaneous migration intensity, and the shorter the duration, the higher the migration priority.

5. The multi-process control method for a data center precision air conditioning system according to claim 1, characterized in that, S5 includes: The control commands are sent to each precision air conditioning actuator. Collect return air temperature, supply air temperature and heat load feedback data for each zone after the command is executed; The control effect is verified based on the feedback data, and the feedback data is used as the input to construct the continuously changing characteristic quantity in the next control cycle to form a closed-loop control.

6. An electronic device, characterized in that, The electronic device includes a memory and at least one processor, the memory storing a computer program, and the processor executing the computer program to implement the multi-processor control method for a data center precision air conditioning system according to any one of claims 1-5.

7. A computer storage medium, characterized in that, It stores a computer program, which, when executed, implements the multi-process control method for a data center precision air conditioning system according to any one of claims 1-5.

Images