A Terminal Control Method for Data Center Air Conditioning Based on Deep Learning and Thermodynamic Mechanism Models
An improved LTC method combining deep learning and thermodynamic mechanism models dynamically adjusts the operating parameters of air conditioning terminals, solving the problem of adjustment deviation in the control of computer room air conditioning and achieving efficient energy saving and safe control of computer room air conditioning.
Patent Information
- Application Number
- CN202611512428.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-28
- Publication Date
- 2026-11-17
AI Technical Summary
Existing methods for controlling air conditioning in computer rooms make it difficult to accurately determine the actual impact of adjusting a single air conditioning terminal on different areas, leading to deviations in the timing and magnitude of air conditioning terminal adjustments, which affects the stability and security of energy-saving control in computer rooms.
By combining deep learning and thermodynamic mechanism models, an improved LTC is constructed to predict the actual impact of each air conditioning terminal on the controlled heat zone, determine the control sequence and output adjustment amount, adjust the operating parameters of the air conditioning terminal one by one, form terminal control commands, and correct net heat change based on the heat conservation relationship.
It improves the accuracy of temperature prediction and the stability of continuous control in the controlled hot zone, reduces the impact of load fluctuations and heat accumulation on temperature prediction results, and ensures the safety and effectiveness of energy-saving control of computer room air conditioning.
Smart Images

Figure CN123110728A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data center air conditioning control technology, and in particular to a data center air conditioning terminal control method based on deep learning and thermodynamic mechanism models. Background Technology
[0002] In the field of data center air conditioning control technology, the operating load of computing equipment in data center computer rooms is constantly increasing, leading to a continuous increase in heat generation within the computer room and significant differences in temperature variations across different areas. Existing data center air conditioning control methods typically adjust the air supply status based on temperature monitoring results and predict subsequent temperatures by combining this with changes in equipment load, in order to reduce the risk of localized overheating and control cooling energy consumption.
[0003] In actual operation, the heat inside the computer room continuously changes with equipment load, air supply status, and airflow. The temperature change relationship established under historical operating conditions is difficult to maintain stability over a long period. When the equipment load changes rapidly and local heat continues to accumulate, the predicted results are prone to deviating from the actual temperature response, leading to deviations in the timing and magnitude of air conditioning terminal adjustments.
[0004] The air supply ranges of different air conditioning terminals in a computer room overlap, and temperature changes in one area are often influenced by neighboring air conditioning terminals. Existing methods typically determine the adjustment target directly based on the area temperature and equipment operating status, making it difficult to accurately judge the actual impact of adjusting a single air conditioning terminal on different areas. When multiple air conditioning terminals are adjusted consecutively, the previous adjustment will change the relationship between the temperature and air supply in subsequent areas, easily causing the original judgment results to lose accuracy, thereby affecting the stability and security of energy-saving control in the computer room.
[0005] Therefore, how to provide a data center air conditioning terminal control method based on deep learning and thermodynamic mechanism models is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0006] One objective of this invention is to propose a control method for computer room air conditioning terminals based on deep learning and thermodynamic mechanism models. This invention combines thermodynamic mechanism analysis with an improved LTC to predict the thermal state of the computer room. Based on the actual impact of each air conditioning terminal on the controlled heat zone, the control sequence and output adjustment amount are determined. This method has the advantages of high prediction accuracy, good energy-saving effect, and high control safety.
[0007] The data center air conditioning terminal control method based on deep learning and thermodynamic mechanism model according to an embodiment of the present invention includes the following steps: Collect controlled hot zone temperature, IT load, air conditioning terminal operating parameters and operating power, and use thermodynamic mechanism model combined with temperature changes to form thermal mechanism mismatch; An improved LTC is constructed, which divides the liquid state into air-side liquid state and heat storage-side liquid state. The air-side liquid state is updated according to the operating parameters of the air conditioning terminal, and the heat storage-side liquid state is updated according to the IT load and the temperature of the controlled hot zone. The residual heat state on the air side is transferred to the heat storage-side liquid state. Based on the thermal mechanism mismatch adjustment of the two types of liquid time constants, the net heat change and future temperature of each controlled heat zone are predicted to form a thermal safety margin; The operating parameters of each air conditioning terminal are restored to their pre-control values, and the predicted net heat change before and after restoration is compared to form the initial marginal cooling capacity. The initial marginal cooling capacity is corrected for thermal conservation based on the thermodynamic mechanism model, eliminating duplicate attribution of the same net heat change and forming a cooling capacity attribution relationship; Based on the cooling capacity attribution relationship, the air conditioning terminals to be reduced are screened, and the reduction sequence is determined by combining the thermal safety margin and operating power. The output adjustment amount of the air conditioning terminals is determined one by one to form the adjusted air conditioning terminal operating parameters. After each formation, the net heat change is re-predicted, the cooling capacity attribution relationship is updated, and the terminal control command is formed. Execute the end-point control command, and re-establish the thermal mechanism mismatch based on the controlled thermal zone temperature after control, updating the two types of liquid states and the liquid time constant.
[0008] Optionally, the formation of the thermal mechanism mismatch includes the following steps: Arrange the controlled thermal zone temperatures according to the acquisition time, and select adjacent acquisition times; Starting from the controlled hot zone temperature at the previous acquisition time, the corresponding IT load and air conditioning terminal operating parameters are input into the thermodynamic mechanism model to extrapolate the controlled hot zone temperature at the next acquisition time. The predicted temperature of the controlled thermal zone is compared with the temperature of the controlled thermal zone at the next data acquisition time, resulting in a mismatch in thermal mechanism.
[0009] Optionally, the construction of the improved LTC includes the following steps: The continuous liquid state of LTC is divided into air-side liquid state and heat storage-side liquid state, and corresponding liquid time constants are established for each. The air-side liquid state is updated based on the air conditioning terminal operating parameters, and the heat storage-side liquid state is updated based on the IT load and the temperature of the controlled hot zone. The state change that is not offset by the current heat extraction after the air-side liquid state is updated is retained as the air-side residual heat state and transferred to the heat storage side liquid state.
[0010] Optionally, the residual heat state on the air side is screened out from the changes in the liquid state on the air side at adjacent acquisition times, and the portion that is not offset by the heat extraction effect of the corresponding air conditioning terminal is selected. The residual heat state on the air side is mapped according to the state dimension of the liquid state on the heat storage side, and the liquid state on the heat storage side at the next acquisition time is updated based on the liquid state on the heat storage side at the previous acquisition time and the mapping result.
[0011] Optionally, the formation of the thermal safety margin includes the following steps: Arrange thermal mechanism mismatches according to the time of data collection, and extract adjacent changes and continuous changes in the same direction; The liquid time constant on the air side is adjusted based on adjacent changes, and the liquid time constant on the heat storage side is adjusted based on continuous changes in the same direction. The corresponding liquid state is updated using two types of adjusted liquid time constants to predict the net heat change of each controlled heat zone; Based on net heat change and controlled heat zone temperature, future temperatures are projected, and a thermal safety margin is formed by the interval between the future temperature and the upper limit of the controlled heat zone temperature.
[0012] Optionally, the formation of the initial marginal cooling capacity includes the following steps: Select each air conditioning terminal one by one, save the corresponding operating parameters, and restore the corresponding operating parameters to the values before adjustment; The net heat change was re-predicted using the improved LTC, and the predicted net heat change before and after the recovery was compared to obtain the differences in net heat change in each controlled heat zone. The corresponding operating parameters are restored to the saved values, and the net heat change difference is mapped to the current air conditioning terminal and the controlled heat zone to form the initial marginal cooling capacity.
[0013] Optionally, the formation of the cooling load attribution relationship includes the following steps: Based on the initial marginal cooling capacity of each air conditioning terminal in the controlled heat zone, obtain the corresponding predicted net heat change; By performing thermal conservation matching between each initial marginal cooling capacity and the corresponding predicted net heat change, the common net heat change corresponding to different air conditioning terminals can be located. Remove the common corresponding portion from each initial marginal cooling capacity, and retain the marginal cooling capacity corresponding to a single air conditioning terminal; Based on the retention results of air conditioning terminals and controlled hot zones, a cooling capacity attribution relationship is formed.
[0014] Optionally, determining the cutting order includes the following steps: The controlled heat zone and marginal cooling capacity of each air conditioning terminal are obtained according to the cooling capacity attribution relationship. Compare the thermal safety margins of each controlled heat zone corresponding to the same air conditioning terminal to obtain the minimum thermal safety margin; Select air conditioning terminals to be reduced based on marginal cooling capacity and minimum thermal safety margin; The order of reduction is determined by combining the minimum thermal safety margin and operating power of each air conditioning terminal to be reduced.
[0015] Optionally, the generation of the end-point control command includes the following steps: Determine the output adjustment amount of the current air conditioning terminal to be reduced according to the reduction order, and obtain the adjusted air conditioning terminal operating parameters; Based on the adjusted air conditioning terminal operating parameters, the improved LTC is used to re-predict the net heat change of each controlled heat zone and update the heat safety margin. The operating parameters of the remaining air conditioning terminals to be reduced are restored one by one, the predicted net heat changes before and after restoration are compared, the initial marginal cooling capacity is reconstructed, and the cooling capacity attribution relationship is updated after heat conservation correction. Based on the updated thermal safety margin and cooling capacity attribution relationship, the reduction sequence of the remaining air conditioning terminals to be reduced is re-determined, and the terminal control instructions are formed by summarizing the output adjustments of each time.
[0016] Optionally, the updating of the two types of liquid states and the liquid time constant includes the following steps: Execute the terminal control command and collect the temperature of the controlled thermal zone after control. The controlled hot zone temperature before the execution of the terminal control command is taken as the starting point for the deduction. The controlled hot zone temperature is deduced based on the corresponding IT load and the air conditioning terminal operating parameters after the control. By comparing the predicted temperature of the controlled thermal zone with the temperature of the controlled thermal zone after adjustment, a new mismatch in the thermal mechanism is formed; The newly formed thermal mechanism mismatch is integrated into the existing thermal mechanism mismatch sequence, and adjacent changes and continuous unidirectional changes are updated. The liquid time constant on the air side and the liquid time constant on the heat storage side are adjusted respectively, and the corresponding liquid states are updated.
[0017] The beneficial effects of this invention are: (1) This invention combines thermodynamic mechanism analysis with improved LTC, dynamically adjusts the two types of liquid time constants according to the thermal mechanism mismatch, characterizes the rapid thermal response on the air side of the computer room and the slow thermal change on the heat storage side, reduces the impact of load fluctuation and heat accumulation on temperature prediction results, and improves the accuracy of temperature prediction and continuous control stability of the controlled heat zone.
[0018] (2) The present invention restores the operating parameters of the air conditioning terminal one by one to form the initial marginal cooling capacity, and combines the heat conservation relationship to eliminate the repeated attribution of different air conditioning terminals to the same net heat change, identify the actual cooling capacity contribution of each air conditioning terminal to different controlled heat zones, and reduce the control judgment deviation caused by the cross of air supply action.
[0019] (3) The present invention determines the order of air conditioning terminal reduction based on the cooling capacity attribution relationship, thermal safety margin and operating power, and re-predicts the net heat change and updates the cooling capacity attribution relationship after determining the output adjustment amount. It continuously corrects the subsequent control decision based on the thermal state after the previous adjustment, ensuring the temperature safety of the controlled hot zone while reducing unnecessary cooling output and improving the energy-saving control effect of the computer room air conditioning. Attached Figure Description
[0020] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is an overall flowchart of the data center air conditioning terminal control method based on deep learning and thermodynamic mechanism model proposed in this invention; Figure 2 This is a schematic diagram of the air supply coverage of the controlled heat zone by the air conditioning terminal in this invention; Figure 3 This is a schematic diagram of the attribution relationship of cooling capacity after thermal conservation correction in this invention. Detailed Implementation
[0021] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.
[0022] refer to Figures 1-3 A terminal control method for computer room air conditioning based on deep learning and thermodynamic mechanism models includes the following steps: Collect controlled hot zone temperature, IT load, air conditioning terminal operating parameters and operating power, and use thermodynamic mechanism model combined with temperature changes to form thermal mechanism mismatch; An improved LTC is constructed, which divides the liquid state into air-side liquid state and heat storage-side liquid state. The air-side liquid state is updated according to the operating parameters of the air conditioning terminal, and the heat storage-side liquid state is updated according to the IT load and the temperature of the controlled hot zone. The residual heat state on the air side is transferred to the heat storage-side liquid state. Based on the thermal mechanism mismatch adjustment of the two types of liquid time constants, the net heat change and future temperature of each controlled heat zone are predicted to form a thermal safety margin; The operating parameters of each air conditioning terminal are restored to their pre-control values, and the predicted net heat change before and after restoration is compared to form the initial marginal cooling capacity. The initial marginal cooling capacity is corrected for thermal conservation based on the thermodynamic mechanism model, eliminating duplicate attribution of the same net heat change and forming a cooling capacity attribution relationship; Based on the cooling capacity attribution relationship, the air conditioning terminals to be reduced are screened, and the reduction sequence is determined by combining the thermal safety margin and operating power. The output adjustment amount of the air conditioning terminals is determined one by one to form the adjusted air conditioning terminal operating parameters. After each formation, the net heat change is re-predicted, the cooling capacity attribution relationship is updated, and the terminal control command is formed. Execute the end-point control command, and re-establish the thermal mechanism mismatch based on the controlled thermal zone temperature after control, updating the two types of liquid states and the liquid time constant.
[0023] In this embodiment, the formation of the thermal mechanism mismatch includes the following steps: Arrange the controlled thermal zone temperatures according to the acquisition time, and select adjacent acquisition times; A time series is established based on the acquisition times of the controlled heat zone temperatures. Temperature records are arranged consecutively according to the same controlled heat zone, and the operating intervals are defined by the acquisition times corresponding to adjacent temperature records. The IT load and air conditioning terminal operating parameters are mapped to each operating interval according to the timestamps. When the sampling times are inconsistent, the continuously changing operating parameters are interpolated according to the time of adjacent records, and the most recent valid record is used for the operating parameters that have not changed, thus establishing the time correspondence between the controlled heat zone temperature, IT load, and air conditioning terminal operating parameters.
[0024] Starting from the controlled hot zone temperature at the previous acquisition time, the corresponding IT load and air conditioning terminal operating parameters are input into the thermodynamic mechanism model to extrapolate the controlled hot zone temperature at the next acquisition time. A continuous historical operating range is selected. The heat output per unit time of the controlled hot zone is determined based on the IT load of each operating range. The heat extraction per unit time of the air conditioning terminal is determined based on the air volume of the air conditioning terminal and the temperature difference between the supply air temperature and the return air temperature. The temperature change of the controlled hot zone and the temperature difference between adjacent controlled hot zones are recorded. Based on the heat balance relationship of the controlled hot zone, the equivalent heat capacity of the controlled hot zone and the heat transfer parameters of adjacent controlled hot zones are simultaneously fitted. The deviation between the temperature change corresponding to the fitted parameters and the historical temperature change is used as the fitting target to obtain the parameters used in the thermodynamic mechanism model.
[0025] Using the temperature of the controlled hot zone at the previous data acquisition moment as the starting point for temperature propagation, the IT load within the corresponding operating range is converted into the heat output per unit time of the controlled hot zone. The heat output per unit time of the air conditioning terminal is determined based on the air volume of the air conditioning terminal and the temperature difference between the supply air temperature and the return air temperature. The net heat input per unit time within the operating range is determined based on the heat output per unit time of the controlled hot zone, the heat output per unit time of the air conditioning terminal, and the heat exchange per unit time between adjacent controlled hot zones. The temperature of the controlled hot zone at the next data acquisition moment is then extrapolated by combining the equivalent heat capacity of the controlled hot zone and the length of the operating range.
[0026] The predicted temperature of the controlled thermal zone is compared with the temperature of the controlled thermal zone at the next data acquisition time, resulting in a mismatch in thermal mechanism.
[0027] Compare the predicted temperature of the controlled thermal zone at the next acquisition time with the temperature of the controlled thermal zone at the same acquisition time, and record the positive and negative directions and amplitudes of the difference between the two. This serves as the thermal mechanism mismatch for the corresponding controlled thermal zone at the corresponding acquisition time. The thermal mechanism mismatch is continuously saved according to the controlled thermal zone and acquisition time for use when adjusting the two types of liquid time constants in the improved LTC.
[0028] In this embodiment, the construction of the improved LTC includes the following steps: The continuous liquid state of LTC is divided into air-side liquid state and heat storage-side liquid state, and corresponding liquid time constants are established for each. Following the dimensions of the improved LTC continuous liquid state, one part of the state dimension is used to characterize the liquid state on the air side, and another part is used to characterize the liquid state on the heat storage side. Both parts of the state dimension remain fixed during training. Historical operating intervals where the operating parameters of the air conditioning terminal change are selected. Two exponentially decaying terms containing independent time constant parameters are fitted to the controlled hot zone temperature response sequence after the change in operating parameters, obtaining two time constant parameters. The time constant parameter with the smaller value is used as the initial liquid time constant for the air side liquid state, and the time constant parameter with the larger value is used as the initial liquid time constant for the heat storage side liquid state.
[0029] The air-side liquid state is updated based on the air conditioning terminal operating parameters, and the heat storage-side liquid state is updated based on the IT load and the temperature of the controlled hot zone. The operating parameters of the air conditioning terminal are normalized and encoded according to a fixed field order, and the encoded results are mapped to the corresponding dimension of the air-side liquid state. The air-side liquid state is updated based on the air-side liquid state at the previous acquisition time and the current encoding result. The IT load and controlled hot zone temperature are normalized and encoded, and the encoded results are mapped to the corresponding dimension of the heat storage side liquid state. The heat storage side liquid state is updated based on the heat storage side liquid state at the previous acquisition time and the current encoding result. The heat extraction effect is formed based on the air volume and the temperature difference between the supply air temperature and the return air temperature at the air conditioning terminal. The heat extraction effect is normalized and encoded, and the encoded results are mapped to the corresponding dimension of the air-side liquid state. When training the improved LTC, the mapping parameters corresponding to the heat extraction effect are adjusted according to the predicted deviation of the controlled hot zone temperature.
[0030] The state change that is not offset by the current heat extraction after the air-side liquid state is updated is retained as the air-side residual heat state and transferred to the heat storage side liquid state.
[0031] The state change is obtained based on the air-side liquid state at the current and previous acquisition times. The current heat extraction is then converted into a heat extraction state response according to the mapping parameters. The state change and the heat extraction state response are compared in the same state dimension. If the directions are opposite, the smaller of the two amplitudes is used as the cancellation factor. The original direction is retained after subtracting the cancellation factor from the state change amplitude. If the directions are the same, the state change is retained. The retained state changes in each state dimension form the residual heat state on the air side.
[0032] In this embodiment, the residual heat state on the air side is filtered out from the changes in the liquid state on the air side at adjacent acquisition times, and the portion that is not offset by the heat extraction effect of the corresponding air conditioning terminal is selected. The residual heat state on the air side is mapped according to the state dimension of the liquid state on the heat storage side, and the liquid state on the heat storage side at the next acquisition time is updated based on the liquid state on the heat storage side at the previous acquisition time and the mapping result.
[0033] When training the improved LTC, the residual thermal state on the air side is transformed to the corresponding dimension of the liquid state on the heat storage side using a state mapping matrix. The transformation result, IT load, and controlled heat zone temperature are used together to update the liquid state on the heat storage side. The updated liquid state on the heat storage side and the corresponding liquid state on the air side are used together to predict net heat change. The controlled heat zone temperature is extrapolated based on the predicted net heat change. The parameters of the state mapping matrix are adjusted according to the deviation between the extrapolated controlled heat zone temperature and the controlled heat zone temperature at the corresponding acquisition time to obtain the state mapping matrix.
[0034] The current residual thermal state on the air side is transformed to the corresponding dimension of the liquid state on the heat storage side using a state mapping matrix. The transformation result, IT load, and controlled heat zone temperature are used together for continuous state updates from the liquid state on the heat storage side at the previous acquisition time to the next acquisition time. The residual thermal state on the air side is reconstructed based on the updated liquid state on the air side at the next acquisition time.
[0035] In this embodiment, the formation of the thermal safety margin includes the following steps: Arrange thermal mechanism mismatches according to the time of data collection, and extract adjacent changes and continuous changes in the same direction; Thermal mechanism mismatches are arranged according to the controlled thermal regions. The difference between the thermal mechanism mismatch at the current acquisition time and the thermal mechanism mismatch at the previous acquisition time is used to record adjacent changes. When the thermal mechanism mismatches at consecutive acquisition times maintain the same positive and negative directions, the corresponding amplitudes are accumulated sequentially. When the positive and negative directions change, the current accumulation ends, and accumulation starts again from the acquisition time after the direction change to obtain continuous changes in the same direction.
[0036] The liquid time constant on the air side is adjusted based on adjacent changes, and the liquid time constant on the heat storage side is adjusted based on continuous changes in the same direction. For the air-side liquid state, within the historical operating range where the operating parameters of the air conditioning terminal change, the moment of the parameter change is taken as the response start point, and the moment when the amplitude of the controlled hot zone temperature change rate first falls back to the range of the temperature change rate amplitude within the previous operating range is taken as the response end point. The time interval between the two is taken as the temperature response duration. Based on the fitted correspondence between adjacent change amplitudes and temperature response durations in the historical operating range, the air-side liquid time constant is determined from the current adjacent change amplitude.
[0037] For the liquid state on the thermal storage side, the starting point of the response is defined as the moment when the thermal mechanism mismatch begins and the data collection time continuously maintains the same positive and negative direction, and the ending point is defined as the last data collection time before the positive and negative directions change. The time interval between the two is defined as the duration of the continuous thermal response. Based on the correlation between the cumulative degree of continuous unidirectional changes in historical operating intervals and the duration of the continuous thermal response, the liquid time constant on the thermal storage side is determined by the current cumulative degree of continuous unidirectional changes. Both types of liquid time constants are limited to the corresponding value range formed by historical operating data.
[0038] The corresponding liquid state is updated using two types of adjusted liquid time constants to predict the net heat change of each controlled heat zone; The air-side liquid time constant and the heat storage-side liquid time constant are substituted into the continuous-time update process of the corresponding liquid states to obtain the updated two types of liquid states. These two types of liquid states are then concatenated and input into the output mapping layer, which outputs the predicted net heat change according to the controlled heat zone. The output mapping layer is trained synchronously during the improved LTC training process. Based on the controlled heat zone temperature change, the equivalent heat capacity of the controlled heat zone, and the length of the operating interval at adjacent acquisition times, the historical net heat change is deduced. This historical net heat change is used as the supervision result, and the output mapping parameters are adjusted based on the deviation between the predicted net heat change and the supervision result. For continuous prediction, the updated two types of liquid states at the previous prediction time are used as the starting point for the next prediction time. The predicted net heat change is represented by the net heat input per unit time.
[0039] Based on net heat change and controlled heat zone temperature, future temperatures are projected, and a thermal safety margin is formed by the interval between the future temperature and the upper limit of the controlled heat zone temperature.
[0040] Using the current temperature of the controlled hot zone as the starting point for temperature propagation, and following the heat balance relationship between net heat change, propagation time, and equivalent heat capacity of the controlled hot zone, the predicted net heat change is converted into temperature change for the corresponding prediction period, and the future temperature is obtained by advancing hourly. The upper limit of the controlled hot zone temperature adopts the upper limit of temperature control executed by the corresponding controlled hot zone during the operation of the computer room, and the remaining interval between the future temperature and the corresponding upper limit of temperature is used as the thermal safety margin; when there are consecutive prediction times, the thermal safety margin is saved according to the prediction time for subsequent comparison of thermal safety constraints of different controlled hot zones and prediction times.
[0041] In this embodiment, the formation of the initial marginal cold amount includes the following steps: Select each air conditioning terminal one by one, save the corresponding operating parameters, and restore the corresponding operating parameters to the values before adjustment; For each air conditioning terminal, locate the most recent adjustment time that formed the current operating parameter in the historical operating parameter record, take the last valid operating parameter before the corresponding adjustment time as the previous value, and save the current operating parameter according to the air conditioning terminal identifier. When processing the current air conditioning terminal, replace the corresponding operating parameter with the previous value, and keep the current operating parameter for other air conditioning terminals; after completing the processing of the current air conditioning terminal, move on to the next air conditioning terminal.
[0042] The net heat change was re-predicted using the improved LTC, and the predicted net heat change before and after the recovery was compared to obtain the differences in net heat change in each controlled heat zone. A net heat change prediction is performed using the current operating parameters of all air conditioning terminals, and the prediction results for each controlled heat zone are saved. After restoring the current operating parameters of the air conditioning terminals, the improved LTC is run again using the same air-side liquid state, heat storage-side liquid state, IT load, and controlled heat zone temperature as the prediction starting point. The net heat change difference for the controlled heat zone corresponding to the two prediction results is formed according to the controlled heat zone and the prediction time. The net heat change difference is saved according to the direction of the predicted net heat change after restoration relative to the predicted net heat change before restoration. An increase in net heat input corresponds to an increase in heat extraction effect of the current air conditioning terminal relative to the previous value, and a decrease in net heat input corresponds to a decrease in heat extraction effect.
[0043] The corresponding operating parameters are restored to the saved values, and the net heat change difference is mapped to the current air conditioning terminal and the controlled heat zone to form the initial marginal cooling capacity.
[0044] After extracting the net heat change difference of the current air conditioning terminal, the corresponding operating parameters are restored to the saved current operating parameters before processing the next air conditioning terminal. Based on the saved net heat change differences for each air conditioning terminal and controlled heat zone, each net heat change difference is used as the initial marginal cooling capacity of the corresponding air conditioning terminal for subsequent heat conservation correction. The initial marginal cooling capacity adopts the same dimensions as the predicted net heat change, characterized by the change in net heat input per unit time.
[0045] In this embodiment, the formation of the cold energy attribution relationship includes the following steps: Based on the initial marginal cooling capacity of each air conditioning terminal in the controlled heat zone, obtain the corresponding predicted net heat change; The initial marginal cooling capacity of each air conditioning terminal is compiled according to the controlled heat zone and the forecast time, and the predicted net heat change of the same controlled heat zone at the same forecast time is obtained. The heat exchange per unit time flowing from adjacent controlled heat zones into the current controlled heat zone is recorded as heat input, and the heat exchange per unit time flowing out of the current controlled heat zone is recorded as heat output. The heat generation per unit time of the controlled heat zone formed by IT load and the net inflow heat exchange of adjacent controlled heat zones are summarized, and the net heat input per unit time corresponding to the predicted net heat change is deducted. The result is used as the heat output per unit time of the air conditioning terminal under the current operating parameters.
[0046] Based on the direction of heat loss change corresponding to the initial marginal cooling capacity, the air conditioning terminals in the same controlled heat zone and at the predicted time are respectively classified into a heat loss increase group and a heat loss decrease group. Keeping the IT load, controlled heat zone temperature, and heat exchange conditions of adjacent controlled heat zones constant, the current operating parameters of the air conditioning terminals within the same direction group are simultaneously replaced with their corresponding pre-control values, while other air conditioning terminals retain their current operating parameters. The heat loss per unit time before and after the replacement is determined using a thermodynamic mechanism model. The difference in heat loss per unit time before and after the replacement within the same direction group is used as the heat conservation boundary for that direction.
[0047] By performing thermal conservation matching between each initial marginal cooling capacity and the corresponding predicted net heat change, the common net heat change corresponding to different air conditioning terminals can be located. Within the same controlled heat zone and prediction time, the initial marginal cooling capacity amplitude is accumulated according to the direction of heat transfer change, and the accumulated result is compared with the corresponding heat conservation boundary. If the accumulated result does not exceed the corresponding heat conservation boundary, the original assignment is retained; if the accumulated result exceeds the corresponding heat conservation boundary, the excess portion is identified as duplicate assignment caused by different air conditioning terminals jointly corresponding to the same net heat change.
[0048] Remove the common corresponding portion from each initial marginal cooling capacity, and retain the marginal cooling capacity corresponding to a single air conditioning terminal; A historical operating period was selected in which only one air conditioning terminal's operating parameter changed while the operating parameters of other air conditioning terminals remained unchanged. The heat extraction change of the corresponding air conditioning terminal was determined based on the supply air volume and the temperature difference between the supply air temperature and return air temperature before and after the change, thus obtaining the net heat extraction change of each controlled heat zone during the same period. Thermodynamic mechanism models were used to deduct the impact of IT load changes and heat exchange changes between adjacent controlled heat zones on the net heat extraction change, yielding the heat extraction change of the corresponding air conditioning terminal acting on each controlled heat zone. The amplitude of the same-direction heat extraction change was calculated, and the heat extraction allocation ratio was determined according to the proportion of the heat extraction change amplitude of each controlled heat zone in the total same-direction heat extraction change amplitude of the corresponding air conditioning terminal. Based on the supply air volume and the temperature difference between the supply air temperature and return air temperature of the current air conditioning terminal under the current operating parameters and the previous control values, the heat extraction change and its direction relative to the previous control values were determined. A heat extraction allocation ratio consistent with the current heat extraction change direction was selected to allocate the current heat extraction change of the air conditioning terminal to the corresponding controlled heat zone.
[0049] For controlled heat zones with duplicate assignments, the duplicate portion is allocated according to the proportion of the heat transfer change amplitude of each air conditioning terminal to the current controlled heat zone in the total heat transfer change amplitude in the same direction. The duplicate portion is then deducted from the corresponding initial marginal cooling capacity amplitude, retaining the original heat transfer change direction. The processed marginal cooling capacity is retained according to the corresponding air conditioning terminal and controlled heat zone. The cumulative amplitude of the same heat transfer change direction in the same controlled heat zone is checked, and the cumulative amplitude does not exceed the heat conservation boundary of the corresponding direction.
[0050] Based on the retention results of air conditioning terminals and controlled hot zones, a cooling capacity attribution relationship is formed.
[0051] The marginal cooling capacity after repeated attribution is saved according to the air conditioning terminal, controlled heat zone, and prediction time, while the corresponding heat extraction change direction is also retained. The results generated by the same air conditioning terminal in different controlled heat zones are saved separately. The marginal cooling capacity and heat extraction change direction are associated with the correspondence between the air conditioning terminal and the controlled heat zone to form a cooling capacity attribution relationship, which is used to obtain the corresponding air conditioning terminal and marginal cooling capacity according to the controlled heat zone in the future.
[0052] In this embodiment, determining the cutting order includes the following steps: The controlled heat zone and marginal cooling capacity of each air conditioning terminal are obtained according to the cooling capacity attribution relationship. Using the air conditioning terminal as an index, the corresponding controlled heat zone, prediction time, and marginal cooling capacity are obtained from the cooling capacity attribution relationship. When an air conditioning terminal corresponds to different controlled heat zones, the prediction time and marginal cooling capacity corresponding to each controlled heat zone are retained respectively.
[0053] Compare the thermal safety margins of each controlled heat zone corresponding to the same air conditioning terminal to obtain the minimum thermal safety margin; In each controlled heat zone corresponding to the current air conditioning terminal, the thermal safety margin for the corresponding prediction time is obtained. When a controlled heat zone corresponds to consecutive prediction times, the thermal safety margin within the corresponding prediction period is first compared, and then compared with other controlled heat zones corresponding to the current air conditioning terminal. The thermal safety margin with the smallest value is retained, and the corresponding controlled heat zone and prediction time are saved.
[0054] Select air conditioning terminals to be reduced based on marginal cooling capacity and minimum thermal safety margin; Based on the minimum thermal safety margin and the equivalent heat capacity of the corresponding controlled heat zone, the amount of heat increase that can be tolerated during the corresponding forecast period is determined, and then converted into the allowable net heat input increase per unit time based on the corresponding forecast duration. According to the cooling capacity attribution relationship, the marginal cooling capacity of the current air conditioning terminal in the aforementioned controlled heat zone is obtained. The marginal cooling capacity amplitude is compared with the allowable net heat input increase per unit time to determine the output adjustment range of the current air conditioning terminal. If an output adjustment range that meets the thermal safety limit exists, the current air conditioning terminal is retained; otherwise, it is excluded.
[0055] The order of reduction is determined by combining the minimum thermal safety margin and operating power of each air conditioning terminal to be reduced.
[0056] The air conditioning terminals to be reduced are arranged in descending order of minimum thermal safety margin. If the minimum thermal safety margin values are the same, they are then arranged in descending order of operating power to obtain the reduction order for this round.
[0057] In this embodiment, the formation of the end-point control command includes the following steps: Determine the output adjustment amount of the current air conditioning terminal to be reduced according to the reduction order, and obtain the adjusted air conditioning terminal operating parameters; Select the historical operating range of the air conditioning terminal to be reduced, determine the heat taken per unit time based on the air volume and the temperature difference between the supply air temperature and the return air temperature corresponding to each combination of operating parameters, and obtain the corresponding operating power, and establish the correspondence between the combination of operating parameters and the heat taken per unit time and the operating power.
[0058] The heat output per unit time is determined based on the supply air volume and the temperature difference between the supply and return air temperatures corresponding to the current operating parameters. Using the current heat output per unit time as a benchmark, the heat output reduction per unit time for each combination of operating parameters is determined. According to the proportion of the marginal cooling capacity amplitude of the current air conditioning terminal in each controlled heat zone to the total marginal cooling capacity amplitude of the current air conditioning terminal in the same direction in the cooling capacity attribution relationship, the heat output reduction per unit time is allocated to the corresponding controlled heat zones. Operating parameter combinations whose heat output reduction per unit time allocated to the controlled heat zones corresponding to the minimum thermal safety margin does not exceed the corresponding allowable increase in net heat input per unit time are selected. The operating parameter combination with the lowest operating power is selected as the adjusted air conditioning terminal operating parameters. The output adjustment amount is determined based on the difference between the selected operating parameters and the current operating parameters.
[0059] Based on the adjusted air conditioning terminal operating parameters, the improved LTC is used to re-predict the net heat change of each controlled heat zone and update the heat safety margin. The improved LTC is operated using the adjusted air conditioning terminal operating parameters, current IT load, and controlled heat zone temperature as inputs, with the current air-side liquid state and heat storage-side liquid state as the starting point. Based on the re-obtained net heat change, the future temperature of each controlled heat zone is extrapolated, and a new thermal safety margin is formed based on the interval between the future temperature and the upper limit of the controlled heat zone temperature, replacing the corresponding result before this round of adjustments.
[0060] The operating parameters of the remaining air conditioning terminals to be reduced are restored one by one, the predicted net heat changes before and after restoration are compared, the initial marginal cooling capacity is reconstructed, and the cooling capacity attribution relationship is updated after heat conservation correction. When re-evaluating a remaining air conditioning terminal to be reduced, air conditioning terminals that have already undergone parameter adjustments in this round retain their adjusted operating parameters, while other air conditioning terminals retain their current operating parameters. The operating parameters of the currently evaluated air conditioning terminal are saved, and only the corresponding operating parameters are restored to their pre-adjustment values. Predictions are made using the same IT load, controlled heat zone temperature, prediction time, air-side liquid state, and heat storage-side liquid state before and after restoration. Based on the difference in net heat change between the two predictions, the initial marginal cooling capacity of the current air conditioning terminal in each controlled heat zone is recalculated. After completing the evaluation of the current air conditioning terminal, the corresponding operating parameters are restored to the saved values before evaluating the next remaining air conditioning terminal to be reduced.
[0061] After all remaining air conditioning terminals to be reduced are re-evaluated, the initial marginal cooling capacity formed in this round replaces the results of the previous round. When redefining the heat conservation boundary, the air conditioning terminals that have completed parameter adjustments in this round will use the adjusted operating parameters. According to the direction of heat extraction change corresponding to the initial marginal cooling capacity formed in this round, the remaining air conditioning terminals to be reduced will be grouped. The current operating parameters of the air conditioning terminals in the same direction group will be simultaneously replaced with the corresponding values before adjustment. Other air conditioning terminals will retain the operating parameters of this round. The heat extraction per unit time of the same controlled heat zone before and after replacement will be compared to form the heat conservation boundary for the current round in the corresponding direction. Based on the heat conservation boundary of the current round, the net heat change corresponding to different air conditioning terminals will be re-identified. The duplicate attribution will be deducted, and the updated cooling capacity attribution relationship will be formed according to the marginal cooling capacity saved and processed by the air conditioning terminal, the controlled heat zone, and the prediction time.
[0062] Based on the updated thermal safety margin and cooling capacity attribution relationship, the reduction sequence of the remaining air conditioning terminals to be reduced is re-determined, and the terminal control instructions are formed by summarizing the output adjustments of each time.
[0063] Based on the updated cooling capacity attribution relationship, the controlled heat zone and marginal cooling capacity corresponding to the remaining air conditioning terminals to be reduced are re-obtained, and the reduction space is re-evaluated using the updated thermal safety margin. For air conditioning terminals that still have reduction space, the minimum thermal safety margin and operating power are re-compared to form the subsequent reduction sequence. If there are remaining air conditioning terminals to be reduced with reduction space, the next round of output adjustment determination, net heat change prediction, thermal safety margin update, and cooling capacity attribution relationship update is carried out according to the subsequent reduction sequence; if there are no remaining air conditioning terminals to be reduced with reduction space, the successive adjustment ends.
[0064] After each adjustment is completed, the results of each adjustment of the operating parameters of the air conditioning terminal are summarized. The operating parameters after the last adjustment are used as the control value for the same air conditioning terminal, and the terminal control command is formed by combining them.
[0065] In this embodiment, the updating of the two types of liquid states and the liquid time constant includes the following steps: Execute the terminal control command and collect the temperature of the controlled thermal zone after control. Using the moment when the terminal control command begins execution as the feedback starting point, the controlled heat zone temperature after control is obtained according to the original acquisition cycle of the controlled heat zone. Based on the actual change time of the air conditioning terminal operating parameters, a time correspondence is established between the controlled air conditioning terminal operating parameters and the corresponding controlled heat zone temperature.
[0066] The controlled hot zone temperature before the execution of the terminal control command is taken as the starting point for the deduction. The controlled hot zone temperature is deduced based on the corresponding IT load and the air conditioning terminal operating parameters after the control. The temperature progression starts at the last controlled hot zone temperature acquisition time before the execution of the terminal control command, and ends at the acquisition time corresponding to the controlled hot zone temperature after control. The heat output per unit time of the controlled hot zone is determined based on the IT load of the corresponding operating interval. The heat extraction per unit time of the air conditioning terminal is determined based on the controlled air volume and the temperature difference between the supply and return air temperatures. The temperature is then continuously advanced by combining the heat exchange per unit time of adjacent controlled hot zones, the equivalent heat capacity of the controlled hot zone, and the length of the operating interval. When the operating parameters of the air conditioning terminal change during the feedback period, the operating interval is divided based on the time of parameter change, and the corresponding parameters for each operating interval are used to continuously advance to the controlled hot zone temperature acquisition time after control.
[0067] By comparing the predicted temperature of the controlled thermal zone with the temperature of the controlled thermal zone after adjustment, a new mismatch in the thermal mechanism is formed; The newly formed thermal mechanism mismatch is saved according to the temperature acquisition time of the controlled thermal zone and the temperature of the controlled thermal zone after regulation. The corresponding acquisition time is used as the sequence position of the thermal mechanism mismatch, so that the inferred temperature of the controlled thermal zone, the temperature of the controlled thermal zone after regulation and the thermal mechanism mismatch are at the same acquisition time.
[0068] The newly formed thermal mechanism mismatch is integrated into the existing thermal mechanism mismatch sequence, and adjacent changes and continuous unidirectional changes are updated. The liquid time constant on the air side and the liquid time constant on the heat storage side are adjusted respectively, and the corresponding liquid states are updated.
[0069] The newly formed thermal mechanism mismatch is appended to the end of the thermal mechanism mismatch sequence within the same controlled thermal region. The difference between the newly formed thermal mechanism mismatch and the thermal mechanism mismatch at the previous acquisition time is calculated and added as a new adjacent change. When the thermal mechanism mismatches at two acquisition times maintain the same positive and negative directions, the amplitude of the newly formed thermal mechanism mismatch is added to the current continuous same-direction change; when the positive and negative directions change, the current accumulation ends, and a new continuous same-direction change begins from the newly formed thermal mechanism mismatch.
[0070] The air-side liquid time constant is obtained based on the fitting relationship between the updated adjacent change amplitude and the temperature response duration. The heat storage-side liquid time constant is obtained based on the fitting relationship between the updated cumulative degree of continuous unidirectional change and the sustained thermal response duration. The updated air-side liquid time constant is substituted into the continuous-time update process of the air-side liquid state, and the air-side liquid state is updated using the adjusted air conditioning terminal operating parameters. The updated air-side liquid state is compared with the air-side liquid state at the previous acquisition time to obtain the corresponding state change. The residual thermal state on the air side is reconstructed according to the cancellation relationship between the heat extraction state response and the state change, and then transformed to the corresponding dimension of the heat storage-side liquid state through a state mapping matrix.
[0071] The updated liquid time constant on the thermal storage side is substituted into the continuous time update process of the liquid state on the thermal storage side. The liquid state on the thermal storage side is updated using the IT load of the corresponding operating range, the controlled heat zone temperature after adjustment, and the converted residual heat state on the air side. The updated liquid state on the air side and the liquid state on the thermal storage side are used as the starting point of the state at the next acquisition time, and the updated liquid time constant on the air side and the liquid time constant on the thermal storage side are used for continuous state updates at the next acquisition time.
[0072] Example 1: To verify the feasibility of this invention in practice, it was applied to an energy-saving control scenario for the air conditioning terminals in a data center computer room. For example... Figure 2 As shown, server racks in the data center are arranged according to the supply and return air direction. Adjacent rack areas are divided into controlled heat zones A, B, and C, with temperature acquisition points set up in each zone. Air conditioning terminals A, B, and C supply air to their respective zones. The effective ranges of different air conditioning terminals overlap, meaning a single controlled heat zone can receive combined heat from various terminals. During operation, the temperature of the controlled heat zones, IT load, air conditioning terminal operating parameters, and operating power are collected, and a corresponding relationship is established based on the time of data collection.
[0073] The continuously collected temperatures of controlled heat zones are arranged, with the temperature at the previous collection time serving as the starting point for the thermodynamic mechanism model. Corresponding IT loads and air conditioning terminal operating parameters are incorporated into the heat calculation. The heating effect of the controlled heat zones is determined based on the IT load, and the heat extraction effect is determined based on the air conditioning terminal operating parameters. The temperature at the next collection time is then extrapolated by combining the heat transfer between adjacent controlled heat zones. The extrapolated temperature is compared with the actual temperature at the next collection time. The difference between the two is used to identify thermal mechanism mismatches, which are continuously saved according to the collection time to reflect the deviation between the calculated mechanism results and the actual thermal response of the computer room.
[0074] An improved LTC (Liquid Temperature Coefficient) system is used to process data from the data center. In the improved LTC, the air-side liquid state handles the rapid thermal response caused by changes in air conditioning terminal operating parameters, while the heat storage-side liquid state handles the slow thermal response caused by IT load, controlled heat zone temperature, and continuous heat accumulation. Each time new operating data is acquired, the air-side liquid state is updated based on the current air conditioning terminal operating parameters, and the heat storage-side liquid state is updated based on the current IT load and controlled heat zone temperature. The portion of the air-side liquid state change that is not offset by the current heat extraction action forms the air-side residual thermal state, which is transferred to the heat storage-side liquid state according to the existing state mapping relationship. The persistent thermal impact after the air conditioning terminal operation ends can be included in subsequent temperature prediction.
[0075] Based on the continuously formed thermal mechanism mismatch, adjacent changes and continuous unidirectional changes are extracted. The liquid time constant on the air side is adjusted according to adjacent changes, and the liquid time constant on the heat storage side is adjusted according to continuous unidirectional changes. The corresponding liquid states are updated using the adjusted two types of liquid time constants to predict the net heat change of each controlled heat zone. Combined with the current temperature of the controlled heat zone, the future temperature within the prediction period is further extrapolated, and a thermal safety margin is formed by the interval between the future temperature and the upper limit of the controlled heat zone temperature. When heat continues to accumulate in a controlled heat zone, the state on the heat storage side can retain this continuous change, avoiding the situation where it is judged that the controlled heat zone still has a large reduction space based solely on the current temperature.
[0076] Subsequently, the actual heat extraction effect of each air conditioning terminal was evaluated one by one. After selecting an air conditioning terminal, its current operating parameters were saved, and then the corresponding operating parameters were restored to the effective values before the most recent adjustment. Other air conditioning terminals maintained their current states. An improved LTC was used to predict the net heat change of each controlled heat zone before and after restoration, and the difference between the two was mapped to the current air conditioning terminal and the controlled heat zone to form the initial marginal cooling capacity. After the evaluation was completed, the operating parameters of the current air conditioning terminal were restored to the saved values before processing the next air conditioning terminal, thereby avoiding interference from the marginal effect judgment caused by the simultaneous changes of different air conditioning terminals.
[0077] For example, the current temperature of a controlled heat zone is 24.6℃, with a maximum temperature of 26.0℃. The improved LTC predicts that the highest future temperature of this controlled heat zone will be 25.3℃, corresponding to a thermal safety margin of 0.7℃. After restoring air conditioning terminal A to its pre-control value, the predicted net heat change in this controlled heat zone increases by 2.1kW, and after restoring air conditioning terminal B to its pre-control value, it increases by 1.4kW, with a cumulative initial marginal cooling capacity of 3.5kW. The heat conservation boundary formed in the same direction based on the current operating state is 2.9kW, therefore, 0.6kW belongs to the overlapping portion formed by different air conditioning terminals corresponding to the same net heat change. After completing the heat conservation correction, the marginal cooling capacities retained by air conditioning terminals A and B in this controlled heat zone are 1.8kW and 1.1kW, respectively, and the cumulative results are consistent with the corresponding heat conservation boundary.
[0078] Because different air conditioning terminals can collectively affect the same controlled heat zone, the initial marginal cooling capacity cannot be directly considered as an independent heat extraction contribution. For example... Figure 3 As shown, heat conservation matching is performed according to the direction of heat extraction change corresponding to the initial marginal cooling capacity, identifying the net heat change common to different air conditioning terminals within the same controlled heat zone. When the cumulative result of the initial marginal cooling capacity of each air conditioning terminal exceeds the heat conservation boundary in the corresponding direction, the excess portion is identified as a duplicate attribution and deducted from the corresponding initial marginal cooling capacity according to the heat extraction distribution relationship in the same direction. After correction, the actual marginal cooling capacity and direction of change corresponding to each air conditioning terminal are retained, and a cooling capacity attribution relationship is formed according to the air conditioning terminal and the controlled heat zone. This enables... Figure 3The duplicate attribution portion in the data is corrected and converted into the retained attribution cooling capacity corresponding to each air conditioning terminal.
[0079] After establishing the cooling capacity attribution relationship, the controlled heat zone and marginal cooling capacity corresponding to each air conditioning terminal are obtained, and the thermal safety margin of the corresponding controlled heat zone is compared. For air conditioning terminals that simultaneously affect different controlled heat zones, the smallest thermal safety margin is used as the criterion for reduction. Based on the heat capacity of the controlled heat zone corresponding to this thermal safety margin and the predicted time period, the range of net heat change increase that can be tolerated is determined. Air conditioning terminals whose marginal cooling capacity can meet the constraints of this range are included in the reduction range. Then, the reduction order is determined by combining the thermal safety margin and operating power, with air conditioning terminals with larger thermal safety margins being given priority for subsequent adjustment.
[0080] When processing the current air conditioning terminals to be reduced according to the reduction order, the combination of operating parameters that can reduce operating power is selected from the existing operating records, and the heat reduction amount relative to the current operating state is calculated. Based on the current cooling capacity attribution, this heat reduction amount is allocated to the corresponding controlled heat zone. Only when the heat reduction amount allocated to the controlled heat zone corresponding to the minimum thermal safety margin does not exceed the range of acceptable net heat change increase is the corresponding operating parameter combination retained. Then, the state with lower operating power is selected from the operating parameter combinations that meet the thermal safety limits, and the output adjustment amount is formed by the difference between the current operating parameters and the selected operating parameters.
[0081] After determining the output adjustment amount for each air conditioning terminal, the adjusted operating parameters are input into the improved LTC (Light Temperature Control) system to re-predict the net heat change and future temperature of each controlled heat zone, thus re-establishing the thermal safety margin. Air conditioning terminals that have already undergone adjustment maintain their adjusted operating state. For the remaining air conditioning terminals to be reduced, the operating parameter recovery and net heat change comparison are re-executed to re-establish the initial marginal cooling capacity. Heat conservation correction is then performed again to update the cooling capacity attribution relationship. Based on the updated thermal safety margin and cooling capacity attribution relationship, the reduction order of the remaining air conditioning terminals is rearranged, and the output adjustment amount for the next air conditioning terminal is determined. Reduction stops when there is no subsequent adjustment state that meets the thermal safety limit. The final determined operating parameters of each air conditioning terminal are summarized to form the terminal control command.
[0082] After executing the terminal control command, the temperature of each controlled hot zone continues to be collected. Using the temperature of the controlled hot zone before control as the starting point for deduction, the temperature of the controlled hot zone is re-deduced based on the corresponding IT load and the operating parameters of the air conditioning terminal after control. The deduced temperature is compared with the actual controlled hot zone temperature after control to re-establish the thermal mechanism mismatch, which is then integrated into the existing thermal mechanism mismatch sequence. The liquid time constant on the air side and the liquid time constant on the heat storage side are adjusted according to the updated adjacent changes and continuous unidirectional changes, respectively. Simultaneously, the two types of liquid states are updated based on the operating parameters of the air conditioning terminal after control, the IT load, and the temperature of the controlled hot zone. The temperature response generated by this actual control continues to participate in the next round of data center air conditioning terminal control.
[0083] To compare the application effects of different control methods under the same computer room operating conditions, the results are shown in Table 1.
[0084] Table 1 Comparison of the application effects of terminal control methods for computer room air conditioning
[0085] As shown in Table 1, the present invention outperforms both the terminal control method based on a fixed temperature threshold and the terminal control method based on historical trend prediction in terms of controlled heat zone temperature control stability, terminal regulation effectiveness, and cooling energy saving effect. The average temperature deviation of the controlled heat zone in the present invention is 0.32℃, which is 0.24℃ and 0.11℃ lower than the two comparative methods, respectively. The over-temperature occurrence rate of the controlled heat zone is 2.4%, which is reduced by 3.4 percentage points and 1.5 percentage points, respectively. The above results indicate that by adjusting the liquid time constant on the air side and the liquid time constant on the heat storage side through thermal mechanism mismatch, and using the two types of liquid states to jointly predict the future temperature of the controlled heat zone, the rapid heat extraction effect of the air conditioning terminal and the continuous heat storage effect of the controlled heat zone can be more accurately reflected, thereby reducing the over-temperature risk caused by temperature prediction deviation.
[0086] Regarding stability, the temperature fluctuation amplitude after regulation in this invention is 0.41℃, lower than the 0.67℃ and 0.53℃ of the two comparative methods. This is because when determining the reduction sequence and output adjustment amount of the air conditioning terminals, this invention not only makes judgments based on the current thermal safety margin, but also re-predicts the net heat change after each adjustment of the air conditioning terminal operating parameters, updating the thermal safety margin and cooling capacity attribution relationship. Subsequent air conditioning terminal adjustments can adapt to the thermal state changes caused by the previous adjustment, reducing excessive temperature fluctuations during continuous reduction.
[0087] Regarding the effectiveness of terminal regulation and energy saving, the ineffective terminal regulation rate of this invention is 3.9%, which is 4.8 percentage points and 2.3 percentage points lower than the two comparative methods, respectively. Simultaneously, the cooling power reduction rate reaches 10.6%. This result corresponds to the invention's method of restoring the initial marginal cooling capacity by reconstructing the operating parameters of each air conditioning terminal and using heat conservation correction to eliminate duplicate attributions of the same net heat change by different air conditioning terminals. By forming a more accurate cooling capacity attribution relationship, ineffective regulation by air conditioning terminals that contribute little to actual heat extraction can be reduced. Under the condition of meeting the thermal safety constraints of the controlled heat zone, priority can be given to reducing air conditioning terminals with energy-saving potential, thereby balancing the safety of the computer room temperature and the reduction of cooling power.
[0088] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A terminal control method for computer room air conditioning based on deep learning and thermodynamic mechanism models, characterized in that, Includes the following steps: Collect controlled hot zone temperature, IT load, air conditioning terminal operating parameters and operating power, and use thermodynamic mechanism model combined with temperature changes to form thermal mechanism mismatch; An improved LTC is constructed, which divides the liquid state into air-side liquid state and heat storage-side liquid state. The air-side liquid state is updated according to the operating parameters of the air conditioning terminal, and the heat storage-side liquid state is updated according to the IT load and the temperature of the controlled hot zone. The residual heat state on the air side is transferred to the heat storage-side liquid state. Based on the thermal mechanism mismatch adjustment of the two types of liquid time constants, the net heat change and future temperature of each controlled heat zone are predicted to form a thermal safety margin; The operating parameters of each air conditioning terminal are restored to their pre-control values, and the predicted net heat change before and after restoration is compared to form the initial marginal cooling capacity. The initial marginal cooling capacity is corrected for thermal conservation based on the thermodynamic mechanism model, eliminating duplicate attribution of the same net heat change and forming a cooling capacity attribution relationship; Based on the cooling capacity attribution relationship, the air conditioning terminals to be reduced are screened, and the reduction sequence is determined by combining the thermal safety margin and operating power. The output adjustment amount of the air conditioning terminals is determined one by one to form the adjusted air conditioning terminal operating parameters. After each formation, the net heat change is re-predicted, the cooling capacity attribution relationship is updated, and the terminal control command is formed. Execute the end-point control command, and re-establish the thermal mechanism mismatch based on the controlled thermal zone temperature after control, updating the two types of liquid states and the liquid time constant.
2. The data center air conditioning terminal control method based on deep learning and thermodynamic mechanism model according to claim 1, characterized in that, The formation of the aforementioned thermal mechanism mismatch includes the following steps: Arrange the controlled thermal zone temperatures according to the acquisition time, and select adjacent acquisition times; Starting from the controlled hot zone temperature at the previous acquisition time, the corresponding IT load and air conditioning terminal operating parameters are input into the thermodynamic mechanism model to extrapolate the controlled hot zone temperature at the next acquisition time. The predicted temperature of the controlled thermal zone is compared with the temperature of the controlled thermal zone at the next data acquisition time, resulting in a mismatch in thermal mechanism.
3. The data center air conditioning terminal control method based on deep learning and thermodynamic mechanism model according to claim 1, characterized in that, The construction of the improved LTC includes the following steps: The continuous liquid state of LTC is divided into air-side liquid state and heat storage-side liquid state, and corresponding liquid time constants are established for each. The air-side liquid state is updated based on the air conditioning terminal operating parameters, and the heat storage-side liquid state is updated based on the IT load and the temperature of the controlled hot zone. The state change that is not offset by the current heat extraction after the air-side liquid state is updated is retained as the air-side residual heat state and transferred to the heat storage side liquid state.
4. The data center air conditioning terminal control method based on deep learning and thermodynamic mechanism model according to claim 3, characterized in that, The residual heat state on the air side is obtained by filtering out the portion of the liquid state on the air side that is not offset by the heat extraction effect of the corresponding air conditioning terminal from the changes in the liquid state on the air side at adjacent acquisition times. The residual heat state on the air side is mapped according to the state dimension of the liquid state on the heat storage side. The liquid state on the heat storage side at the next acquisition time is updated based on the liquid state on the heat storage side at the previous acquisition time and the mapping result.
5. The data center air conditioning terminal control method based on deep learning and thermodynamic mechanism model according to claim 1, characterized in that, The formation of the thermal safety margin includes the following steps: Arrange thermal mechanism mismatches according to the time of data collection, and extract adjacent changes and continuous changes in the same direction; The liquid time constant on the air side is adjusted based on adjacent changes, and the liquid time constant on the heat storage side is adjusted based on continuous changes in the same direction. The corresponding liquid state is updated using two types of adjusted liquid time constants to predict the net heat change of each controlled heat zone; Based on net heat change and controlled heat zone temperature, future temperatures are projected, and a thermal safety margin is formed by the interval between the future temperature and the upper limit of the controlled heat zone temperature.
6. The data center air conditioning terminal control method based on deep learning and thermodynamic mechanism model according to claim 1, characterized in that, The formation of the initial marginal coldness includes the following steps: Select each air conditioning terminal one by one, save the corresponding operating parameters, and restore the corresponding operating parameters to the values before adjustment; The net heat change was re-predicted using the improved LTC, and the predicted net heat change before and after the recovery was compared to obtain the differences in net heat change in each controlled heat zone. The corresponding operating parameters are restored to the saved values, and the net heat change difference is mapped to the current air conditioning terminal and the controlled heat zone to form the initial marginal cooling capacity.
7. The data center air conditioning terminal control method based on deep learning and thermodynamic mechanism model according to claim 1, characterized in that, The formation of the cold energy attribution relationship includes the following steps: Based on the initial marginal cooling capacity of each air conditioning terminal in the controlled heat zone, obtain the corresponding predicted net heat change; By performing thermal conservation matching between each initial marginal cooling capacity and the corresponding predicted net heat change, the common net heat change corresponding to different air conditioning terminals can be located. Remove the common corresponding portion from each initial marginal cooling capacity, and retain the marginal cooling capacity corresponding to a single air conditioning terminal; Based on the retention results of air conditioning terminals and controlled hot zones, a cooling capacity attribution relationship is formed.
8. The data center air conditioning terminal control method based on deep learning and thermodynamic mechanism model according to claim 1, characterized in that, Determining the cutting order includes the following steps: The controlled heat zone and marginal cooling capacity of each air conditioning terminal are obtained according to the cooling capacity attribution relationship. Compare the thermal safety margins of each controlled heat zone corresponding to the same air conditioning terminal to obtain the minimum thermal safety margin; Select air conditioning terminals to be reduced based on marginal cooling capacity and minimum thermal safety margin; The order of reduction is determined by combining the minimum thermal safety margin and operating power of each air conditioning terminal to be reduced.
9. The data center air conditioning terminal control method based on deep learning and thermodynamic mechanism model according to claim 8, characterized in that, The generation of the terminal control command includes the following steps: Determine the output adjustment amount of the current air conditioning terminal to be reduced according to the reduction order, and obtain the adjusted air conditioning terminal operating parameters; Based on the adjusted air conditioning terminal operating parameters, the improved LTC is used to re-predict the net heat change of each controlled heat zone and update the heat safety margin. The operating parameters of the remaining air conditioning terminals to be reduced are restored one by one, the predicted net heat changes before and after restoration are compared, the initial marginal cooling capacity is reconstructed, and the cooling capacity attribution relationship is updated after heat conservation correction. Based on the updated thermal safety margin and cooling capacity attribution relationship, the reduction sequence of the remaining air conditioning terminals to be reduced is re-determined, and the terminal control instructions are formed by summarizing the output adjustments of each time.
10. The data center air conditioning terminal control method based on deep learning and thermodynamic mechanism model according to claim 1, characterized in that, The updating of the two types of liquid states and the liquid time constant includes the following steps: Execute the terminal control command and collect the temperature of the controlled thermal zone after control. The controlled hot zone temperature before the execution of the terminal control command is taken as the starting point for the deduction. The controlled hot zone temperature is deduced based on the corresponding IT load and the air conditioning terminal operating parameters after the control. By comparing the predicted temperature of the controlled thermal zone with the temperature of the controlled thermal zone after adjustment, a new mismatch in the thermal mechanism is formed; The newly formed thermal mechanism mismatch is integrated into the existing thermal mechanism mismatch sequence, and adjacent changes and continuous unidirectional changes are updated. The liquid time constant on the air side and the liquid time constant on the heat storage side are adjusted respectively, and the corresponding liquid states are updated.
