Efficient machine room energy-saving control system based on global optimization

Abstract

The invention discloses an efficient machine room energy-saving control system based on global optimization, which belongs to the technical field of temperature control and energy conservation of central air-conditioning machine rooms and comprises a multi-source data acquisition module, a refrigeration optimization decision module, an external temperature interactive scheduling module, a control module and a feedback module. The multi-source data acquisition module is used for collecting and fusing multi-source data in an all-weather manner, including real-time power consumption and operation states of power equipment and refrigeration equipment in a machine room, temperatures of all areas of the machine room, and external environment temperature and humidity and weather forecast data acquired through an API; according to the efficient machine room energy-saving control system based on global optimization, on-demand allocation of refrigeration equipment is achieved, a natural cold source is fully utilized, invalid consumption of refrigeration energy consumption is reduced from the source, temperature control accuracy and energy saving performance of a machine room are improved, the system has the self-optimization capacity, long-term stable and efficient operation of the machine room is guaranteed, and the service life of the system is prolonged. The problems that a traditional machine room is extensive in refrigeration control and serious in energy waste are effectively solved.

Description

Technical Field

[0001] This invention belongs to the field of central air conditioning room temperature control and energy-saving technology, specifically a high-efficiency energy-saving control system for computer rooms based on global optimization. Background Technology

[0002] As a core supporting facility of the central air conditioning system, the central air conditioning computer room continuously generates heat during the operation of its internal power equipment, requiring refrigeration equipment to regulate the temperature to ensure stable system operation. However, current computer room refrigeration control mostly adopts a traditional, extensive mode, lacking a global optimization design, resulting in significant energy waste. Traditional control methods cannot accurately predict the computer room temperature rise curve based on the actual number of power equipment in operation and its heating characteristics. The start-up, shutdown, and operating parameter settings of refrigeration equipment lack scientific basis, easily leading to situations where a large number of refrigeration equipment are started even when only some power equipment is in operation, resulting in ineffective consumption of refrigeration energy. At the same time, traditional control does not fully utilize the cooling factors of the external natural environment and ignores the energy-saving potential of natural cold sources, often causing refrigeration equipment to operate excessively, resulting in high overall refrigeration energy consumption in the computer room. Existing control methods, due to the lack of comprehensive consideration of internal and external factors of the computer room, make it difficult to achieve a balance between temperature control requirements and energy-saving goals. Therefore, there is an urgent need to develop a high-efficiency computer room energy-saving control system based on global optimization to solve the above-mentioned industry pain points. Summary of the Invention

[0003] This invention aims to address at least one of the technical problems existing in the prior art. To this end, this invention proposes a high-efficiency energy-saving control system for data centers based on global optimization, enabling relevant personnel to easily address issues such as high cooling energy consumption and inaccurate temperature control strategies in data centers.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: A high-efficiency data center energy-saving control system based on global optimization includes a multi-source data acquisition module, a cooling optimization decision module, an external temperature interactive scheduling module, a control module, and a feedback module. The multi-source data acquisition module is used to collect and integrate multi-source data around the clock, including real-time power consumption and operating status of power and cooling equipment inside the computer room, temperature in various areas of the computer room, and external environmental temperature and humidity and weather forecast data obtained through API. The cooling optimization decision module predicts the temperature rise curve of the computer room based on historical data and establishes the performance model of each cooling device. With the upper limit of the computer room temperature as a constraint and the minimum total cooling energy consumption of the system as the goal, the module solves the optimal collaborative control strategy of the cooling devices in real time through optimization algorithms to achieve a dynamic balance of low energy consumption to offset the temperature rise. The external temperature interactive scheduling module accesses and processes external ambient temperature and weather forecast data, quantifies and analyzes the rate and magnitude of their impact on the temperature inside the computer room, and, combined with the start-up patterns of power equipment in different time periods in historical data, predicts the specific time periods during which continuous external cooling can have an effective impact on the inside of the computer room, and determines in advance the number of cooling equipment that can be adjusted during that time period. The control module is used to convert the generated policy instructions into commands that can be executed by specific devices; The feedback module continuously collects actual operating data after the execution of the control strategy and evaluates the effect, feeding it back to the cooling optimization decision module and the external temperature interaction scheduling module.

[0005] Furthermore, the multi-source data acquisition module is used to collect and fuse multi-source data around the clock, including real-time power consumption and operating status of the power and cooling equipment inside the computer room, temperature in various areas of the computer room, and external environmental temperature and humidity and weather forecast data obtained through API. The processing procedure is as follows: Through various sensor networks deployed inside the computer room, the system monitors and collects the operating data of various power and refrigeration equipment inside the computer room in real time around the clock, continuously acquiring parameters such as real-time power consumption, operating frequency, and start / stop status of each chiller unit, water pump, and cooling tower fan; as well as acquiring temperature sensor data from different functional areas within the computer room. External environmental data is obtained regularly from the meteorological service platform, including real-time outdoor temperature and humidity, wind speed and direction, and weather forecast information for the next 12 hours. The system aligns and integrates two types of data streams—internal sensors and external APIs—according to a unified timestamp, and automatically performs data cleaning to remove outliers, fill in reasonable missing values, and standardize data of different formats and dimensions into a unified format that can be processed within the system, forming a complete global data view with timestamps.

[0006] Furthermore, the cooling optimization decision module predicts the computer room temperature rise curve based on historical data and establishes performance models for each cooling device. With the upper limit of the computer room temperature safety as a constraint and the minimum total cooling energy consumption as the objective, it uses an optimization algorithm to solve for the optimal collaborative control strategy of the cooling devices in real time. The dynamic balance process of offsetting temperature rise with low energy consumption is as follows: It receives and integrates current data from multi-source data acquisition modules and similar scenario data from historical databases in real time, and uses machine learning algorithms to predict the temperature rise trend and speed of the computer room in the future under the condition of heat generation from existing power equipment operation, forming a dynamic computer room temperature rise curve. Refrigeration equipment performance models are established by combining raw performance data from refrigeration equipment manufacturers with historical data from long-term operation. These models accurately describe the relationship between the cooling capacity and real-time energy consumption of each chiller, precision air conditioner, and other equipment under different load rates and set parameters; The computer room temperature rise curve is coupled with the performance model of each cooling device. The upper limit of the temperature that the computer room must maintain is used as the core constraint, and the minimum total power consumption of the entire cooling system is set as the optimization objective. An optimization problem is constructed and solved by an optimization algorithm to find the combination of cooling devices that should be started in the future and the setting parameters of each operating device should be adjusted.

[0007] Furthermore, the process involves coupling the computer room temperature rise curve with the performance models of each cooling device, using the upper limit of the safe temperature that the computer room must maintain as the core constraint, and setting the minimum total power consumption of the entire cooling system as the optimization objective. An optimization problem is then constructed and solved using an optimization algorithm to determine the optimal combination of cooling devices to be activated in the future, and the appropriate adjustment of the set parameters for each operating device. The process is as follows: By integrating current equipment operation data with historical data, a predictive model is used to calculate the precise trajectory of the computer room temperature rise over time under the current heat load in the future, i.e., the computer room temperature rise curve. The built-in performance model of the cooling equipment is invoked, and the predicted temperature rise curve is dynamically correlated with the performance models of all available cooling equipment. The requirement that the computer room temperature must be maintained below the safe upper limit is transformed into an inequality constraint for the optimization problem. With the goal of minimizing the sum of the total power consumption of all possible cooling equipment during the prediction period, a mixed integer optimization problem containing continuous variables (setting parameters) and discrete variables (start and stop states) is constructed. Then, the optimal decision sequence in each future control cycle is obtained through a real-time optimization solution algorithm. The objective function is modeled as follows: ; The constraints that need to be met include dynamic thermal equilibrium constraints, temperature safety constraints, and equipment operation constraints. Among them, the dynamic constraint of thermal equilibrium is: ;

[0008] Temperature safety constraints include: ; Among them, equipment operation constraints: ; in, It is the total energy consumption over the predicted time. To predict the number of time-domain steps, This represents the total number of refrigeration equipment. It is equipment At any moment power consumption, It is a moment Predict the temperature of the computer room. It is equipment In setting The cooling capacity below, This is a binary decision variable for starting and stopping the device; a value of 0 indicates it is off, and a value of 1 indicates it is on. It controls the step length. For the equivalent heat capacity of the computer room, For the first The total heat output of all power equipment in the computer room at each prediction step size. The maximum allowable temperature for the safe operation of computer room equipment. Each refrigeration unit Its adjustable physical lower and upper limits for set parameters; Furthermore, the external temperature interaction scheduling module accesses and processes external ambient temperature and weather forecast data, quantitatively analyzes the rate and magnitude of their impact on the internal temperature of the computer room, and, combined with historical data on the start-up patterns of power equipment at different time periods, predicts the specific time periods during which continuous external cooling can effectively affect the internal temperature of the computer room, and determines in advance the number of adjustable cooling equipment during that time period. The processing procedure is as follows: Through the pre-built application programming interface, it continuously accesses and processes real-time monitoring data and short-term forecast data from the meteorological department, including outdoor dry-bulb temperature, wet-bulb temperature, wind speed and temperature change curves for the next 12 hours, forming a structured external environment data sequence. A heat transfer quantification model is trained based on the thermal performance of the data center envelope and historical data. The external temperature data is converted into a function that directly affects the internal temperature of the data center. The model outputs the rate and potential magnitude of temperature drop in the data center caused by the external environment per unit time. Synchronously retrieve historical databases, analyze the start-up patterns and power consumption curves of power equipment in the computer room under different date types and time periods, establish a time series model of internal heat load for typical days, and clarify the pattern of heat generation in the computer room itself changing over time; The external temperature drop influence function is coupled with the internal heat load time series model for calculation. Through simulation comparison, it is predicted which consecutive periods of external natural cooling capacity will be sufficient to offset or significantly reduce the net heat load in the computer room. The specific time windows for actively utilizing natural cold sources are identified, and a natural cold source utilization time table is obtained. Based on the predicted available time period of natural cold source, and further based on the estimated cooling output of natural cooling during that time period and the performance characteristics of the refrigeration equipment, the specific number, type and capacity of mechanical refrigeration equipment that can be safely shut down or derated are calculated to form a pre-adjustment strategy. The generated natural cold source utilization schedule and corresponding refrigeration equipment adjustment suggestions are used as a forward-looking constraint and optimization target and transmitted to the control module in real time. This allows the system to plan ahead and seamlessly integrate with natural cooling solutions when formulating global control strategies, thereby minimizing the total energy consumption of the system.

[0009] Furthermore, the heat transfer quantification model trained based on the thermal performance of the data center envelope and historical data is used to calculate and transform the external temperature data into a function that directly affects the internal temperature of the data center. The processing steps for outputting the rate and potential magnitude of temperature drop in the data center caused by the external environment per unit time are as follows: Based on the building drawings and material parameters of the computer room, a thermal physical model of its enclosure structure, including walls and roof, is established to describe the theoretical relationship between outdoor and indoor temperatures. By calling up long-term stored historical data, including the outdoor temperature series of the same period and the corresponding indoor temperature change series of the computer room, the physical model is aligned with these actual data on the time axis. Then, the parameter estimation method is used to back-infer and calibrate the comprehensive heat transfer coefficient and effective heat capacity of the room in the physical model using historical data, so that the output of the model achieves the best fit with the historical real temperature change curve. The calibrated physical model with defined parameters is encapsulated into a computable influence function. This function takes real-time outdoor temperature data as input and, by solving the heat balance equation, directly outputs the rate of temperature drop in the computer room per unit time under current and predicted external temperature conditions, as well as the final possible steady-state temperature difference.

[0010] The quantitative model, trained and validated with data, becomes a reliable prediction tool that can translate external meteorological conditions into specific effects on the internal temperature of the computer room in real time, providing accurate quantitative basis for decisions on the utilization of natural cooling sources.

[0011] Furthermore, the synchronous retrieval of historical databases analyzes the start-up patterns and power consumption curves of power equipment in the computer room for different date types and time periods, establishes a time-series model of internal heat load for a typical day, and clarifies the processing procedure for the change in heat generation of the computer room itself over time. The process is as follows: Structured operation logs within a specified time range are extracted from historical databases. These logs include precise timestamps, start / stop status of each power device, and real-time power readings. The data is then cleaned and time-series aligned to form a continuous and complete sequence of equipment load data. The data is automatically categorized into different date type templates based on date attributes, identifying potential differences in equipment operation modes under different social activity cycles. The data is then analyzed at the hourly time granularity to calculate the average start-up probability of power equipment and typical load rate and power consumption distribution in each time period under each date type, and to draw load characteristic curves for different time periods. Data is automatically categorized into different date type templates based on date attributes, such as regular weekdays, weekends, and public holidays; By extracting representative load patterns for each date type using regression algorithms, an internal heat load time series model with time as the independent variable is constructed to quantitatively describe the typical pattern of the data center's own heat generation power changing with the time of day.

[0012] Furthermore, the external temperature drop influence function is coupled with the internal heat load time series model for calculation. Through simulation comparison, it is predicted which consecutive periods of external natural cooling capacity will be sufficient to offset or significantly reduce the net heat load in the computer room, and the specific time windows for actively utilizing natural cold sources are identified. The process for obtaining the natural cold source utilization time table is as follows: The natural temperature drop rate for each future time period, based on weather forecasts and outputted by the external temperature drop influence function, is simultaneously superimposed with the heat generation rate of the computer room itself for the corresponding time period predicted by the internal heat load time series model to obtain the net heat load at each future time point. Set a threshold to identify consecutive time periods when the net heat load is consistently below the threshold. These periods indicate that the external natural cooling capacity is sufficient to reduce the heat load in the computer room. These consecutive time periods that meet the criteria are merged and optimized, and segments that are too short to allow the system to switch stably are removed. Combined with the minimum start-up and shutdown time of the refrigeration equipment and the engineering constraints of the system's thermal inertia, a natural cold source utilization time period table is generated, and the specific start and end time windows for starting to utilize and maximizing the utilization of the natural cold source and ending the utilization of the natural cold source are marked.

[0013] Furthermore, the feedback module continuously collects actual operating data after the execution of the control strategy and evaluates the effect, feeding it back to the cooling optimization decision module and the external temperature interaction scheduling module. The processing procedure is as follows: Through the sensor network and device controllers deployed in the computer room, the actual operating data generated after the execution of the latest control strategy is continuously and in real time collected, including the final stable temperature of each area of ​​the computer room, the instantaneous and cumulative energy consumption of each cooling device, and the actual operating status and parameters of the device. The collected actual data is compared with the predicted data made by the previous cooling optimization decision module, the deviation of key indicators is calculated, and the deviation information is fed back to the cooling optimization decision module and the external temperature interaction scheduling module.

[0014] Compared with the prior art, the beneficial effects of the present invention are: When the central air conditioning is turned on, the power equipment in the computer room begins to operate. This invention accurately determines the heating status of each power device in the computer room and the cooling characteristics of each cooling device based on historical data. Then, based on the actual number of power devices turned on and the current real-time temperature of the computer room, it predicts the temperature rise curve of the computer room under the current power equipment operation. This allows for precise planning of the appropriate number of cooling devices to be turned on and their operating parameters, achieving precise dynamic matching between the heat generation and heating process of the power devices and the cooling process of the cooling devices. This method effectively avoids the energy waste problem of starting a large number of cooling devices when only some power devices are turned on. Under the premise of meeting the cooling needs of the central air conditioning computer room, it achieves on-demand allocation of cooling devices, reducing the ineffective consumption of cooling energy from the source. By accessing and processing real-time external environmental temperature and humidity data, weather forecasts, and meteorological data, combined with a heat transfer quantification model trained on historical data, the system can accurately quantify and analyze the rate and magnitude of the impact of external cooling on the internal temperature of the computer room. It clarifies the threshold for an effective temperature difference between the external environment and the computer room interior, determining whether external cooling can actually lower the temperature inside the computer room. Simultaneously, the system analyzes the operating patterns and heat generation of power equipment at different time periods in historical data, precisely mapping the number of operating power equipment and heat generation at different times to the periods of natural external cooling. When it is determined that external cooling can effectively impact the computer room, the system can calculate and determine in advance the number of cooling devices that can be safely shut down or operated at reduced capacity, making advance adjustments to the cooling equipment. This fully utilizes natural cold sources to replace mechanical refrigeration, avoiding the situation of ignoring the cooling potential of external temperature differences and operating too many cooling devices, maximizing the energy-saving effect of natural cold sources, and thus significantly reducing the overall cooling energy consumption of the computer room. Attached Figure Description

[0015] Figure 1 This is a block diagram of a high-efficiency data center energy-saving control system based on global optimization according to the present invention.

[0016] Figure 2 This is a flowchart of the cooling optimization decision module of the present invention.

[0017] Figure 3 This is a flowchart of the external temperature interaction scheduling module of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] like Figure 1As shown, a high-efficiency data center energy-saving control system based on global optimization includes a multi-source data acquisition module, a cooling optimization decision module, an external temperature interactive scheduling module, a control module, and a feedback module. The multi-source data acquisition module is used to collect and integrate multi-source data around the clock, including real-time power consumption and operating status of power and cooling equipment inside the computer room, temperature in various areas of the computer room, and external environmental temperature and humidity and weather forecast data obtained through API. In this embodiment, the multi-source data acquisition module is used to collect and fuse multi-source data around the clock, including real-time power consumption and operating status of the power and cooling equipment inside the computer room, temperature in various areas of the computer room, and external environmental temperature and humidity and weather forecast data obtained through API. The processing procedure is as follows: Through various sensor networks deployed inside the computer room, the system monitors and collects the operating data of various power and refrigeration equipment inside the computer room in real time around the clock, continuously acquiring parameters such as real-time power consumption, operating frequency, and start / stop status of each chiller unit, water pump, and cooling tower fan; as well as acquiring temperature sensor data from different functional areas within the computer room. External environmental data is obtained regularly from the meteorological service platform, including real-time outdoor temperature and humidity, wind speed and direction, and weather forecast information for the next 12 hours. The system aligns and integrates two types of data streams—internal sensors and external APIs—according to a unified timestamp, and automatically performs data cleaning to remove outliers, fill in reasonable missing values, and standardize data of different formats and dimensions into a unified format that can be processed within the system, forming a complete global data view with timestamps.

[0020] The cooling optimization decision module predicts the temperature rise curve of the computer room based on historical data and establishes the performance model of each cooling device. With the upper limit of the computer room temperature as a constraint and the minimum total cooling energy consumption of the system as the goal, the module solves the optimal collaborative control strategy of the cooling devices in real time through optimization algorithms to achieve a dynamic balance of low energy consumption to offset the temperature rise. In this embodiment, the cooling optimization decision module predicts the temperature rise curve of the computer room based on historical data and establishes a performance model for each cooling device. With the upper limit of the computer room temperature as a constraint and the minimum total cooling energy consumption of the system as the objective, the module uses an optimization algorithm to solve the optimal collaborative control strategy of the cooling devices in real time. The dynamic balance process of low energy consumption offsetting temperature rise is as follows: It receives and integrates current data from multi-source data acquisition modules and similar scenario data from historical databases in real time, and uses machine learning algorithms to predict the temperature rise trend and speed of the computer room in the future under the condition of heat generation from existing power equipment operation, forming a dynamic computer room temperature rise curve. Refrigeration equipment performance models are established using raw performance data from refrigeration equipment manufacturers and historical data from long-term operation. These models accurately describe the relationship between the cooling capacity and real-time energy consumption of each chiller unit, precision air conditioner, and other equipment under different load rates and set parameters. The computer room temperature rise curve is coupled with the performance model of each cooling device. The upper limit of the temperature that the computer room must maintain is used as the core constraint, and the minimum total power consumption of the entire cooling system is set as the optimization objective. An optimization problem is constructed and solved by an optimization algorithm to find the combination of cooling devices that should be started in the future and the setting parameters of each operating device should be adjusted.

[0021] It should be noted that the core of this strategy is to dynamically offset the temperature rise caused by the heat generated by the equipment with the lowest possible overall energy consumption, while ensuring that the temperature of the computer room does not exceed the limit. This achieves a precise balance between temperature stability and energy saving, and the strategy instructions are then output to the control module for execution.

[0022] In this embodiment, the temperature rise curve of the computer room is coupled with the performance model of each cooling device. The upper limit of the safe temperature that the computer room must maintain is used as the core constraint, and the minimum total power consumption of the entire cooling system is set as the optimization objective. An optimization problem is constructed and solved using an optimization algorithm to determine the combination of cooling devices that should be activated in the future, and the setting parameters of each operating device should be adjusted accordingly. The processing procedure is as follows: By integrating current equipment operation data with historical data, a predictive model is used to calculate the precise trajectory of the computer room temperature rise over time under the current heat load in the future, i.e., the computer room temperature rise curve. The built-in performance model of the cooling equipment is invoked, and the predicted temperature rise curve is dynamically correlated with the performance models of all available cooling equipment. The requirement that the computer room temperature must be maintained below the safe upper limit is transformed into an inequality constraint for the optimization problem. With the goal of minimizing the sum of the total power consumption of all possible cooling equipment during the prediction period, a mixed integer optimization problem containing continuous variables (setting parameters) and discrete variables (start and stop states) is constructed. Then, the optimal decision sequence in each future control cycle is obtained through a real-time optimization solution algorithm. These models mathematically define the relationship between the cooling efficiency and instantaneous power consumption of each chiller, precision air conditioner, and other equipment under different load rates and set parameters. The output results directly determine which cooling equipment should be started or shut down in the future, as well as the set parameters of each operating device.

[0023] The objective function is modeled as follows: ; The constraints that need to be met include dynamic thermal equilibrium constraints, temperature safety constraints, and equipment operation constraints. Among them, the dynamic constraint of thermal equilibrium is: ;

[0024] Temperature safety constraints include: ; Among them, equipment operation constraints: ; in, It is the total energy consumption over the predicted time. To predict the number of time-domain steps, This represents the total number of refrigeration equipment. It is equipment At any moment power consumption, It is a moment Predict the temperature of the computer room. It is equipment In setting The cooling capacity below, This is a binary decision variable for starting and stopping the device; a value of 0 indicates it is off, and a value of 1 indicates it is on. It controls the step length. For the equivalent heat capacity of the computer room, For the first The total heat output of all power equipment in the computer room at each prediction step size. The maximum allowable temperature for the safe operation of computer room equipment. Each refrigeration unit Its adjustable physical lower and upper limits for set parameters; It should be noted that the formula means that the future heat load is known. Under the premise of external environmental influences, how to dynamically schedule multiple cooling devices to ensure that the computer room temperature never exceeds the standard? Under the given hard constraints, find the most energy-efficient operating strategy; The external temperature interactive scheduling module accesses and processes external ambient temperature and weather forecast data, quantitatively analyzes the rate and magnitude of their impact on the internal temperature of the computer room, and combines the start-up patterns of power equipment in different time periods in historical data to predict the specific time periods during which continuous external cooling can have an effective impact on the internal temperature of the computer room, and predicts in advance the number of cooling equipment that can be adjusted during that time period. In this embodiment, the external temperature interactive scheduling module accesses and processes external ambient temperature and weather forecast data, quantifies and analyzes the rate and magnitude of their impact on the internal temperature of the computer room, and combines historical data on the start-up patterns of power equipment at different time periods to predict the specific time periods during which continuous external cooling can have an effective impact on the internal temperature of the computer room. It then determines in advance the number of adjustable cooling devices available during those time periods. The processing procedure is as follows: Through the pre-built application programming interface, it continuously accesses and processes real-time monitoring data and short-term forecast data from the meteorological department, including outdoor dry-bulb temperature, wet-bulb temperature, wind speed and temperature change curves for the next 12 hours, forming a structured external environment data sequence. A heat transfer quantification model is trained based on the thermal performance of the data center envelope and historical data. The external temperature data is converted into a function that directly affects the internal temperature of the data center. The model outputs the rate and potential magnitude of temperature drop in the data center caused by the external environment per unit time. In this embodiment, a heat transfer quantification model is trained based on the thermal performance of the data center envelope and historical data. The external temperature data is then calculated and transformed into a function directly influencing the internal temperature of the data center. The output of the rate and potential magnitude of temperature drop in the data center caused by the external environment per unit time is as follows: Based on the building drawings and material parameters of the computer room, a thermal physical model of its enclosure structure, including walls and roof, is established to describe the theoretical relationship between outdoor and indoor temperatures. By calling up long-term stored historical data, including the outdoor temperature series of the same period and the corresponding indoor temperature change series of the computer room, the physical model is aligned with these actual data on the time axis. Then, the parameter estimation method is used to back-infer and calibrate the comprehensive heat transfer coefficient and effective heat capacity of the room in the physical model using historical data, so that the output of the model achieves the best fit with the historical real temperature change curve. The calibrated physical model with defined parameters is encapsulated into a computable influence function. This function takes real-time outdoor temperature data as input and, by solving the heat balance equation, directly outputs the rate of temperature drop in the computer room per unit time under current and predicted external temperature conditions, as well as the final possible steady-state temperature difference. Synchronously retrieve historical databases, analyze the start-up patterns and power consumption curves of power equipment in the computer room under different date types and time periods, establish a time series model of internal heat load for typical days, and clarify the pattern of heat generation in the computer room itself changing over time; In this embodiment, historical databases are retrieved simultaneously to analyze the start-up patterns and power consumption curves of power equipment in the computer room under different date types and time periods. An internal heat load time series model for a typical day is established, and the processing procedure for clarifying the change in the computer room's own heat generation over time is as follows: Structured operation logs within a specified time range are extracted from historical databases. These logs include precise timestamps, start / stop status of each power device, and real-time power readings. The data is then cleaned and time-series aligned to form a continuous and complete sequence of equipment load data. The data is automatically categorized into different date type templates based on date attributes, identifying potential differences in equipment operation modes under different social activity cycles. The data is then analyzed at the hourly time granularity to calculate the average start-up probability of power equipment and typical load rate and power consumption distribution in each time period under each date type, and to draw load characteristic curves for different time periods. By extracting representative load patterns for each date type using regression algorithms, an internal heat load time series model with time as the independent variable is constructed to quantitatively describe the typical pattern of the data center's own heat generation power changing with the time of day.

[0025] It should be noted that after the heat load time series model is established, the model's prediction accuracy needs to be tested using the part of historical data that was not involved in the modeling through cross-validation. The model parameters should be fine-tuned based on error analysis to ensure that it can reliably reflect the dynamic changes of the computer room's heat load, thereby providing accurate internal heat source input for subsequent natural cooling source utilization decisions. The external temperature drop influence function is coupled with the internal heat load time series model for calculation. Through simulation comparison, it is predicted which consecutive periods of external natural cooling capacity will be sufficient to offset or significantly reduce the net heat load in the computer room. The specific time windows for actively utilizing natural cold sources are identified, and a natural cold source utilization time table is obtained. In this embodiment, the external temperature drop influence function is coupled with the internal heat load time series model for calculation. Through simulation comparison, it is predicted which consecutive time periods in the future will have sufficient external natural cooling capacity to offset or significantly reduce the net heat load in the computer room. Specific time windows where natural cooling sources can be actively utilized are identified, and the natural cooling source utilization time schedule is obtained. The processing procedure is as follows: The natural temperature drop rate for each future time period, based on weather forecasts and outputted by the external temperature drop influence function, is simultaneously superimposed with the heat generation rate of the computer room itself for the corresponding time period predicted by the internal heat load time series model to obtain the net heat load at each future time point. Set a threshold to identify consecutive time periods when the net heat load is consistently below the threshold. These periods indicate that the external natural cooling capacity is sufficient to reduce the heat load in the computer room. It should be noted that, typically, by analyzing a large amount of historical operating data, reviewing the periods in which natural cooling was successfully and safely utilized, calculating the statistical distribution of net heat load during these periods, and then selecting a low quantile as an initial threshold, the natural cooling capacity is ensured to be sufficient and stable in most cases. At the same time, a safety correction amount must be added to this threshold, which needs to take into account the error of temperature prediction, system response delay, and the sensitivity of the equipment to temperature fluctuations, and finally determine a value of the minimum cooling requirement of the equipment. In addition, in actual operation, this threshold can be adjusted and dynamically fine-tuned based on feedback from long-term operation, thereby maximizing the utilization time of natural cold sources while ensuring the absolute safety of the computer room temperature. These consecutive time periods that meet the conditions are merged and optimized, and segments that are too short to allow the system to switch stably are removed. Combined with the minimum start and stop time of the refrigeration equipment and the engineering constraints of the system's thermal inertia, a natural cold source utilization time period table is generated, and the specific start and end time windows for starting to utilize and maximizing the utilization of the natural cold source and ending the utilization of the natural cold source are marked. Based on the predicted available time period of natural cold source, and further based on the estimated cooling output of natural cooling during that time period and the performance characteristics of the refrigeration equipment, the specific number, type and capacity of mechanical refrigeration equipment that can be safely shut down or derated are calculated to form a pre-adjustment strategy. It should be noted that, based on the predicted external temperature for this period and the established heat transfer model of the computer room, the maximum cooling capacity that the natural cold source can provide is calculated. Then, the performance model of the refrigeration equipment is called to analyze the real-time cooling capacity and power consumption characteristics of each mechanical refrigeration device under the current computer room load. Next, the estimated cooling capacity of the natural cold source is balanced with the predicted internal heat load of the computer room for this period to determine the remaining cooling capacity that needs to be supplemented or undertaken by the mechanical refrigeration system. Then, based on the remaining cooling capacity demand, starting from high-energy-consuming or low-efficiency equipment, mechanical refrigeration equipment that can be safely shut down or derated is selected in sequence, and it is calculated whether the cooling capacity released after shutting down or derated is within the safe range that the natural cold source can cover, while ensuring that the reserved mechanical refrigeration capacity is sufficient to cope with possible load fluctuations. Finally, based on the above calculation results, a pre-adjustment strategy containing a list of specific operable equipment, adjustment methods, and execution time windows is generated, and this strategy is sent to the global optimization decision module as an optimization constraint for final approval and coordinated execution.

[0026] The generated natural cold source utilization schedule and corresponding refrigeration equipment adjustment suggestions are used as a forward-looking constraint and optimization target and transmitted to the control module in real time. This allows the system to plan ahead and seamlessly integrate with natural cooling solutions when formulating global control strategies, thereby minimizing the total energy consumption of the system.

[0027] The control module is used to convert the generated policy instructions into commands that can be executed by specific devices; The feedback module continuously collects actual operating data after the execution of the control strategy, and evaluates the effect before feeding it back to the cooling optimization decision module and the external temperature interaction scheduling module. In this embodiment, the feedback module continuously collects actual operating data after the execution of the control strategy, and evaluates the effect before feeding it back to the cooling optimization decision module and the external temperature interaction scheduling module. The processing procedure is as follows: Through the sensor network and device controllers deployed in the computer room, the actual operating data generated after the execution of the latest control strategy is continuously and in real time collected, including the final stable temperature of each area of ​​the computer room, the instantaneous and cumulative energy consumption of each cooling device, and the actual operating status and parameters of the device. The collected actual data is compared with the predicted data made by the previous cooling optimization decision module, the deviation of key indicators is calculated, and the deviation information is fed back to the cooling optimization decision module and the external temperature interaction scheduling module.

[0028] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A high-efficiency data center energy-saving control system based on global optimization, characterized in that, It includes a multi-source data acquisition module, a cooling optimization decision module, an external temperature interaction scheduling module, a control module, and a feedback module; The multi-source data acquisition module is used to collect and integrate multi-source data around the clock, including real-time power consumption and operating status of power and cooling equipment inside the computer room, temperature in various areas of the computer room, and external environmental temperature and humidity and weather forecast data obtained through API. The cooling optimization decision module predicts the temperature rise curve of the computer room based on historical data and establishes the performance model of each cooling device. With the upper limit of the computer room temperature as a constraint and the minimum total cooling energy consumption of the system as the goal, the module solves the optimal collaborative control strategy of the cooling devices in real time through optimization algorithms to achieve a dynamic balance of low energy consumption to offset the temperature rise. The external temperature interactive scheduling module accesses and processes external ambient temperature and weather forecast data, quantifies and analyzes the rate and magnitude of their impact on the temperature inside the computer room, and, combined with the start-up patterns of power equipment in different time periods in historical data, predicts the specific time periods during which continuous external cooling can have an effective impact on the inside of the computer room, and determines in advance the number of cooling equipment that can be adjusted during that time period. The control module is used to convert the generated policy instructions into commands that can be executed by specific devices; The feedback module continuously collects actual operating data after the execution of the control strategy and evaluates the effect, feeding it back to the cooling optimization decision module and the external temperature interaction scheduling module.

2. The high-efficiency data center energy-saving control system based on global optimization according to claim 1, characterized in that, The multi-source data acquisition module is used to collect and integrate multi-source data around the clock, including real-time power consumption and operating status of power and cooling equipment inside the computer room, temperature in various areas of the computer room, and external environmental temperature and humidity and weather forecast data obtained through API. The processing procedure is as follows: Through various sensor networks deployed inside the computer room, the system monitors and collects the operating data of various power and refrigeration equipment inside the computer room in real time around the clock, continuously acquiring parameters such as real-time power consumption, operating frequency, and start / stop status of each chiller unit, water pump, and cooling tower fan; as well as acquiring temperature sensor data from different functional areas within the computer room. External environmental data is obtained regularly from the meteorological service platform, including real-time outdoor temperature and humidity, wind speed and direction, and weather forecast information for the next 12 hours. The system aligns and integrates two types of data streams—internal sensors and external APIs—according to a unified timestamp, and automatically performs data cleaning to remove outliers, fill in reasonable missing values, and standardize data of different formats and dimensions into a unified format that can be processed within the system, forming a complete global data view with timestamps.

3. The high-efficiency data center energy-saving control system based on global optimization according to claim 1, characterized in that, The cooling optimization decision module predicts the temperature rise curve of the computer room based on historical data and establishes performance models for each cooling device. With the upper limit of the computer room temperature as a constraint and the minimum total cooling energy consumption as the objective, it uses an optimization algorithm to solve for the optimal collaborative control strategy of the cooling devices in real time. The dynamic balance process of low energy consumption offsetting temperature rise is as follows: It receives and integrates current data from multi-source data acquisition modules and similar scenario data from historical databases in real time, and uses machine learning algorithms to predict the temperature rise trend and speed of the computer room in the future under the condition of heat generation from existing power equipment operation, forming a dynamic computer room temperature rise curve. Refrigeration equipment performance models are established by combining raw performance data from refrigeration equipment manufacturers with historical data from long-term operation. The computer room temperature rise curve is coupled with the performance model of each cooling device. The upper limit of the temperature that the computer room must maintain is used as the core constraint, and the minimum total power consumption of the entire cooling system is set as the optimization objective. An optimization problem is constructed and solved by an optimization algorithm to find the combination of cooling devices that should be started in the future and the setting parameters of each operating device should be adjusted.

4. The high-efficiency data center energy-saving control system based on global optimization according to claim 3, characterized in that, The process involves coupling the computer room temperature rise curve with the performance models of each cooling device, using the upper limit of the safe temperature that the computer room must maintain as the core constraint, and setting the minimum total power consumption of the entire cooling system as the optimization objective. An optimization problem is then constructed and solved using an optimization algorithm to determine the optimal combination of cooling devices to be activated in the future, as well as the appropriate adjustment parameters for each operating device. The process is as follows: By integrating current equipment operation data with historical data, a predictive model is used to calculate the precise trajectory of the computer room temperature rise over time under the current heat load in the future, i.e., the computer room temperature rise curve. The built-in performance model of the cooling equipment is invoked, and the predicted temperature rise curve is dynamically correlated with the performance models of all available cooling equipment. The requirement that the computer room temperature must be maintained below the safe upper limit is transformed into an inequality constraint for the optimization problem. With the goal of minimizing the sum of the total power consumption of all possible cooling equipment during the prediction period, a mixed integer optimization problem containing continuous variables (setting parameters) and discrete variables (start and stop states) is constructed. Then, the optimal decision sequence in each future control cycle is obtained through a real-time optimization solution algorithm. The objective function model formula is as follows: ； The constraints that need to be met include dynamic thermal equilibrium constraints, temperature safety constraints, and equipment operation constraints. Among them, the dynamic constraint of thermal equilibrium is: ； ； Temperature safety constraints include: ； Among them, equipment operation constraints: ； in, It is the total energy consumption over the predicted time. To predict the number of time-domain steps, This represents the total number of refrigeration equipment. It is equipment At any moment power consumption, It is a moment Predict the temperature of the computer room. It is equipment In setting The cooling capacity below, This is a binary decision variable for starting and stopping the device; a value of 0 indicates it is off, and a value of 1 indicates it is on. It controls the step length. For the equivalent heat capacity of the computer room, For the first The total heat output of all power equipment in the computer room at each prediction step size. The maximum allowable temperature for the safe operation of computer room equipment. Each refrigeration unit Its adjustable physical lower and upper limits for parameters.

5. The high-efficiency data center energy-saving control system based on global optimization according to claim 1, characterized in that, The external temperature interactive scheduling module receives and processes external ambient temperature and weather forecast data, quantifies and analyzes the rate and magnitude of their impact on the internal temperature of the computer room, and combines historical data on the start-up patterns of power equipment at different time periods to predict the specific time periods during which continuous external cooling can effectively affect the internal temperature of the computer room. It then determines in advance the number of adjustable cooling devices available during those periods. The processing procedure is as follows: Through the pre-built application programming interface, it continuously accesses and processes real-time monitoring data and short-term forecast data from the meteorological department, including outdoor dry-bulb temperature, wet-bulb temperature, wind speed and temperature change curves for the next 12 hours, forming a structured external environment data sequence. A heat transfer quantification model is trained based on the thermal performance of the data center envelope and historical data. The external temperature data is converted into a function that directly affects the internal temperature of the data center. The model outputs the rate and potential magnitude of temperature drop in the data center caused by the external environment per unit time. Synchronously retrieve historical databases, analyze the start-up patterns and power consumption curves of power equipment in the computer room under different date types and time periods, establish a time series model of internal heat load for typical days, and clarify the pattern of heat generation in the computer room itself changing over time; The external temperature drop influence function is coupled with the internal heat load time series model for calculation. Through simulation comparison, it is predicted which consecutive periods of external natural cooling capacity will be sufficient to offset or significantly reduce the net heat load in the computer room. The specific time windows for actively utilizing natural cold sources are identified, and a natural cold source utilization time table is obtained. Based on the predicted available time period of natural cold source, and further based on the estimated cooling output of natural cooling during that time period and the performance characteristics of the refrigeration equipment, the specific number, type and capacity of mechanical refrigeration equipment that can be safely shut down or derated are calculated to form a pre-adjustment strategy. The generated natural cold source utilization schedule and corresponding refrigeration equipment adjustment suggestions are used as a forward-looking constraint and optimization target and transmitted to the control module in real time. This allows the system to plan ahead and seamlessly integrate with natural cooling solutions when formulating global control strategies, thereby minimizing the total energy consumption of the system.

6. The high-efficiency data center energy-saving control system based on global optimization according to claim 5, characterized in that, The heat transfer quantification model trained based on the thermal performance of the data center envelope and historical data is used to calculate and transform the external temperature data into a function that directly affects the internal temperature of the data center. The processing steps for outputting the rate and potential magnitude of temperature drop in the data center caused by the external environment per unit time are as follows: Based on the building drawings and material parameters of the computer room, a thermal physical model of its enclosure structure, including walls and roof, is established to describe the theoretical relationship between outdoor and indoor temperatures. By calling up long-term stored historical data, including the outdoor temperature series of the same period and the corresponding indoor temperature change series of the computer room, the physical model is aligned with these actual data on the time axis. Then, the parameter estimation method is used to back-infer and calibrate the comprehensive heat transfer coefficient and effective heat capacity of the room in the physical model using historical data, so that the output of the model achieves the best fit with the historical real temperature change curve. The calibrated physical model with defined parameters is encapsulated into a computable influence function. This function takes real-time outdoor temperature data as input and, by solving the heat balance equation, directly outputs the rate of temperature drop in the computer room per unit time under current and predicted external temperature conditions, as well as the final possible steady-state temperature difference.

7. A high-efficiency data center energy-saving control system based on global optimization according to claim 5, characterized in that, The synchronous retrieval of historical databases analyzes the start-up patterns and power consumption curves of power equipment in the computer room for different date types and time periods, establishes a time-series model of internal heat load for a typical day, and clarifies the processing procedure for the change of heat generation in the computer room over time. Structured operation logs within a specified time range are extracted from historical databases. These logs include precise timestamps, start / stop status of each power device, and real-time power readings. The data is then cleaned and time-series aligned to form a continuous and complete sequence of equipment load data. The data is automatically categorized into different date type templates based on date attributes, identifying potential differences in equipment operation modes under different social activity cycles. The data is then analyzed at the hourly time granularity to calculate the average start-up probability of power equipment and typical load rate and power consumption distribution in each time period under each date type, and to draw load characteristic curves for different time periods. By extracting representative load patterns for each date type using regression algorithms, an internal heat load time series model with time as the independent variable is constructed to quantitatively describe the typical pattern of the data center's own heat generation power changing with the time of day.

8. A high-efficiency data center energy-saving control system based on global optimization according to claim 5, characterized in that, The process of coupling the external temperature drop influence function with the internal heat load time series model and predicting which consecutive periods of external natural cooling capacity will be sufficient to offset or significantly reduce the net heat load in the computer room through simulation comparison is as follows: The specific time windows for actively utilizing natural cooling sources are identified, and the natural cooling source utilization time schedule is obtained. The natural temperature drop rate for each future time period, based on weather forecasts and outputted by the external temperature drop influence function, is simultaneously superimposed with the heat generation rate of the computer room itself for the corresponding time period predicted by the internal heat load time series model to obtain the net heat load at each future time point. Set a threshold to identify consecutive time periods when the net heat load is consistently below the threshold. These periods indicate that the external natural cooling capacity is sufficient to reduce the heat load in the computer room. These consecutive time periods that meet the criteria are merged and optimized, and segments that are too short to allow the system to switch stably are removed. Combined with the minimum start-up and shutdown time of the refrigeration equipment and the engineering constraints of the system's thermal inertia, a natural cold source utilization time period table is generated, and the specific start and end time windows for starting to utilize and maximizing the utilization of the natural cold source and ending the utilization of the natural cold source are marked.

9. A high-efficiency data center energy-saving control system based on global optimization according to claim 1, characterized in that, The feedback module continuously collects actual operating data after the execution of the control strategy, and evaluates the effect before feeding it back to the cooling optimization decision module and the external temperature interaction scheduling module. The processing procedure is as follows: Through the sensor network and device controllers deployed in the computer room, the actual operating data generated after the execution of the latest control strategy is continuously and in real time collected, including the final stable temperature of each area of ​​the computer room, the instantaneous and cumulative energy consumption of each cooling device, and the actual operating status and parameters of the device. The collected actual data is compared with the predicted data made by the previous cooling optimization decision module, the deviation of key indicators is calculated, and the deviation information is fed back to the cooling optimization decision module and the external temperature interaction scheduling module.

Images