Data center evaporation cold total heat recovery double-energy-storage air conditioning system and control method

Abstract

According to the data center evaporation cold total heat recovery double-energy-storage air conditioning system and the control method, the obvious energy efficiency improvement and economic optimization effects are achieved, the future server heat load, environment parameters and electricity price changes are accurately pre-judged through a composite prediction model of the central control module, and the real-time charge state of the double-energy-storage module is combined, so that the energy efficiency is improved. A collaborative operation instruction with the optimal cost is generated; by means of deep coupling of the evaporative cooling module and the total heat recovery module, a natural cold source is utilized to the maximum extent, and the running time of a conventional refrigeration module is shortened; through cooperative scheduling of the cold energy storage unit and the heat storage unit, efficient cold / heat storage is achieved in the electricity price valley value period, stored energy is released preferentially in the peak value period, and frequent starting and stopping of equipment are avoided; and on the basis of the multi-dimensional constraint optimization decision model, the total operation cost is minimized on the premise that the system meets the safety temperature and humidity requirements of the data center.

Description

Technical Field

[0001] This invention relates to the field of data center air conditioning technology, and in particular to a dual-energy storage air conditioning system and control method for data center evaporative cooling and total heat recovery. Background Technology

[0002] As a core infrastructure of the digital economy, data centers face increasingly prominent energy consumption issues, with air conditioning systems accounting for approximately 30%-40% of total data center energy consumption. To reduce operating costs and carbon emissions, existing technologies commonly employ evaporative cooling, heat recovery, or energy storage. Evaporative cooling technology utilizes the principle of water evaporation to absorb heat, using ambient air with lower wet-bulb temperatures to cool the data center, significantly reducing the need for mechanical cooling. Total heat recovery technology recovers sensible and latent heat from server exhaust air using rotary or plate heat exchangers, which is then used to preheat fresh air or domestic hot water. Energy storage technology leverages time-of-use electricity pricing mechanisms, using off-peak electricity hours to produce ice or cool phase change materials (such as paraffin wax), releasing cooling energy during peak hours. When applied individually, these technologies have already improved energy efficiency to some extent; for example, the PUE (Power Usage Effectiveness) of data centers using evaporative cooling can be reduced to below 1.5.

[0003] However, the existing structure has the following key problems:

[0004] Insufficient technological synergy: Evaporative cooling, heat recovery, and energy storage systems typically operate independently or in simple series, lacking a dynamic coordination mechanism. When the ambient wet-bulb temperature rises, the evaporative cooling efficiency drops sharply, requiring the immediate activation of conventional refrigeration modules to compensate for the cooling, resulting in a sharp increase in energy consumption. Furthermore, if the waste heat generated by the heat recovery module is not utilized immediately (such as during low-load periods at night), it is directly released into the environment, causing energy waste (heat recovery efficiency loss of approximately 40%).

[0005] Load response lag: Traditional control strategies rely on real-time monitoring data, but data center heat load fluctuates drastically due to task scheduling, and ambient temperature and humidity are also time-varying. Existing systems lack predictive capabilities and cannot adjust their operating modes in a timely manner, often resulting in mismatches in the timing of energy storage unit charging and discharging (such as being forced to release cooling during peak electricity price periods), or frequent start-ups and shutdowns of conventional cooling modules (increasing equipment wear by 10%-15%).

[0006] Lack of economic optimization: Time-of-use pricing mechanisms are not deeply integrated. Although conventional cold storage systems store cold during off-peak hours, they do not take into account future heat load and environmental parameter forecasts, which can easily lead to insufficient or excessive energy storage (cold energy waste rate as high as 20%). Furthermore, the coupling relationship between heat recovery and cold energy storage is not coordinated, making it impossible to minimize the total operating cost of the system.

[0007] Therefore, the following problems urgently need to be solved in this field: how to construct a highly integrated air conditioning system that dynamically balances the operation of evaporative cooling, heat recovery, conventional refrigeration and dual energy storage modules through multi-module intelligent collaboration and forward-looking optimization control, so as to completely solve the technical bottlenecks of low energy utilization, high operating costs and slow load response. Summary of the Invention

[0008] The purpose of this invention is to provide a dual-energy storage air conditioning system and control method for data center evaporative cooling and total heat recovery, in order to solve the problems existing in the prior art.

[0009] To achieve the above objectives, the present invention provides the following solution:

[0010] This invention provides a dual-energy storage air conditioning system for data centers, comprising:

[0011] The data acquisition module is used to collect server heat load data and return air temperature and humidity data inside the data center in real time, as well as ambient dry bulb temperature data, ambient wet bulb temperature data and time-of-use electricity price data outside the data center.

[0012] The evaporative cooling module includes a fresh air inlet and an air outlet, wherein the air outlet is connected to the cold aisle of the data center via an air supply duct.

[0013] The total heat recovery module includes an exhaust inlet and a processed air outlet. The exhaust inlet is connected to the hot aisle of the data center through an exhaust duct, and the processed air outlet is connected to the fresh air inlet of the evaporative cooling module.

[0014] A conventional refrigeration module includes a compressor, a condenser, a throttling device, and an evaporator. The evaporator is located in the first air supply duct and is used to deeply cool the air processed by the evaporative cooling module.

[0015] The dual energy storage module includes a cold energy storage unit and a heat energy storage unit; the cold energy storage unit is connected to the evaporator of the conventional refrigeration module through a first supply and return water pipeline, and provides cold energy to the evaporator or stores the cold energy generated by the evaporator; the heat energy storage unit is connected to the total heat recovery module through a second supply and return water pipeline, and is used to store the heat recovered by the total heat recovery module;

[0016] The central control module is connected to the data acquisition module, evaporative cooling module, total heat recovery module, conventional refrigeration module, and dual energy storage module. It is used to construct and store a composite prediction model based on historical data collected by the data acquisition module, predicting server heat load, ambient temperature and humidity, and time-of-use electricity prices within a preset time period. During system operation, it acquires real-time data collected by the data acquisition module and uses the composite prediction model to generate a predicted data sequence for the preset time period. Based on the predicted data sequence and the current energy storage state of the dual energy storage module, it solves for the objective function of minimizing the total system operating cost, generating coordinated operation control commands for the evaporative cooling module, total heat recovery module, conventional refrigeration module, and dual energy storage module. It then sends these coordinated operation control commands to control the start / stop, operating power, and switching of operating modes of each module.

[0017] Preferably, the data acquisition module includes:

[0018] A first temperature and humidity sensor is deployed within the hot aisle of the data center.

[0019] A second temperature and humidity sensor is deployed outside the data center;

[0020] A load monitoring unit is connected to the server cluster management system interface to obtain real-time total power or CPU utilization as server thermal load data.

[0021] The electricity price acquisition unit is connected to the data platform interface of the power grid company and is used to acquire time-of-use electricity price data.

[0022] Preferably, the total heat recovery module is a rotary total heat recovery unit. The rotor inside the rotary total heat recovery unit is made of moisture-absorbing material. During the rotation process, the rotor alternately contacts the exhaust air from the heat channel and the fresh air from the external environment, thereby achieving dual recovery of heat and humidity.

[0023] Preferably, the cold energy storage unit is a phase change energy storage tank, which is filled with encapsulated phase change material. The conventional refrigeration module operates during off-peak electricity prices and stores the cold energy in the phase change material through the circulating medium in the first supply and return water pipeline.

[0024] Preferably, the composite prediction model is a long short-term memory network model. The central control module trains the long short-term memory network model by constructing a multi-dimensional time series from historical server heat load data, environmental dry-bulb temperature data, environmental wet-bulb temperature data, and time-of-use electricity price data to obtain prediction capabilities.

[0025] The present invention also provides a data center air conditioning control method, comprising the following steps:

[0026] S1: Data acquisition and preprocessing, real-time acquisition of server heat load, return air temperature and humidity inside the data center, as well as dry and wet bulb temperatures of the external environment and the time-of-use electricity price at the current moment; and normalization and outlier filtering of the acquired raw data to form standardized real-time status data.

[0027] S2: Multidimensional prediction, calling a pre-built composite prediction model, taking the standardized real-time status data and historical status data as input, and outputting server heat load prediction, environmental dry and wet bulb temperature prediction, and time-of-use electricity price prediction for multiple time steps within a future preset scheduling cycle.

[0028] S3: Optimize decision-making, construct a function with the total operating cost of the system within the scheduling period as the optimization objective, take the predicted values ​​of each parameter generated by the multi-dimensional prediction step and the current state of charge of the dual energy storage module as constraints, solve the objective function through an optimization algorithm, and obtain the optimal combination of operating strategies for the evaporative cooling module, conventional refrigeration module, cold storage unit and heat storage unit at each time step within the scheduling period;

[0029] S4: Instruction execution. Based on the optimal operating strategy combination, specific control instructions are generated and sent to the actuators of the evaporative cooling module, the total heat recovery module, the conventional refrigeration module, and the dual energy storage module. The control instructions include the fan speed, water pump flow rate, compressor start / stop and operating frequency, and pipeline valve opening degree of each module.

[0030] Preferably, the formula with the objective function of minimizing the total system operating cost is as follows:

[0031]

[0032] Where t is the time step index within the scheduling period, N is the total number of time steps, and P elec (t) represents the predicted electricity price at time t, E fan (t), E pump (t), E comp (t) represents the predicted total energy consumption of all fans, pumps, and compressors in the system at time t. The predicted total energy consumption is calculated based on the power model of each component under its corresponding operating strategy. M switch (t) is the penalty term for the number of start-stop switching times of the main components of the system at time t, and α is the weight coefficient of the penalty term.

[0033] Preferably, the constraints include at least:

[0034] Data center supply air temperature constraint: The supply air temperature T after system processing at any time step t. supply (t) must satisfy T min ≤T supply (t)≤T max T min and T max The lower and upper temperature limits required for the safe operation of data center equipment;

[0035] Cooling balance constraint: At any time step t, the sum of the cooling capacity provided by the evaporative cooling module, the cooling capacity provided by the conventional refrigeration module, and the cooling capacity released by the cooling storage unit must be greater than or equal to the predicted value of the server's heat load.

[0036] Energy storage unit state constraints: The state of charge of cold energy storage units and heat energy storage units must be between their designed minimum and maximum capacities at any time step t.

[0037] Preferably, the system switches between at least four modes based on the optimal combination of operating strategies:

[0038] Mode 1: Full evaporative cooling mode. When the predicted ambient wet-bulb temperature meets the conditions for direct cooling, the evaporative cooling module is turned on to maximum power and the conventional cooling module is turned off.

[0039] Mode 2: Energy storage priority mode. When the electricity price forecast is at its lowest point and the data center heat load is low, the conventional cooling module is turned on to charge the cold storage unit, while evaporative cooling is used to meet the real-time cooling load.

[0040] Mode 3: Combined cooling mode. When evaporative cooling cannot meet the heat load demand independently, the evaporative cooling module and the conventional refrigeration module are turned on together or the cold energy storage unit is released to provide cooling in a coordinated manner.

[0041] Mode 4: Total Heat Recovery Mode. When the external ambient temperature is low and preheating of the fresh air is required, the total heat recovery module is activated to recover the waste heat of the exhaust air and store the excess heat in the heat storage unit.

[0042] Preferably, the method for constructing the composite prediction model includes:

[0043] Obtain a historical dataset with the same dimensions as that in step S1, which is at least one year old.

[0044] The historical dataset is divided to generate a training set, a validation set, and a test set.

[0045] Define the network structure of the Long Short-Term Memory (LSTM) network model, including the input layer dimension, the number of hidden layers, the number of neurons per layer, and the output layer dimension;

[0046] The long short-term memory network model is iteratively trained using the training set, and the model performance is evaluated using the validation set after each iteration. The hyperparameters of the model are then adjusted based on the evaluation results.

[0047] Once the model's performance on the validation set reaches the preset convergence criterion, training stops, and the test set is used to perform a final evaluation of the model's generalization ability. If the model passes the evaluation, it is then stored in the central control module.

[0048] The present invention achieves the following beneficial technical effects compared to the prior art:

[0049] This invention provides a dual-energy storage air conditioning system and control method for data centers, featuring evaporative cooling and total heat recovery. It significantly improves energy efficiency and optimizes economic performance. Through a composite prediction model in the central control module, it accurately predicts future server heat load, environmental parameters, and electricity price changes. Combined with the real-time state of charge of the dual energy storage modules, it generates cost-optimized collaborative operation commands. By leveraging the deep coupling of the evaporative cooling module and the total heat recovery module, it maximizes the utilization of natural cold sources and reduces the operating time of conventional cooling modules. Through the coordinated scheduling of cold and heat storage units, it efficiently stores cold / heat during off-peak electricity periods and prioritizes energy release during peak periods, avoiding frequent equipment start-ups and shutdowns. Based on a multi-dimensional constraint-based optimization decision model, it ensures that the system minimizes total operating costs while meeting the safe temperature and humidity requirements of the data center. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 A schematic diagram of the structure of the data center evaporative cooling and total heat recovery dual energy storage air conditioning system provided by the present invention;

[0052] Figure 2 A flowchart of the data center air conditioning control method provided by the present invention. Detailed Implementation

[0053] The serial numbers assigned to components in this document, such as "first," "second," etc., are merely used to distinguish the described objects and have no sequential or technical meaning. The terms "connection" and "linkage" used in this application, unless otherwise specified, include both direct and indirect connections (linkages). In the description of this invention, it should be understood that the terms "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention.

[0054] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "over," and "on top" of the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "beneath" of the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0056] The purpose of this invention is to provide a dual-energy storage air conditioning system and control method for data center evaporative cooling and total heat recovery, in order to solve the problems existing in the prior art.

[0057] Example 1:

[0058] To make the objectives, technical solutions, and advantages of this invention clearer, the following description is provided in conjunction with the appendix. Figure 1 and attached Figure 2 This paper provides a detailed description of a typical embodiment of the data center evaporative cooling and total heat recovery dual-energy storage air conditioning system and control method of the present invention. It should be emphasized that this embodiment is merely a preferred example to fully disclose the present invention and does not constitute any limitation on the technical solution. Those skilled in the art, after understanding the essence of the present invention, can adaptively adjust the component models, capacity ratios, connection methods, and control parameters according to site conditions, load scale, climate characteristics, and electricity pricing policies without inventive effort.

[0059] The air conditioning system described in this embodiment is installed in a Tier III data center with a building area of ​​approximately 3000m². 2 The system is designed with a rack power density of 8kW / rack, operating continuously for 8760 hours a year. The location is in a hot-summer, cold-winter region, with a peak-to-valley electricity price difference of approximately 4:1. The system is as follows: Figure 1 As shown. The server room of Data Center 1 adopts a closed hot and cold aisle design. The ceiling of the hot aisle is tightly connected to the exhaust inlet of the total heat recovery module 2 through exhaust duct 5; the cold aisle is supplied with air from under the floor, and the low-temperature air processed by the evaporative cooling module 3 is evenly delivered into the cold aisle through the supply air duct 4. The fresh air inlet of the evaporative cooling module 3 is directly connected to the processed air outlet of the total heat recovery module 2 through a short-distance duct, so that the fresh air after energy exchange has a high enthalpy value before entering the evaporative cooling module 3, thereby effectively reducing the initial load of the evaporative cooling module 3. The evaporator 61 of the conventional cooling module 6 is built into the supply air duct 4 and arranged in series with the evaporative cooling module 3, which can provide additional sensible cooling under extreme wet-bulb temperature conditions. The dual energy storage module 7 is arranged in the auxiliary equipment room of the server room, connected to the evaporator 61 through the first supply and return water pipe 71, and connected to the waste heat recovery section of the total heat recovery module 2 through the second supply and return water pipe 72, with an overall footprint of less than 30m². 2 It has a volume significantly lower than that of traditional water-cooled storage tanks.

[0060] Specifically, the data acquisition module physically consists of a first temperature and humidity sensor, a second temperature and humidity sensor, a load monitoring unit, and an electricity price acquisition unit. The first temperature and humidity sensor uses a wall-mounted multi-probe structure, installed at the center of the top of the hot aisle, with the probes extending into the enclosed hot aisle. It collects return air temperature and humidity data every 30 seconds and uploads it in real time via an RS-485 bus. The second temperature and humidity sensor is an integrated louvered box type, installed next to the fresh air intake on the roof, used to acquire the dry and wet bulb temperatures of the environment, also with a sampling frequency of 30 seconds. The load monitoring unit interfaces with the computer room's environmental monitoring platform via the SNMP protocol, real-time analyzing the instantaneous total power and CPU utilization output by the server cluster management system and converting them into heat load power. The electricity price acquisition unit uses an HTTPS encrypted interface to periodically retrieve the 15-minute time-of-use electricity prices published by the local power grid company and caches them in a local database to prevent data loss for decision-makers due to network interruptions.

[0061] Furthermore, in this embodiment, the evaporative cooling module 3 adopts a two-stage direct evaporative cooling method, consisting of a primary filter, a variable frequency blower, a packed-type evaporative cooling section, and a water baffle. The packing material is flame-retardant polymer honeycomb paper packing with a thickness of 300mm and a face velocity of 2.8m / s. The blower is an EC backward centrifugal fan with a rated air volume of 80,000 m³ / s. 3The fan operates at a speed of / h, with stepless speed regulation within the range of 20%-100%. When the central control module determines that the ambient wet-bulb temperature is below 18℃ and the heat load is below 60% of the design value, the fan operates at the minimum allowable frequency, and the needs of the computer room can be met by evaporative cooling alone. At this time, the conventional cooling module and the dual energy storage module are both in standby mode.

[0062] Furthermore, the total heat recovery module 2 uses a rotary total heat recovery unit with a rotor diameter of 1.5m and a thickness of 200mm. The core is made of aluminum foil substrate coated with a molecular sieve moisture-absorbing coating, achieving a sensible heat efficiency of ≥75% and a latent heat efficiency of ≥65%. The rotor is driven by a brushless DC motor, and its speed can be continuously adjusted within the range of 0-20r / min. When the outdoor temperature is below 10℃ and the machine room still needs to introduce fresh air, the central control module starts the rotor and adjusts the rotor speed and fresh air ratio in real time according to the target supply air temperature, ensuring that the temperature difference between the fresh air preheated by the rotor and the return air temperature of the machine room is less than 3℃, significantly reducing reheat energy consumption. The recovered heat is circulated to the heat storage unit through an ethylene glycol solution in the second supply and return water pipeline; this unit is a pressurized stainless steel water tank with a volume of 2m³. 3 It features an internal U-shaped heat exchange coil, with the coil filled with a paraffin-expanded graphite composite phase change material with a melting point of 55℃, achieving a heat storage density of up to 180 MJ / m³. 3 .

[0063] Furthermore, the conventional refrigeration module 6 adopts a variable frequency air-cooled chiller unit with a fully enclosed scroll compressor. It has a nominal cooling capacity of 260kW and can operate efficiently within a load range of 20%-100%. The condenser is a copper tube aluminum fin type, equipped with a variable frequency EC fan that automatically adjusts the airflow according to the condensing pressure. An electronic expansion valve is used as the throttling device, with a response time of less than 1 second, ensuring precise control of the evaporation temperature. The evaporator 61 is a copper tube aluminum fin dry evaporator, placed in the air supply duct 4 and connected in series with the evaporative cooling module 3. When the evaporative cooling outlet air temperature is still more than 2°C higher than the set air supply temperature, the central control module starts the compressor and adjusts the compressor frequency and electronic expansion valve opening as needed to achieve deep cooling.

[0064] Furthermore, the cold energy storage unit 73 of the dual energy storage module 7 is a phase change energy storage water tank, which is arranged in the return water pipeline of the conventional refrigeration module. The water tank has a volume of 5m³. 3The internal encapsulation contains phase change microcapsules with a melting point of 8℃ and a latent heat of phase change of 215kJ / kg. The first supply and return water pipeline 71 adopts a primary pump variable flow system, with a 25% ethylene glycol solution as the circulating medium and a designed temperature difference of 6℃. When the central control module determines that the electricity price is at its lowest point within the next 4 hours and the predicted heat load is 70% lower than the average, the compressor is started to operate at its rated power, causing the evaporator outlet solution temperature to drop to 3℃, and the phase change material continues to solidify and store cold. During this process, the evaporative cooling module independently undertakes the real-time cooling load of the machine room. When the electricity price enters its peak period or the heat load suddenly increases, the central control module prioritizes the use of the cold storage unit 73 to release cold, and the compressor load gradually decreases or even stops, achieving "peak shaving and valley filling". The heat storage unit 74 is connected to the total heat recovery module 2 through the second supply and return water pipeline 72 and is used to store the heat recovered by the total heat recovery module 2.

[0065] The central control module hardware platform adopts an industrial-grade ARM Cortex-A72 quad-core processor with a main frequency of 1.5GHz, 8GB of LPDDR4 RAM and 64GB of eMMC, and runs on a real-time Linux kernel operating system, ensuring a control cycle of less than 1 second. The composite prediction model is built based on a Long Short-Term Memory (LSTM) network, with a 5-dimensional input dimension: server heat load, ambient dry-bulb temperature, ambient wet-bulb temperature, time-of-use electricity price, and timestamp. The hidden layer uses a two-layer LSTM structure with 128 neurons per layer, and the output dimension predicts the server heat load, ambient dry-bulb and wet-bulb temperatures, and time-of-use electricity price for 96 15-minute time steps within the next 24 hours. The training dataset is taken from operational data from the past two years, totaling approximately 70GB. A sliding window method is used to construct the samples, with a window length of 96 steps and a step size of 15 minutes. During training, the Adam optimizer is used with an initial learning rate of 0.001. After 200 epochs of training, the RMSE of the model on the validation set decreased to 3.8%, meeting engineering accuracy requirements. The predicted results and real-time data are input into a mixed-integer linear programming (MILP) solver. The objective function comprehensively considers electricity costs, equipment start-up and shutdown penalties, and equipment wear and tear costs. Constraints include upper and lower limits of supply air temperature, the heat balance equation, and the charge state boundary of the energy storage unit. The solver uses a branch and bound algorithm, with an average solution time of 45 seconds, and outputs the optimal operating strategy for each module in 15-minute increments over the next 24 hours.

[0066] Example 2:

[0067] This embodiment provides a control method for the above system, the process of which is as follows: Figure 2As shown. After the system powers on, it first executes step S1: data acquisition and preprocessing. The central control module periodically reads data from the first temperature and humidity sensor, the second temperature and humidity sensor, the load monitoring unit, and the electricity price acquisition unit. It performs moving average filtering and 3σ outlier removal on the raw data, followed by normalization to unify the data range of each dimension to the [0,1] interval. After preprocessing, a standardized real-time state vector is formed and cached in a circular buffer for subsequent prediction.

[0068] Step S2: Multidimensional Prediction. The central control module concatenates the latest 96-step historical state vector with the real-time state vector and inputs it into the trained LSTM network. The network forward propagates and outputs a prediction sequence for the next 96 steps, including server heat load, ambient dry and wet bulb temperatures, and time-of-use electricity prices. To avoid error accumulation, the prediction is re-executed every 15 minutes to achieve rolling optimization.

[0069] The process then proceeds to step S3: optimization decision-making. The central control module calls the Gurobi solver, inputting the predicted sequence, the current state of charge (SOE) of the dual energy storage modules, and the physical parameters of the equipment into the MILP model. The objective function is as described in the invention, where the electricity price is the predicted value, and the energy consumption of the fan, water pump, and compressor is fitted using a quadratic polynomial to obtain the power model. The weight α of the start-stop penalty term is set to 0.05. Regarding constraints, the upper and lower limits of the supply air temperature are set to 24℃ and 18℃, respectively; the cooling balance equation ensures that the real-time cooling capacity is greater than or equal to the predicted heat load; the SOE of the energy storage unit is limited to between 0.1 and 0.9 to prevent overcharging and over-discharging. The solver provides the hourly strategy for the next 96 steps within 45 seconds, including Boolean variables for the start-stop of each module, fan speed percentage, water pump flow percentage, compressor frequency, and valve opening.

[0070] As one implementation method, this embodiment uses minimizing the total system operating cost as the objective function, and the formula is:

[0071]

[0072] Where t is the time step index within the scheduling period, N is the total number of time steps, and P elec (t) represents the predicted electricity price at time t, E fan (t), E pump (t), E comp (t) represents the predicted total energy consumption of all fans, pumps, and compressors in the system at time t. The predicted total energy consumption is calculated based on the power model of each component under its corresponding operating strategy. M switch (t) is the penalty term for the number of start-stop switching times of the main components of the system at time t, and α is the weight coefficient of the penalty term.

[0073] As one implementation method, the constraints include at least:

[0074] Data center supply air temperature constraint: The supply air temperature T after system processing at any time step t. supply (t) must satisfy T min ≤T supply (t)≤T max T min and T max For the minimum and maximum temperatures required for the safe operation of data center equipment, this embodiment uses 18°C ​​and 24°C;

[0075] Cooling balance constraint: At any time step t, the sum of the cooling capacity provided by the evaporative cooling module, the cooling capacity provided by the conventional refrigeration module, and the cooling capacity released by the cooling storage unit must be greater than or equal to the predicted value of the server's heat load.

[0076] Energy storage unit state constraints: The state of charge of cold energy storage units and heat energy storage units must be between their designed minimum and maximum capacities at any time step t.

[0077] As one implementation method, the system switches between at least four modes based on the optimal combination of operating strategies:

[0078] Mode 1: Full evaporative cooling mode. When the predicted ambient wet-bulb temperature meets the conditions for direct cooling, the evaporative cooling module is turned on to maximum power and the conventional cooling module is turned off.

[0079] Mode 2: Energy storage priority mode. When the electricity price forecast is at its lowest point and the data center heat load is low, the conventional cooling module is turned on to charge the cold storage unit, while evaporative cooling is used to meet the real-time cooling load.

[0080] Mode 3: Combined cooling mode. When evaporative cooling cannot meet the heat load demand independently, the evaporative cooling module and the conventional refrigeration module are turned on together or the cold energy storage unit is released to provide cooling in a coordinated manner.

[0081] Mode 4: Total Heat Recovery Mode. When the external ambient temperature is low and preheating of the fresh air is required, the total heat recovery module is activated to recover the waste heat of the exhaust air and store the excess heat in the heat storage unit.

[0082] Finally, the process proceeds to step S4: instruction execution. The central control module parses the solved strategy into Modbus-RTU and CANopen instructions, communicating with field inverters, electric valves, and compressor drivers via the RS-485 bus. For example, when the strategy requires the evaporative cooling module to operate independently, the instruction sets the frequency of the evaporative cooling fan to the predicted value, the frequency of the conventional refrigeration module compressor to 0Hz, and simultaneously shuts down the dual energy storage module's cold release pump. When the strategy requires combined cooling, the evaporative cooling fan maintains a high speed, the compressor frequency is linearly adjusted according to the cooling capacity gap, and the flow rate of the cold storage unit's cold release pump is dynamically adjusted based on the model output to achieve precise matching.

[0083] As one implementation method, the construction method of the composite prediction model includes:

[0084] Obtain a historical dataset with the same dimensions as that in step S1, which is at least one year old.

[0085] The historical dataset is divided to generate a training set, a validation set, and a test set.

[0086] Define the network structure of the Long Short-Term Memory (LSTM) network model, including the input layer dimension, the number of hidden layers, the number of neurons per layer, and the output layer dimension;

[0087] The long short-term memory network model is iteratively trained using the training set, and the model performance is evaluated using the validation set after each iteration. The hyperparameters of the model are then adjusted based on the evaluation results.

[0088] Once the model's performance on the validation set reaches the preset convergence criterion, training stops, and the test set is used to perform a final evaluation of the model's generalization ability. If the model passes the evaluation, it is then stored in the central control module.

[0089] To visually demonstrate the technical effects of this embodiment, continuous comparative tests were conducted from July 1st to July 31st, 2025. During the tests, the system operated according to the control method of this invention (experimental group), and its energy consumption was compared with that of a traditional PID feedback control scheme (control group). The average monthly PUE of the experimental group was 1.27, while that of the control group was 1.48; the average electricity cost during peak hours in the experimental group was 0.68 yuan / kWh, while that in the control group was 0.95 yuan / kWh; the number of compressor start-stop cycles in the experimental group was reduced from 52 times per day in the control group to 6 times per day, significantly reducing equipment wear. Simultaneously, the heat storage unit recovered 1120kWh of waste heat overnight for preheating fresh air the following morning, equivalent to a saving of 310kWh of electric heating power. The data clearly demonstrates that this invention, through predictive-optimization-execution closed-loop control, achieves a synergistic improvement in the energy efficiency and economy of the data center air conditioning system.

[0090] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0091] It should be noted that the components mentioned in the above embodiments are all general standard parts or components known to those skilled in the art. Their structures and principles can be learned by those skilled in the art through technical manuals or conventional experimental methods.

[0092] This invention has illustrated its principles and implementation methods using specific examples. The descriptions of these embodiments are merely illustrative of the method and its core ideas; furthermore, those skilled in the art will recognize that modifications may be made to the specific implementation methods and application scope based on the principles of this invention. Therefore, the content of this specification should not be construed as limiting the invention.

Claims

1. A data center evaporative cooling and total heat recovery dual-energy storage air conditioning system, characterized in that, include: The data acquisition module is used to collect server heat load data, return air temperature and humidity data inside the data center, and ambient dry-bulb temperature data, ambient wet-bulb temperature data, and time-of-use electricity price data outside the data center in real time. The evaporative cooling module includes a fresh air inlet and an air outlet, wherein the air outlet is connected to the cold aisle of the data center via an air supply duct. The total heat recovery module includes an exhaust inlet and a processed air outlet. The exhaust inlet is connected to the hot aisle of the data center through an exhaust duct, and the processed air outlet is connected to the fresh air inlet of the evaporative cooling module. A conventional refrigeration module includes a compressor, a condenser, a throttling device, and an evaporator. The evaporator is located in the first air supply duct and is used to deeply cool the air processed by the evaporative cooling module. The dual energy storage module includes a cold energy storage unit and a heat energy storage unit; the cold energy storage unit is connected to the evaporator of the conventional refrigeration module through a first supply and return water pipeline, and provides cold energy to the evaporator or stores the cold energy generated by the evaporator; the heat energy storage unit is connected to the total heat recovery module through a second supply and return water pipeline, and is used to store the heat recovered by the total heat recovery module; The central control module is connected to the data acquisition module, evaporative cooling module, total heat recovery module, conventional refrigeration module, and dual energy storage module. It is used to construct and store a composite prediction model based on historical data collected by the data acquisition module, predicting server heat load, ambient temperature and humidity, and time-of-use electricity prices within a preset time period. During system operation, it acquires real-time data collected by the data acquisition module and uses the composite prediction model to generate a predicted data sequence for the preset time period. Based on the predicted data sequence and the current energy storage state of the dual energy storage module, it solves for the objective function of minimizing the total system operating cost, generating coordinated operation control commands for the evaporative cooling module, total heat recovery module, conventional refrigeration module, and dual energy storage module. It then sends these coordinated operation control commands to control the start / stop, operating power, and switching of operating modes of each module.

2. The data center evaporative cooling and total heat recovery dual-energy storage air conditioning system according to claim 1, characterized in that, The data acquisition module includes: A first temperature and humidity sensor is deployed within the hot aisle of the data center. A second temperature and humidity sensor is deployed outside the data center; A load monitoring unit is connected to the server cluster management system interface to obtain real-time total power or CPU utilization as server thermal load data. The electricity price acquisition unit is connected to the data platform interface of the power grid company and is used to acquire time-of-use electricity price data.

3. The data center evaporative cooling and total heat recovery dual-energy storage air conditioning system according to claim 1, characterized in that, The total heat recovery module is a rotary total heat recovery unit. The rotor inside the rotary total heat recovery unit is made of moisture-absorbing material. During the rotation process, the rotor alternately contacts the exhaust air from the heat channel and the fresh air from the external environment, thereby achieving dual recovery of heat and humidity.

4. The data center evaporative cooling and total heat recovery dual-energy storage air conditioning system according to claim 1, characterized in that, The cold energy storage unit is a phase change energy storage tank, which is filled with encapsulated phase change material. The conventional refrigeration module operates during off-peak electricity prices and stores the cold energy in the phase change material through the circulating medium in the first supply and return water pipeline.

5. The data center evaporative cooling and total heat recovery dual-energy storage air conditioning system according to claim 1, characterized in that, The composite prediction model is a long short-term memory network model. The central control module trains the long short-term memory network model by constructing a multi-dimensional time series from historical server heat load data, environmental dry-bulb temperature data, environmental wet-bulb temperature data, and time-of-use electricity price data to obtain prediction capabilities.

6. A data center air conditioning control method for the system according to any one of claims 1-5, characterized in that, Includes the following steps: S1: Data acquisition and preprocessing, real-time acquisition of server heat load, return air temperature and humidity inside the data center, as well as dry and wet bulb temperatures of the external environment and the time-of-use electricity price at the current moment; and normalization and outlier filtering of the acquired raw data to form standardized real-time status data. S2: Multidimensional prediction, calling a pre-built composite prediction model, taking the standardized real-time status data and historical status data as input, and outputting server heat load prediction, environmental dry and wet bulb temperature prediction, and time-of-use electricity price prediction for multiple time steps within a future preset scheduling cycle. S3: Optimize decision-making, construct a function with the total operating cost of the system within the scheduling period as the optimization objective, take the predicted values ​​of each parameter generated by the multi-dimensional prediction step and the current state of charge of the dual energy storage module as constraints, solve the objective function through an optimization algorithm, and obtain the optimal combination of operating strategies for the evaporative cooling module, conventional refrigeration module, cold storage unit and heat storage unit at each time step within the scheduling period; S4: Instruction execution. Based on the optimal operating strategy combination, specific control instructions are generated and sent to the actuators of the evaporative cooling module, the total heat recovery module, the conventional refrigeration module, and the dual energy storage module. The control instructions include the fan speed, water pump flow rate, compressor start / stop and operating frequency, and pipeline valve opening degree of each module.

7. The data center air conditioning control method according to claim 6, characterized in that, In step S3, the formula with the objective function of minimizing the total system operating cost is as follows: Where t is the time step index within the scheduling period, N is the total number of time steps, and P elec (t) represents the predicted electricity price at time t, E fan (t), E pump (t), E comp (t) represents the predicted total energy consumption of all fans, pumps, and compressors in the system at time t. The predicted total energy consumption is calculated based on the power model of each component under its corresponding operating strategy. M switch (t) is the penalty term for the number of start-stop switching times of the main components of the system at time t, and α is the weight coefficient of the penalty term.

8. The data center air conditioning control method according to claim 7, characterized in that, In step S3, the constraints include at least: Data center supply air temperature constraint: The supply air temperature T after system processing at any time step t. supply (t) must satisfy T min ≤T supply (t)≤T max T min and T max The lower and upper limits of temperature required for the safe operation of data center equipment; Cooling balance constraint: At any time step t, the sum of the cooling capacity provided by the evaporative cooling module, the cooling capacity provided by the conventional refrigeration module, and the cooling capacity released by the cooling storage unit must be greater than or equal to the predicted value of the server's heat load. Energy storage unit state constraints: The state of charge of cold energy storage units and heat energy storage units must be between their designed minimum and maximum capacities at any time step t.

9. The data center air conditioning control method according to claim 6, characterized in that, In step S4, based on the optimal combination of operating strategies, the system switches between at least four modes: Mode 1: Full evaporative cooling mode. When the predicted ambient wet-bulb temperature meets the conditions for direct cooling, the evaporative cooling module is turned on to maximum power and the conventional cooling module is turned off. Mode 2: Energy storage priority mode. When the electricity price forecast is at its lowest point and the data center heat load is low, the conventional cooling module is turned on to charge the cold storage unit, while evaporative cooling is used to meet the real-time cooling load. Mode 3: Combined cooling mode. When evaporative cooling cannot meet the heat load demand independently, the evaporative cooling module and the conventional refrigeration module are turned on together or the cold energy storage unit is released to provide cooling in a coordinated manner. Mode 4: Total Heat Recovery Mode. When the external ambient temperature is low and preheating of the fresh air is required, the total heat recovery module is activated to recover the waste heat of the exhaust air and store the excess heat in the heat storage unit.

10. The data center air conditioning control method according to claim 6, characterized in that, The method for constructing the composite prediction model includes: Obtain a historical dataset with the same dimensions as that in step S1, spanning at least the past year. The historical dataset is divided to generate a training set, a validation set, and a test set. Define the network structure of the Long Short-Term Memory (LSTM) network model, including the input layer dimension, the number of hidden layers, the number of neurons per layer, and the output layer dimension; The long short-term memory network model is iteratively trained using the training set, and the model performance is evaluated using the validation set after each iteration. The hyperparameters of the model are then adjusted based on the evaluation results. Once the model's performance on the validation set reaches the preset convergence criterion, training stops, and the test set is used to perform a final evaluation of the model's generalization ability. If the model passes the evaluation, it is then stored in the central control module.

Images