Efficient machine room global optimization energy-saving control algorithm based on energy flow conservation principle

Abstract

The invention relates to the technical field of machine room energy conservation, in particular to a high-efficiency machine room global optimization energy-saving control algorithm based on an energy flow conservation principle, and the method comprises the steps: constructing a cold source system mechanism model based on the energy flow conservation principle, the cold source system mechanism model comprises but is not limited to a refrigerator energy efficiency model, a water pump lift and efficiency model and a cooling tower heat dissipation model; constructing an equipment machine learning model, integrating various machine learning methods (such as a random forest, a gradient elevator and LightGBM), and searching an optimal regression model in combination with a voting device; inputting current working condition parameters including real-time load and outdoor meteorological parameters, traversing all possible equipment combinations and parameter combinations, calling the cold source system mechanism model and the equipment machine learning model, and performing global optimization of the equipment combinations by taking minimization of total energy consumption of the system as a target function; and generating an optimal control parameter according to an optimization result, and issuing and executing the optimal control parameter through the Internet of Things platform.

Description

Technical Field

[0001] The present invention relates to the technical field of computer room energy saving, and in particular to a high-efficiency computer room global optimization energy-saving control algorithm based on the energy flow conservation principle. Background Art

[0002] In my country, building energy consumption is increasing year by year. Central air conditioning systems account for a significant portion of total building energy consumption, ranging from 35% to 55%. Of this, refrigeration room equipment (including chillers, refrigeration and cooling water pumps, and cooling towers) accounts for approximately 60% to 70% of central air conditioning system energy consumption. The annual electricity consumption of refrigeration room equipment accounts for 25% to 40% of the building's total electricity consumption.

[0003] As energy consumption becomes increasingly prominent, efficient computer room energy-saving control has become a focus of attention. At present, research on efficient computer room energy-saving control mainly focuses on the following aspects: the classification of energy-saving technologies for refrigeration rooms, the application of intelligent energy-saving control technologies for refrigeration rooms, and practical cases of efficient computer rooms in different industries. However, these studies often lack systematic integration, making it difficult to form a comprehensive and efficient computer room energy-saving control solution. At the same time, model optimization and non-model optimization methods in intelligent energy-saving control technologies for refrigeration rooms also have their own characteristics. Model optimization methods such as intelligent system optimization, load forecasting optimization, and digital twin optimization have achieved certain results in practical applications, but they also face some challenges, such as model accuracy and adaptability. Non-model optimization methods require more practical experience and data analysis to determine the best energy-saving control strategy. Summary of the Invention

[0004] The main purpose of the present invention is to provide a high-efficiency computer room global optimization energy-saving control algorithm based on the principle of energy flow conservation, which effectively solves the above-mentioned problems mentioned in the background technology.

[0005] The technical solutions of the present invention are as follows:

[0006] A high-efficiency global optimization energy-saving control algorithm for computer rooms based on the energy flow conservation principle is proposed. The method includes the following steps:

[0007] S1. Construct a cold source system mechanism model based on the principle of energy flow conservation, wherein the cold source system mechanism model includes but is not limited to a chiller energy efficiency model, a water pump head and efficiency model, and a cooling tower heat dissipation model;

[0008] S2. Build a machine learning model for equipment whose inputs are key variables affecting the efficiency of individual equipment and whose outputs are the calculated energy efficiency of the chiller, the head and efficiency of the water pump, and the temperature below the cooling tower. Use an ensemble learning method combining random forest, gradient boosting, and LightGBM models to generate the optimal regression model through voting integration.

[0009] S3. Input current operating parameters, including real-time load and outdoor meteorological parameters, traverse all possible equipment combinations and parameter combinations, call the cooling system mechanism model and equipment machine learning model, and perform global optimization of equipment combinations with minimizing total system energy consumption as the objective function;

[0010] S4. Generate optimal control parameters based on the optimization results, including the operating parameter combination of the chiller, chilled water pump, cooling water pump and cooling tower, and issue them for execution through the Internet of Things platform.

[0011] A further improvement of the present invention is that the energy efficiency model of the chiller in S1 is based on the chiller operating load rate PLR, the evaporator outlet water temperature T chilled,supply , Cooling water return temperature T cooling,return , cooling water flow L cooling , establish the cooling machine energy efficiency COP chiller The mathematical relationship is:

[0012] COP chiller =F(PLR,T chilled,supply ,T cooling,return ,L cooling )(1)

[0013] Cooling machine power Pe chiller According to the real-time cooling capacity Q l Calculated as:

[0014]

[0015] A further improvement of the present invention is that the cooling tower heat dissipation model in S1 is based on the outdoor wet bulb temperature T wb , Cooling water supply and return water temperature difference dT cooling , cooling water flow L cooling , cooling tower fan speed f tower，fan , establish the cooling tower lower tower temperature T tower,out The mathematical relationship is:

[0016] T tower,out =F(T wb ,dT cooling ,L cooling ,f tower，fan )(3)

[0017] And based on the cooling tower fan power Pe tower,fan and cooling tower fan frequency f tower，fan The current cooling tower fan operating power Pe is calculated by the functional relationship tower,fan =F(f tower，fan ).

[0018] A further improvement of the present invention is that the pump head and efficiency model in S1 is based on the head H of the chilled water pump. chilled,pump , Cooling water pump head H cooling,pump , Efficiency of the chilled water pump Eta chilled,pump , cooling water pump efficiency Eta cooling,pump , expressed as:

[0019] H chilled,pump =F(L chilled,pump ) (4)

[0020] H cooling,pump =F(L cooling,pump ) (5)

[0021] Eta chilled,pump =F(L chilled,pump ) (6)

[0022] Eta cooling, pump =F(L cooling,pump ) (7)

[0023] Chilled water pump operating power Pe chilled,pump , Cooling water pump operating power Pe cooling,pump Expressed as:

[0024]

[0025] Where ρ is the density of water, g is the acceleration due to gravity, and η is the efficiency of the water pump.

[0026] A further improvement of the present invention is that the ensemble learning method in S2 adopts a soft voting mechanism, combines random forest, gradient boosting machine and LightGBM models, optimizes the hyperparameters of each model through cross-validation, and performs weighted averaging on the prediction results. The mechanism principle constraint is introduced through a custom loss function, with the mean absolute error as the main loss function and the other indicators weighted as auxiliary.

[0027] A further improvement of the present invention is that the constraints of the global optimization in S3 include: satisfying the minimum and maximum inlet temperatures of the cooling water of the chiller: T tower,out,min ≤T tower,out ≤T tower,out,max ; Meet the minimum load rate of the cooling machine to prevent cooling machine surge: PLR≤PLR min ;Meet the minimum flow rate of evaporator and condenser: L chilled,min <L chilled <L chilled,max ;L cooling,min <L cooling <L cooling,max .

[0028] A further improvement of the present invention is that the global optimization of equipment combination in S3 includes: the number of operating chillers and load distribution; the start and stop combination and flow regulation of the chilled water pump and the cooling water pump; the number of operating cooling towers and the fan speed matching.

[0029] A further improvement of the present invention is that the device machine learning model collects current data and fills it into the historical database required for each regression, and periodically trains the model offline based on the newly added operating data.

[0030] A further improvement of the present invention is that the safety first principle ensures that the terminal cooling demand is the highest priority, and dynamically adjusts the control parameters through a real-time monitoring and feedback mechanism.

[0031] The technical effects of the present invention are as follows:

[0032] The present invention is based on a high-efficiency global optimization energy-saving control algorithm for a computer room based on the principle of conservation of energy flow. Starting from the equipment and system itself, the equipment and system models are constructed in parallel with the principle of conservation of energy flow and the requirements for safe operation. Compared with conventional data-driven system models, the present invention has higher interpretability and traceability. By identifying the characteristics of the controlled dynamic process under variable load conditions, the operating parameters of the system are adaptively calculated and adjusted, and the optimal operating conditions of equipment such as chillers, chilled water pumps, cooling water pumps, and cooling tower fans are analyzed to achieve global optimization control of the system to obtain the best control effect, thereby matching energy supply with energy demand and minimizing the total energy consumption of the air-conditioning system. Different input loads and outdoor working conditions are achieved, and the current energy efficiency optimal control parameters are issued in combination with the actual operating status of the equipment. Closed-loop control is performed based on the Internet of Things collection and issuance platform, providing improvement measures and decision-making basis for equipment analysis and energy-saving optimization control of cold source systems in multiple application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0034] Figure 1 This is a flow chart of a high-efficiency computer room global optimization energy-saving control algorithm based on the energy flow conservation principle of the present invention;

[0035] Figure 2 This is an application topology diagram of a high-efficiency computer room global optimization energy-saving control algorithm based on the energy flow conservation principle of the present invention;

[0036] Figure 3 A detailed diagram of the global optimization process of a high-efficiency computer room global optimization energy-saving control algorithm based on the energy flow conservation principle of the present invention;

[0037] Figure 4This is a diagram showing the equipment machine learning training results of an efficient global optimization energy-saving control algorithm for a computer room based on the principle of energy conservation of the present invention. DETAILED DESCRIPTION

[0038] The present invention aims to propose an efficient global optimization and energy-saving control algorithm for computer rooms based on the principle of energy conservation. Starting from the equipment and system itself, the equipment and system models are constructed in parallel with the principle of energy conservation and safe operation requirements. Compared with conventional data-driven system models, the algorithm has higher interpretability and traceability. By identifying the characteristics of the controlled dynamic process under variable load conditions, the operating parameters of the system are adaptively calculated and adjusted, and the optimal operating conditions of equipment such as chillers, chilled water pumps, cooling water pumps, and cooling tower fans are analyzed to achieve global optimization control of the system to obtain the best control effect, thereby matching energy supply with energy demand and minimizing the total energy consumption of the air-conditioning system. Different input loads and outdoor working conditions are achieved, and the current energy efficiency optimal control parameters are issued in combination with the actual operating status of the equipment. Closed-loop control is performed based on the Internet of Things collection and issuance platform, providing improvement measures and decision-making basis for equipment analysis and energy-saving optimization control of cold source systems in multiple application scenarios.

[0039] Example 1:

[0040] This embodiment proposes a high-efficiency global optimization energy-saving control algorithm for computer rooms based on the principle of energy conservation. Figure 1-Figure 3 As shown, the following specific steps are included:

[0041] S1. Construct a cold source system mechanism model based on the principle of energy flow conservation, wherein the cold source system mechanism model includes but is not limited to a chiller energy efficiency model, a water pump head and efficiency model, and a cooling tower heat dissipation model;

[0042] S2. Build a machine learning model for equipment whose inputs are key variables affecting the efficiency of individual equipment and whose outputs are the calculated energy efficiency of the chiller, the head and efficiency of the water pump, and the temperature below the cooling tower. Use an ensemble learning method combining random forest, gradient boosting, and LightGBM models to generate the optimal regression model through voting integration.

[0043] S3. Input current operating parameters, including real-time load and outdoor meteorological parameters, traverse all possible equipment combinations and parameter combinations, call the cooling system mechanism model and equipment machine learning model, and perform global optimization of equipment combinations with minimizing total system energy consumption as the objective function;

[0044] S4. Generate optimal control parameters based on the optimization results, including the operating parameter combination of the chiller, chilled water pump, cooling water pump and cooling tower, and issue them for execution through the Internet of Things platform.

[0045] In this embodiment, the energy efficiency model of the chiller in S1 is based on the chiller operating load rate PLR, the evaporator outlet water temperature T chilled,supply , Cooling water return temperature T cooling,return , cooling water flow L cooling , establish the cooling machine energy efficiency COP chiller The mathematical relationship is:

[0046] COP chiller =F(PLR,T chilled,supply ,T cooling,return ,L cooling )(1)

[0047] Cooling machine power Pe chiller According to the real-time cooling capacity Q l Calculated as:

[0048]

[0049] In this embodiment, the cooling tower heat dissipation model in S1 is based on the outdoor wet bulb temperature T wb , Cooling water supply and return water temperature difference dT cooling , cooling water flow L cooling , cooling tower fan speed f tower，fan , establish the cooling tower lower tower temperature T tower,out The mathematical relationship is:

[0050] T tower,out =F(T wb ,dT cooling ,L cooling ,f tower，fan )(3)

[0051] And based on the cooling tower fan power Pe tower,fan and cooling tower fan frequency f tower，fan The current cooling tower fan operating power Pe is calculated by the functional relationship tower,fan =F(f tower，fan ).

[0052] In this embodiment, the pump head and efficiency model in S1 is based on the head H of the chilled water pump. chilled,pump , Cooling water pump head H cooling,pump , Efficiency of the chilled water pump Eta chilled,pump , cooling water pump efficiency Eta cooling,pump , expressed as:

[0053] H chilled,pump =F(L chilled,pump )(4)

[0054] H cooling,pump =F(L cooling,pump )(5)

[0055] Eta chilled,pump =F(L chilled,pump )(6)

[0056] Eta cooling,pump =F(L cooling,pump )(7)

[0057] Chilled water pump operating power Pe chilled,pump , Cooling water pump operating power Pe cooling,pump Expressed as:

[0058]

[0059] Where ρ is the density of water, g is the acceleration due to gravity, and η is the efficiency of the water pump.

[0060] In this example, key historical operational data from a refrigeration plant is collected. Abnormal data identification and processing algorithms are used to improve the quality of the data, preventing issues such as poor model generalization and low accuracy caused by data quality issues during subsequent modeling. Subsequently, a subset of characteristic variables influencing chiller efficiency, water pump characteristics, and cooling tower heat dissipation capacity are analyzed and selected to construct a machine learning model for the equipment. The model inputs are the key variables that influence the efficiency of individual equipment, and the model outputs are the calculated energy efficiency of the chiller, the head and efficiency of the water pump, and the lower tower temperature of the cooling tower.

[0061] A comparison of common machine learning methods revealed that models built using some methods achieved good training results and high fitting accuracy on the test set, but were prone to overfitting and exhibited large deviations in prediction results for areas not included in the dataset, resulting in poor generalization. Therefore, by introducing the Voting Regressor and a custom loss function, combined with conventional machine learning models, the approach integrates multiple machine learning models and adds a penalty for regression mechanism rationality to the original loss function. This prevents the tendency to abandon the original mechanism principle in pursuit of lower error during training. By integrating the regression results of multiple models, the stability and robustness of the model can be improved. Even if a model performs poorly in certain situations, other models can intervene to compensate for the shortfall and improve overall regression performance.

[0062] Ensemble learning is a method that improves prediction performance by combining multiple models (base learners). Voting is a form of ensemble learning that aggregates the predictions of different models to generate a final prediction. The following describes the basic principles and steps for training three models: Random Forest (RF), Gradient Boosting Machine (XGBoost), and LightGBM, and then using Voting to ensemble them into the optimal model. The specific model training steps are as follows:

[0063] (1) Model selection and training

[0064] Random Forest (RF): is an ensemble of multiple decision trees, each trained independently on a random subset of the data. Random Forest can reduce overfitting and provide higher accuracy and stability.

[0065] Gradient Boosting Machine (XGBoost): is a boosting method that iteratively adds new models to correct errors in existing models. XGBoost is known for its efficient computational power and excellent generalization ability.

[0066] LightGBM: It is a gradient boosting framework developed by Microsoft. It uses a histogram algorithm and a leaf growing strategy, making it more efficient and faster when processing large-scale data.

[0067] (2) Training process: For each selected model: Prepare training data. Adjust hyperparameters based on the characteristics of each model. Train each of the three models using the training data. Verify and optimize model performance, for example, through cross-validation.

[0068] (3) Voting Integration

[0069] There are two main types of voting integration: hard voting and soft voting.

[0070] Hard Voting: For classification problems, each model predicts the class of the sample and takes the class with the highest number of occurrences as the final prediction. For regression problems, the median or mode of all model predictions is usually taken.

[0071] Soft Voting: For classification problems, each model gives a probability distribution of the class, and the class with the highest probability is taken as the final prediction. For regression problems, the average of all model predictions can be taken.

[0072] (4) Determination of training objective function

[0073] In machine learning, choosing an appropriate objective function (also known as a loss function or cost function) is crucial for model training and directly impacts the model's training results and ultimate performance. The objective function defines a measure of the difference between the model's predicted value and the true value. The model training process is the process of minimizing this difference. Different tasks and algorithms use different types of objective functions. In regression tasks, mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), and coefficient of determination (R²) are commonly used. In this regression model, mean absolute error is used as the primary loss function, with other weighted indicators serving as supplementary factors. To ensure that the regression results do not deviate from the actual mechanism, a penalty function is set to prevent any deviation between the trained model's calculations and the actual mechanism.

[0074] (5) Selection of the optimal model

[0075] During the integration process, the performance of the integrated model can be evaluated using the validation set, using metrics such as accuracy, F1 score, and mean squared error. If soft voting is used, the probability weights of the predictions from different models can also be considered, assigning different weights to different models to further optimize the performance of the integrated model.

[0076] In this embodiment, the ensemble learning method in S2 adopts a soft voting mechanism, combined with random forest, gradient boosting machine and LightGBM models, optimizes the hyperparameters of each model through cross-validation, and performs weighted averaging on the prediction results. The mechanism principle constraint is introduced through a custom loss function, with the mean absolute error as the main loss function, and the other indicators are weighted as auxiliary.

[0077] Based on the equipment model, a complete calculation mechanism framework for the entire refrigeration room system needs to be established. After inputting the current real-time load and outdoor weather parameters, the framework calls upon various calculation sub-models through the full system mechanism. Through iterative optimization of parameters, it finds all possible equipment and parameter combinations that can meet the load under the same operating conditions. Depending on the complexity of the system, tens of thousands or even hundreds of thousands of combinations are traversed to ultimately arrive at the optimal solution.

[0078] In this embodiment, assuming that the system has a refrigeration unit, b chilled water pumps, c cooling water pumps, and d cooling towers operating in conjunction, the energy-saving control calculation process for the entire refrigeration room is described as follows:

[0079]

[0080] T wb ,dT cooling ,…f tower，fan ;

[0081] L chilled,pump ,H chilled,pump …η chilled,pump ;

[0082] L cooling,pump ,H cooling,pump …η cooling,pump )

[0083] Safety principles must also be considered when performing global optimization calculations. The constraints are as follows: Minimum and maximum inlet temperatures of the chiller cooling water: T tower,out,min ≤T tower,out ≤T tower,out,max ; Meet the minimum load rate of the cooling machine to prevent cooling machine surge: PLR≤PLR min ;Meet the minimum flow rate of evaporator and condenser: L chilled,min <L chilled <L chilled,max ;L cooling,min <L cooling <L cooling,max .

[0084] In this embodiment, the global optimization of equipment combination in S3 includes: the number of operating chillers and load distribution; the start-stop combination and flow regulation of the chilled water pump and the cooling water pump; the number of operating cooling towers and the fan speed matching.

[0085] Global optimization of efficient computer room based on the principle of energy conservation: Figure 1 As shown in FIG, by inputting the actual indoor load and outdoor meteorological parameters, the global optimization main loop can be entered according to the input conditions.

[0086] The optimization calculation results include two parts: one is the optimization of the startup combination of the chiller, water pump and cooling tower, and the other is the optimization of the parameters of the cooling water temperature difference and approximation. The outer layer of the main loop is the cooling water temperature difference (the supply and return temperature difference of the cooling water in the main pipe) and the approximation of the cooling tower (the difference between the tower temperature below the cooling tower and the outdoor wet-bulb temperature). The two are used as optimization parameters. Before the calculation, the calculation range of the parameters is first determined. Within the range, the parameter values ​​that need to be traversed and calculated are determined by fixing the calculation step length. The initial value, final value and optimization step length of the optimization are obtained by discretizing the two, and the parameter set N for traversal optimization is generated. Taking the cooling water temperature difference as an example, the method for determining the initial and final values ​​is to design the cooling water temperature difference as ±a℃, with a step length of b℃. The values ​​of a and b are determined by trial parameter setting, ensuring the calculation efficiency while ensuring the calculation results. The values ​​of a and b are finally determined through test experiments. Then, according to the input load, the chillers that can currently meet the demand are arranged and combined, and the available chiller combination A is generated according to different combinations. In order to achieve the chilled and cooling water circulating water flow required for the operation of the chiller, the corresponding available chilled water pump combination B and cooling water pump combination C under different combinations of chiller A can be obtained.

[0087] Finally, the currently available cooling tower combination D is obtained according to the cooling water flow rate under each combination, completing the preparation before the optimization process.

[0088] The equipment models (chiller energy efficiency model, water pump head / efficiency model, cooling tower heat dissipation model) collect current data and fill it into the historical database required for their respective regressions. The models are trained offline in stages based on the newly added operating data to achieve the timeliness and multi-scenario applicability of the models.

[0089] After completing the optimization sequence of parameter + equipment combination in the above steps, the total system power corresponding to each parameter and equipment combination under the current corresponding working conditions is calculated by calling the main loop, such as Figure 2 The current optimal solution is shown and recorded.

[0090] If the power of the next calculation is less than the previous calculation result, the current calculation will replace the previous one. The calculation will be repeated iteratively until all possible parameter + equipment combinations under the current working conditions are traversed. Finally, the optimal solution in the historical calculation results and the corresponding parameter settings (cooling water temperature difference and cooling tower approximation) and equipment combination (chiller, chilled water pump, cooling water pump, cooling tower) are obtained.

[0091] In this embodiment, the device machine learning model collects current data and fills it into the historical database required for each regression, and periodically trains the model offline based on the newly added operating data.

[0092] In this embodiment, the safety first principle is to ensure that the terminal cooling demand is the highest priority, and dynamically adjust the control parameters through real-time monitoring and feedback mechanism.

[0093] Example 2:

[0094] This embodiment provides an electronic device, including: a processor and a memory, wherein the memory stores a computer program that can be called by the processor; the processor executes the above-mentioned high-efficiency computer room global optimization energy-saving control algorithm based on the principle of energy conservation by calling the computer program stored in the memory.

[0095] The electronic device may have relatively large differences due to different configurations or performances, and may include one or more processors (Central Processing Units, CPU) and one or more memories, wherein the memory stores at least one computer program, which is loaded and executed by the processor to implement a high-efficiency global optimization energy-saving control algorithm for a computer room based on the principle of energy conservation provided by the above-mentioned method embodiment. The electronic device may also include other components for realizing the functions of the device. For example, the electronic device may also have components such as a wired or wireless network interface and an input / output interface for data input and output. This embodiment will not be described in detail here.

[0096] Those skilled in the art will appreciate that the present invention may be implemented as a system, method, or computer program product. Therefore, the present disclosure may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, the present invention may be implemented in the form of a computer program product embodied in one or more computer-readable media containing computer-readable program code.

[0097] Any combination of one or more computer-readable media may be employed. A computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0098] The present invention is described with reference to flowcharts and block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process or block in the flowcharts and block diagrams, as well as combinations of processes and blocks in the flowcharts or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts or block diagrams. Figure 1 A process or multiple processes and boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0099] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 A process or multiple processes and boxes Figure 1 A step that specifies a function in one or more boxes.

[0100] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which are all protected by the present invention.

Claims

1. A high-efficiency global optimization energy-saving control algorithm for computer rooms based on the principle of energy conservation, characterized by: The specific steps include: S1. Construct a cold source system mechanism model based on the principle of energy flow conservation, wherein the cold source system mechanism model includes but is not limited to a chiller energy efficiency model, a water pump head and efficiency model, and a cooling tower heat dissipation model; S2. Build a machine learning model for equipment whose inputs are key variables affecting the efficiency of individual equipment and whose outputs are the calculated energy efficiency of the chiller, the head and efficiency of the water pump, and the temperature below the cooling tower. Use an ensemble learning method combining random forest, gradient boosting, and LightGBM models to generate the optimal regression model through voting integration. S3. Input current operating parameters, including real-time load and outdoor meteorological parameters, traverse all possible equipment combinations and parameter combinations, call the cooling system mechanism model and equipment machine learning model, and perform global optimization of equipment combinations with minimizing total system energy consumption as the objective function; S4. Generate optimal control parameters based on the optimization results, including the operating parameter combination of the chiller, chilled water pump, cooling water pump and cooling tower, and issue them for execution through the Internet of Things platform.

2. The high-efficiency global optimization energy-saving control algorithm for a computer room based on the energy flow conservation principle according to claim 1 is characterized by: The energy efficiency model of the chiller in S1 is based on the chiller operating load rate PLR, the evaporator outlet water temperature T chilled,supply , Cooling water return temperature T cooling,return , cooling water flow L cooling , establish the cooling machine energy efficiency COP chiller The mathematical relationship is: COP chiller =F(PLR,T chilled,supply ,T cooling,return ,L cooling ) (1) Cooling machine power Pe chiller According to the real-time cooling capacity Q l Calculated as:

3. The high-efficiency computer room global optimization energy-saving control algorithm based on the energy flow conservation principle according to claim 2 is characterized by: The cooling tower heat dissipation model in S1 is based on the outdoor wet-bulb temperature T wb , Cooling water supply and return water temperature difference dT cooling , cooling water flow L cooling , cooling tower fan speed f tower，fan , establish the cooling tower lower tower temperature T tower,out The mathematical relationship is: T tower,out =F(T wb ,dT cooling ,L cooling ,f tower，fan ) (3) And based on the cooling tower fan power Pe tower,fan and cooling tower fan frequency f tower，fan The current cooling tower fan operating power Pe is calculated by the functional relationship tower,fan =F(f tower，fan ).

4. The high-efficiency global optimization energy-saving control algorithm for a computer room based on the energy flow conservation principle according to claim 3 is characterized by: The pump head and efficiency model in S1 is based on the chilled water pump head H chilled,pump , Cooling water pump head H cooling,pump , Efficiency of the chilled water pump Eta chilled,pump , cooling water pump efficiency Eta cooling,pump , expressed as: H chilled,pump =F(L chilled,pump ) (4) H cooling,pump =F(L cooling,pump ) (5) And chilled,pump =F(L chilled,pump ) (6) And cooling,pump =F(L cooling,pump ) (7) Chilled water pump operating power Pe chilled,pump , Cooling water pump operating power Pe cooling,pump Expressed as: Where ρ is the density of water, g is the acceleration due to gravity, and η is the efficiency of the water pump.

5. The high-efficiency computer room global optimization energy-saving control algorithm based on the energy flow conservation principle according to claim 4 is characterized by: The ensemble learning method in S2 adopts a soft voting mechanism, combined with random forest, gradient boosting machine and LightGBM models, optimizes the hyperparameters of each model through cross-validation, and performs weighted average on the prediction results. It also introduces mechanism principle constraints through a custom loss function, using the mean absolute error as the main loss function and the other indicators as weighted auxiliary.

6. The high-efficiency computer room global optimization energy-saving control algorithm based on the energy flow conservation principle according to claim 5 is characterized by: The constraints of the global optimization in S3 include: satisfying the minimum and maximum inlet temperatures of the cooling water of the chiller: T tower,out,min ≤T tower,out ≤T tower,out,max ; Meet the minimum load rate of the cooling machine to prevent cooling machine surge: PLR≤PLR min ;Meet the minimum flow rate of evaporator and condenser: L chilled,min <L chilled <L chilled,max ;L cooling,min <L cooling <L cooling,max .

7. The high-efficiency computer room global optimization energy-saving control algorithm based on the energy flow conservation principle according to claim 6 is characterized by: The global optimization of equipment combination in S3 includes: the number of operating chillers and load distribution; the start-stop combination and flow regulation of the chilled water pump and the cooling water pump; and the number of operating cooling towers and fan speed matching.

8. The high-efficiency global optimization energy-saving control algorithm for a computer room based on the energy flow conservation principle according to claim 7 is characterized by: The device machine learning model collects current data and fills it into the historical database required for each regression, and periodically trains the model offline based on the newly added operating data.

9. The high-efficiency computer room global optimization energy-saving control algorithm based on the energy flow conservation principle according to claim 8 is characterized by: The safety-first principle ensures that the terminal cooling demand is the highest priority, and dynamically adjusts the control parameters through real-time monitoring and feedback mechanisms.

Images