AI-based Energy-saving Control System and Method for Water-cooled Central Air Conditioner

Abstract

This application relates to the field of air-conditioning control technology. Specifically, it discloses an AI-based energy-saving control system and method for water-cooled central air conditioners. It uses artificial intelligence technology based on deep learning to comprehensively monitor and perform time-series analysis on the operation data of the refrigeration machine room of the water-cooled central air conditioner, so as to discover the time-series correlation change patterns of multi-source operation parameters, thereby obtaining a comprehensive characterization of the operation state of the machine room. At the same time, in combination with the set value of the chilled water outlet temperature of the chiller, through cross-modal interaction analysis of the set chilled water outlet temperature and the operation state of the machine room, the operation state response characteristics of the refrigeration machine room at the current set temperature are discovered, and on this basis, intelligent recommendation of the chilled water outlet temperature is carried out. In this way, dynamic optimization control of the water-cooled central air-conditioning system can be effectively achieved to adapt to changes in the external environment and internal load, thereby significantly improving the operation efficiency and energy efficiency of the water-cooled central air-conditioning system.

Description

Technical Field

[0001] This application relates to the technical field of air-conditioning control, and more specifically, to an AI-based energy-saving control system and method for water-cooled central air conditioners. Background Art

[0002] With the growth of global energy demand and the increasingly severe environmental problems, improving energy utilization efficiency has become the focus of common concern in various industries. As one of the main components of building energy consumption, the energy efficiency optimization of central air-conditioning systems is of great significance for achieving the goals of energy conservation and emission reduction. Especially for water-cooled central air-conditioning systems used in large commercial buildings, industrial plants, data centers and other places, due to their large scale, high load and high energy consumption, therefore, how to effectively improve the energy efficiency of water-cooled central air-conditioning systems and achieve energy conservation and emission reduction has become an important issue in current research and practice.

[0003] Traditionally, the control of water-cooled central air-conditioning systems usually adopts fixed set values and simple PID control strategies to achieve basic temperature control purposes. However, these control methods generally lack flexibility and are usually difficult to adapt to complex and changeable operating environments. For example, the control method based on fixed set values usually requires manual setting of fixed chilled water outlet temperatures and cannot be dynamically adjusted according to external environmental changes and internal load demands, resulting in low system operating efficiency. Although PID control can adjust the water-cooled central air-conditioning system to a certain extent, in a water-cooled central air-conditioning system, there are complex coupling relationships among multiple variables such as chilled water outlet temperature, cooling water temperature, and indoor temperature. The PID controller usually can only independently control a single variable and cannot effectively handle these coupling effects, resulting in limited adjustment effects and difficulty in achieving optimal control.

[0004] In recent years, with the rapid development of artificial intelligence technology, it has become possible to use AI technology to control the energy saving of water-cooled central air-conditioning systems. Therefore, an AI-based energy-saving control system and method for water-cooled central air conditioners are expected. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. Embodiments of this application provide an AI-based energy-saving control system and method for water-cooled central air conditioners. It uses artificial intelligence technology based on deep learning to comprehensively monitor and perform time-series analysis on the operation data of the refrigeration machine room of the water-cooled central air conditioner, so as to discover the time-series correlation change patterns of multi-source operation parameters, thereby obtaining a comprehensive representation of the operation state of the machine room. At the same time, in combination with the set value of the chilled water outlet temperature of the chiller, through cross-modal interaction analysis of the set temperature of the chilled water outlet and the operation state of the machine room, the operation state response characteristics of the refrigeration machine room at the current set temperature are discovered, and on this basis, intelligent recommendation of the chilled water outlet temperature is carried out. In this way, dynamic optimization control of the water-cooled central air conditioning system can be effectively realized to adapt to changes in the external environment and internal load, thereby significantly improving the operation efficiency and energy efficiency of the water-cooled central air conditioning system.

[0006] According to one aspect of this application, an AI-based energy-saving control method for water-cooled central air conditioners is provided, which includes:

[0007] Collect the operation data of the refrigeration machine room of the water-cooled central air conditioner, and at the same time obtain the set value of the chilled water outlet temperature of the chiller;

[0008] Extract the time-series pattern features of the operation data of the refrigeration machine room of the water-cooled central air conditioner to obtain a semantic encoding feature map of the operation state time series of the machine room;

[0009] Perform one-hot encoding on the set value of the chilled water outlet temperature of the chiller to obtain a one-hot encoding vector of the set temperature of the chilled water outlet;

[0010] Perform cross-modal interaction optimization based on the prompt information on the one-hot encoding vector of the set temperature of the chilled water outlet and the semantic encoding feature map of the operation state time series of the machine room to obtain a joint encoding feature map of the operation state - set temperature of the chilled water outlet response;

[0011] Based on the joint encoding feature map of the operation state - set temperature of the chilled water outlet response, determine the recommended value of the chilled water outlet temperature;

[0012] Update the set value of the chilled water outlet temperature of the chiller to the recommended value of the chilled water outlet temperature.

[0013] According to another aspect of this application, an AI-based energy-saving control system for water-cooled central air conditioners is provided, which includes:

[0014] A data acquisition module for collecting the operation data of the refrigeration machine room of the water-cooled central air conditioner and obtaining the set value of the chilled water outlet temperature of the chiller at the same time;

[0015] A feature extraction module, used to extract the time series pattern features of the operation data of the water-cooled central air-conditioning refrigeration room to obtain a semantic coding feature map of the time series pattern of the room operation status;

[0016] A one-hot encoding module, used for performing one-hot encoding on the set value of the chilled water outlet temperature of the chiller to obtain a one-hot encoding vector for setting the chilled water outlet temperature;

[0017] A cross-modal interactive optimization module, used for performing cross-modal interactive optimization on the chilled water outlet temperature setting unique hot encoding vector and the machine room operation status temporal pattern semantic encoding feature map based on prompt information to obtain a machine room operation status-chilled water outlet temperature setting response joint encoding feature map;

[0018] A temperature value decision module, used to determine a recommended value of the chilled water outlet temperature based on the equipment room operation status-chilled water outlet temperature setting response joint coding characteristic diagram;

[0019] The temperature value updating module is used to update the set value of the chilled water outlet temperature of the chiller to the recommended value of the chilled water outlet temperature.

[0020] Compared with the prior art, the AI-based water-cooled central air-conditioning energy-saving control system and method provided by the present application uses artificial intelligence technology based on deep learning to comprehensively monitor and time-series analyze the operating data of the water-cooled central air-conditioning refrigeration room, so as to mine the time-series correlation change pattern of multi-source operating parameters, thereby obtaining a comprehensive representation of the operating status of the room, and at the same time, combined with the set value of the chilled water outlet temperature of the chiller, through the cross-modal interactive analysis of the chilled water outlet set temperature and the operating status of the room, to mine the operating status response characteristics of the refrigeration room at the current set temperature, and on this basis, make intelligent recommendations for the chilled water outlet temperature. In this way, the dynamic optimization control of the water-cooled central air-conditioning system can be effectively realized to adapt to changes in the external environment and internal load, thereby significantly improving the operating efficiency and energy efficiency of the water-cooled central air-conditioning system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0022] Figure 1 This is a flow chart of an AI-based water-cooled central air-conditioning energy-saving control method according to an embodiment of the present application.

[0023] Figure 2Schematic diagram of data flow for the AI-based energy-saving control method of water-cooled central air conditioners according to an embodiment of the present application.

[0024] Figure 3 Flowchart of sub-step S4 of the AI-based energy-saving control method of water-cooled central air conditioners according to an embodiment of the present application.

[0025] Figure 4 Flowchart of sub-step S41 of the AI-based energy-saving control method of water-cooled central air conditioners according to an embodiment of the present application.

[0026] Figure 5 Flowchart of sub-step S42 of the AI-based energy-saving control method of water-cooled central air conditioners according to an embodiment of the present application.

[0027] Figure 6 Block diagram of the AI-based energy-saving control system of water-cooled central air conditioners according to an embodiment of the present application. Detailed implementation manners

[0028] As shown in the present application and the claims, unless the context clearly indicates an exception, words such as "a", "an", "one", and / or "the" are not specifically singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of the clearly identified steps and elements, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0029] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or the server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0030] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the operations before or below do not necessarily need to be executed precisely in order. On the contrary, as needed, various steps can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several operations can be removed from these processes.

[0031] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0032] In view of the technical problems described in the above background art, the present application proposes an optimized AI-based energy-saving control method for water-cooled central air conditioners. It uses artificial intelligence technology based on deep learning to comprehensively monitor and perform time-series analysis on the operation data of the refrigeration machine room of the water-cooled central air conditioner, so as to discover the time-series correlation change patterns of multi-source operation parameters, thereby obtaining a comprehensive characterization of the operation state of the machine room. At the same time, combined with the set value of the chilled water outlet temperature of the chiller, through cross-modal interaction analysis of the set chilled water outlet temperature and the operation state of the machine room, the operation state response characteristics of the refrigeration machine room at the current set temperature are discovered, and on this basis, an intelligent recommendation of the chilled water outlet temperature is made. In this way, the dynamic optimization control of the water-cooled central air conditioning system can be effectively realized to adapt to the changes in the external environment and internal load, thereby significantly improving the operation efficiency and energy efficiency of the water-cooled central air conditioning system.

[0033] Figure 1 It is a flowchart of the AI-based energy-saving control method for water-cooled central air conditioners according to an embodiment of the present application. Figure 2 It is a schematic diagram of data flow of the AI-based energy-saving control method for water-cooled central air conditioners according to an embodiment of the present application. As Figure 1 and Figure 2 shown, the AI-based energy-saving control method for water-cooled central air conditioners includes the steps: S1, collecting the operation data of the refrigeration machine room of the water-cooled central air conditioner, and at the same time obtaining the set value of the chilled water outlet temperature of the chiller; S2, extracting the time-series mode features of the operation data of the refrigeration machine room of the water-cooled central air conditioner to obtain a time-series mode semantic coding feature map of the operation state of the machine room; S3, performing one-hot encoding on the set value of the chilled water outlet temperature of the chiller to obtain a one-hot encoded vector of the set chilled water outlet temperature; S4, performing cross-modal interaction optimization based on hint information on the one-hot encoded vector of the set chilled water outlet temperature and the time-series mode semantic coding feature map of the operation state of the machine room to obtain a joint coding feature map of the operation state - set chilled water outlet temperature response; S5, determining the recommended value of the chilled water outlet temperature based on the joint coding feature map of the operation state - set chilled water outlet temperature response; S6, updating the set value of the chilled water outlet temperature of the chiller to the recommended value of the chilled water outlet temperature.

[0034] In the above AI-based energy-saving control method for water-cooled central air conditioners, in step S1, the operation data of the refrigeration machine room of the water-cooled central air conditioner is collected, and at the same time, the set value of the chilled water outlet temperature of the chiller is obtained. It should be understood that the chilled water outlet temperature of the chiller directly affects the indoor air temperature and the load of the compressor, and thus has a significant impact on the operation data of the refrigeration machine room of the water-cooled central air conditioner. Therefore, this application expects to realize the energy-saving control of the water-cooled central air conditioning system by analyzing the correlation response mode between the set value of the chilled water outlet temperature and the operation data of the machine room. In a specific example of this application, the operation data includes: the time queue of the outdoor ambient temperature; the time queues of the supply and return water temperatures, supply and return water pressures, and supply and return water flows of the chilled water main pipe; the time series of the switch states and the time queues of the powers of the chiller, chilled water pump, cooling water pump, and cooling tower; the time queues of the inlet and outlet water temperatures, evaporation temperature, condensation temperature, evaporation pressure, and condensation pressure of the chiller; the time queues of the frequencies and inlet and outlet water pressures of the cooling water pump and the cooling water pump. In the technical solution of this application, by real-time monitoring of the above operation parameters, the cooling load demand of the air conditioning system, the response ability of the air conditioning system to the cooling load, the working mode, operation efficiency, refrigeration effect, and energy consumption level of the air conditioning system can be comprehensively understood, thereby providing a strong basis for formulating a more efficient control strategy.

[0035] In the process of collecting the operation data of the refrigeration machine room of the water-cooled central air conditioner and obtaining the set value of the chilled water outlet temperature of the chiller at the same time, an intelligent monitoring platform based on the Internet of Things (IoT) can be used to achieve this. The design and implementation of such a platform need to be comprehensively considered from multiple aspects such as system architecture design, data collection process, user interface development, and security considerations.

[0036] First, in terms of system architecture design, the platform mainly includes five parts: the sensor layer, the data collection layer, the network transmission layer, the cloud service layer, and the user interface layer. The sensor layer is responsible for being installed at various key positions, such as the chiller, cooling tower, etc., for directly measuring relevant physical quantities, including but not limited to the chilled water outlet temperature, the temperature difference between the inlet and outlet water, the water flow velocity, etc. The data collection layer uses a PLC or a dedicated data collection module to collect the raw data from the sensor layer and perform preliminary processing. Next, these processed data are uploaded to the cloud server through wireless communication technologies (such as Wi-Fi, LoRaWAN, etc.) or wired connection methods. The cloud service layer deploys application programs in the cloud, which are responsible for receiving data streams from different locations and performing further data cleaning, storage, and analysis tasks. Finally, the user interface layer provides a Web-based or mobile application for the operator to use, which can display functions such as real-time status information and historical record query.

[0037] In the specific data acquisition process, first select the appropriate sensor type and its accuracy level according to the requirements of the actual application scenario. For example, for the chilled water outlet temperature, it is recommended to use a PT100 platinum resistance thermometer with an accuracy of ±0.1°C. After selecting the hardware, the next step is to program the PLC. This step involves defining input and output points, setting the sampling period, writing data processing logic, etc. This process is implemented through ladder diagram language or other high-level languages (such as Structured Text). When it is necessary to read the chilled water outlet temperature, corresponding code snippets can be added to the program to complete the data reading and packaging, and then send it out through the serial port. To ensure the safe and reliable transmission of data, it is recommended to use a communication method that supports an encrypted transmission protocol (such as HTTPS) and add a verification mechanism (such as CRC verification). After reaching the cloud, these data will be automatically parsed and stored in the database for subsequent query and analysis.

[0038] In order to enable managers to view the current status and historical trends conveniently and quickly, it is also necessary to develop a human-machine interaction interface. This usually includes a real-time monitoring panel that displays the instantaneous values of key parameters; an alarm notification function that can send out alarms in a timely manner when abnormal situations are detected and inform relevant personnel by means of text messages or emails; a report generation tool that allows users to customize time periods to generate detailed reports including statistical charts; in addition, some simple remote control options can be added, such as manually adjusting the set temperature, etc.

[0039] In the design process of the entire system, security is always an issue that cannot be ignored. In addition to the data encryption measures mentioned above, the following strategies should also be adopted: implement strict access control for all devices accessing the network to ensure that only authorized users can access sensitive information; regularly update the software version to repair known vulnerabilities; configure firewall rules to block illegal intrusion attempts; implement a multi-factor authentication mechanism to increase the risk of account theft.

[0040] In summary, through reasonable planning and comprehensive application of existing technical means, it is completely possible to effectively collect and manage the operation data of the chilled water central air-conditioning refrigeration machine room, improve the daily operation and maintenance efficiency of the machine room, and thus promote the optimized operation of the entire air-conditioning system.

[0041] In the above-mentioned AI-based water-cooled central air-conditioning energy-saving control method, the step S2 extracts the time series pattern characteristics of the operating data of the water-cooled central air-conditioning refrigeration room to obtain the time series pattern semantic coding feature map of the room operation state. In a specific example of the present application, the step S2 includes: after the operating data of the water-cooled central air-conditioning refrigeration room is arranged according to the time dimension and the parameter sample dimension into the water-cooled central air-conditioning refrigeration room operation state time series aggregation matrix, the water-cooled central air-conditioning refrigeration room operation state time series aggregation matrix is input into the room operation state time series pattern encoder based on the RNN-LSTM hybrid model to obtain the room operation state time series pattern semantic coding feature map. It should be understood that considering that the operating data of the water-cooled central air-conditioning refrigeration room has a high degree of time series, and there are certain interactions and correlations between the various operating parameters in the operating data, for example, the outdoor ambient temperature directly affects the heat dissipation effect of the cooling tower, and then affects the condensing temperature of the chiller; the temperature difference between the supply and return water of the chilled water reflects the actual cooling capacity of the air-conditioning system, and a larger temperature difference requires more supply and return water flow to meet the cooling load demand. Therefore, in order to capture the associated response patterns and time series change characteristics between various operating parameters, the present application further arranges the operating data of the water-cooled central air-conditioning refrigeration room according to the time dimension and parameter sample dimension into a water-cooled central air-conditioning refrigeration room operating state time series aggregation matrix, so as to maintain the time order and associated structure of the data. Then, the RNN-LSTM hybrid model is used to construct a machine room operating state time series pattern encoder to process the water-cooled central air-conditioning refrigeration room operating state time series aggregation matrix, so as to comprehensively utilize the sequence processing advantages of the RNN model and the long-term memory ability of the LSTM model, dig out the time series association change pattern between multi-source operating parameters, capture the time series dynamic change characteristics of the machine room operating state, and generate a machine room operating state time series pattern semantic coding feature map, thereby providing a comprehensive machine room operating state representation for subsequent energy-saving control.

[0042] In the above AI-based energy-saving control method for water-cooled central air conditioners, in step S3, the set value of the chilled water outlet temperature of the chiller is one-hot encoded to obtain a one-hot encoded vector of the chilled water outlet temperature set value. It should be understood that when using a machine learning model to encode the set value of the chilled water outlet temperature, if the temperature set value in numerical form is directly used, the model may wrongly consider the numerical difference as representing some importance or order relationship, while in fact, the set value of the chilled water outlet temperature is only a reference point. Therefore, in the technical solution of this application, in order to eliminate the influence brought by the numerical magnitude, the set value of the chilled water outlet temperature of the chiller is regarded as a categorical variable, and the one-hot encoding technology is used to process it, so as to convert the temperature set value into a binary vector form, and obtain a one-hot encoded vector of the chilled water outlet temperature set value. Among them, each position in the vector corresponds to a specific temperature, so that each temperature set value is represented by a unique binary vector, thereby ensuring that the model regards the temperature set value as a discrete category rather than a continuous numerical value during processing, and avoiding the influence of numerical magnitude on model training.

[0043] In the above AI-based energy-saving control method for water-cooled central air conditioners, in step S4, cross-modal interaction optimization based on hint information is performed on the one-hot encoded vector of the chilled water outlet temperature set value and the semantic encoding feature map of the timing pattern of the computer room operation state to obtain a joint encoding feature map of the computer room operation state - chilled water outlet temperature set response. That is, this application further reveals the response characteristics of the refrigeration computer room operation state under the current chilled water outlet temperature setting by performing interaction analysis on the one-hot encoded vector of the chilled water outlet temperature set value and the semantic encoding feature map of the timing pattern of the computer room operation state, so as to provide a decision-making basis for the intelligent recommendation of the chilled water outlet temperature. In particular, in order to achieve efficient interaction analysis between the one-hot encoded vector of the chilled water outlet temperature set value and the semantic encoding feature map of the timing pattern of the computer room operation state, and to improve the interaction and fusion effect of multi-modal data, this application proposes a cross-modal interaction optimization method based on hint information, which mines the local fine-grained interaction features between the one-hot encoded vector of the chilled water outlet temperature set value and the semantic encoding feature map of the timing pattern of the computer room operation state as hint information, and guides and optimizes the cross-modal interaction features between the two, so as to enhance the sensitivity and recognition ability of the model to the changes in the computer room operation state under the chilled water outlet temperature setting, and further improve the accuracy of subsequent chilled water outlet temperature control. Among them, Figure 3 is a flowchart of sub-step S4 of the AI-based energy-saving control method for water-cooled central air conditioners according to an embodiment of the present application. As Figure 3As shown, step S4 includes steps: S41, extracting local fine-grained interaction features between the set temperature one-hot encoding vector of the chilled water outlet temperature and the semantic encoding feature map of the timing pattern of the machine room operation status as prompt information to obtain a set of semantic encoding matrices of the cross-modal prompt information of the machine room operation status - set temperature of the chilled water outlet; S42, based on the set of semantic encoding matrices of the cross-modal prompt information of the machine room operation status - set temperature of the chilled water outlet, guiding the cross-modal interaction optimization of the set temperature one-hot encoding vector of the chilled water outlet and the semantic encoding feature map of the timing pattern of the machine room operation status to obtain the joint encoding feature map of the response of the machine room operation status - set temperature of the chilled water outlet.

[0044] Figure 4 It is a flowchart of sub-step S41 of the AI-based energy-saving control method for water-cooled central air conditioners according to an embodiment of the present application. As Figure 4 shown, step S41 includes steps: S411, calculating the product between the set temperature one-hot encoding vector of the chilled water outlet temperature and the transposed vector of the set temperature one-hot encoding vector of the chilled water outlet temperature to obtain the self-correlation encoding matrix of the set temperature of the chilled water outlet; S412, performing feature decoupling on the semantic encoding feature map of the timing pattern of the machine room operation status to obtain a set of local feature matrices of the timing pattern of the machine room operation status; S413, respectively performing cross-modal correlation interaction based on the attention mechanism between the self-correlation encoding matrix of the set temperature of the chilled water outlet and each local feature matrix of the timing pattern of the machine room operation status in the set of local feature matrices of the timing pattern of the machine room operation status to obtain a set of semantic encoding matrices of the cross-modal prompt information of the machine room operation status - set temperature of the chilled water outlet.

[0045] More specifically, step S411 is expressed by the formula:

[0046]

[0047] where, V1 represents the set temperature one-hot encoding vector of the chilled water outlet temperature, represents matrix multiplication operation, (·) T represents the transpose of a vector, M z represents the self-correlation encoding matrix of the set temperature of the chilled water outlet.

[0048] That is, by calculating the product between the set temperature one-hot encoding vector of the chilled water outlet temperature and its own transposed vector, the self-correlation of the set temperature of the chilled water outlet is revealed by using the concept of outer product in matrix operations, and the self-correlation encoding matrix of the set temperature of the chilled water outlet is obtained.

[0049] More specifically, step S412 is expressed by the formula:

[0050] Decouple(X) = {M1, M2,..., M n}

[0051] Wherein, X represents the semantic encoding feature map of the time series pattern of the operation state of the computer room, Decouple(·) represents feature decoupling, and M1, M2, M i and M n respectively represent the first, second, i-th, and n-th local feature matrices of the time series pattern of the operation state of the computer room in the set of local feature matrices of the time series pattern of the operation state of the computer room.

[0052] That is, by performing feature decoupling on the semantic encoding feature map of the time series pattern of the operation state of the computer room, it is decomposed into multiple local feature matrices of the time series pattern of the operation state of the computer room, so as to improve the model's attention to the local details of the operation state of the computer room, reduce the information redundancy between features, enhance the distinguishability of features, and thus contribute to performing more fine-grained cross-modal interaction analysis.

[0053] More specifically, step S413 further includes: performing a linear transformation on the self-correlation encoding matrix of the chilled water outlet temperature setting to obtain a query encoding matrix and a value encoding matrix for the chilled water outlet temperature setting, which is expressed by the formula:

[0054]

[0055] Wherein, W q and W v respectively represent the query embedding matrix and the value embedding matrix, M q and M v respectively represent the query encoding matrix and the value encoding matrix for the chilled water outlet temperature setting, represents matrix multiplication.

[0056] That is, by means of linear mapping, the self-correlation encoding matrix of the chilled water outlet temperature setting is converted into a query encoding matrix and a value encoding matrix for the chilled water outlet temperature setting to meet the input form requirements of the attention mechanism. Among them, the query encoding matrix for the chilled water outlet temperature setting is used to query the correlation with the features of the operation state of the computer room, while the value encoding matrix for the chilled water outlet temperature setting stores the original information of the chilled water outlet temperature setting to be fused.

[0057] More specifically, the step S413 further includes: respectively inputting each local feature matrix of the timing pattern of the computer room operation state in the set of the freezing water outlet temperature setting query coding matrix, the freezing water outlet temperature setting value coding matrix, and the local feature matrix of the timing pattern of the computer room operation state into the cross-modal prompt information encoder based on the transformer structure to obtain the set of cross-modal prompt information semantic coding matrices of the computer room operation state - freezing water outlet temperature setting, which is expressed by the formula:

[0058]

[0059] where S is the characteristic scale value of the local feature matrix of the timing pattern of the computer room operation state, softmax(·) represents the normalized exponential function, and I i represents the i-th cross-modal prompt information semantic coding matrix of the computer room operation state - freezing water outlet temperature setting.

[0060] That is, based on the transformer structure, cross-modal encoding is performed on the freezing water outlet temperature setting query coding matrix, the freezing water outlet temperature setting value coding matrix, and each local feature matrix of the timing pattern of the computer room operation state. Through the self-attention mechanism of the transformer structure, feature fusion and interactive analysis of cross-modal data are realized, effectively capturing the deep fine-grained semantic connection between the freezing water outlet temperature setting and the characteristics of the computer room operation state, generating a set of cross-modal prompt information semantic coding matrices of the computer room operation state - freezing water outlet temperature setting, and using this as prompt information to guide subsequent feature interaction optimization.

[0061] Figure 5 FIG. is a flowchart of sub-step S42 of the AI-based energy-saving control method for water-cooled central air conditioners according to an embodiment of the present application. As Figure 5 shown, the step S42 includes steps: S421, inputting the set of cross-modal prompt information semantic coding matrices of the computer room operation state - freezing water outlet temperature setting into the information gating unit based on the decoder to obtain a set of cross-modal semantic interaction attention weights of the computer room operation state - freezing water outlet temperature setting; S422, inputting the self-correlation coding matrix of the freezing water outlet temperature setting and the set of local feature matrices of the timing pattern of the computer room operation state into the cross-modal interaction unit to obtain a set of cross-modal interaction local feature matrices of the computer room operation state - freezing water outlet temperature setting; S423, inputting the set of cross-modal semantic interaction attention weights of the computer room operation state - freezing water outlet temperature setting and the set of cross-modal interaction local feature matrices of the computer room operation state - freezing water outlet temperature setting into the cross-modal interaction optimization unit to obtain the joint coding feature map of the computer room operation state - freezing water outlet temperature setting response.

[0062] More specifically, the step S421 is expressed by the formula:

[0063] a i = softmax{decoder(I i , W a )}

[0064] where decoder(·) represents the decoder, and W a represents the decoding weight matrix, and a i represents the cross-modal semantic interaction attention weight of the operating state - chilled water supply temperature setting of the i-th computer room.

[0065] That is, the set of cross-modal prompt information semantic encoding matrices of the operating state - chilled water supply temperature setting of the computer room is input into the decoder-based information gating unit for information screening. The information gating unit learns the feature importance of each self-correlation encoding matrix of the chilled water supply temperature setting and the local feature matrix of the temporal pattern of the operating state of the computer room in the set through the decoder, and accordingly performs weight allocation to generate a set of cross-modal semantic interaction attention weights of the operating state - chilled water supply temperature setting of the computer room, so as to effectively highlight the important correlation response pattern between the operating state of the computer room and the chilled water supply temperature during the subsequent feature fusion process, while suppressing the influence of irrelevant or noisy features.

[0066] More specifically, steps S422 and S423 are expressed by the formula:

[0067] F = couple{a1·M1⊙M z , a2·M2⊙M z ,..., a n ·M n ⊙M z}

[0068] where a1, a2, and a n represent the first, second, and n-th cross-modal semantic interaction attention weights of the operating state - chilled water supply temperature setting of the computer room, ⊙ represents element-wise multiplication, couple{·} represents feature concatenation, and F represents the response joint encoding feature map of the operating state - chilled water supply temperature setting of the computer room.

[0069] That is, through element-wise multiplication, direct interaction is carried out between the set of self-correlation encoding matrices of the chilled water supply temperature setting and the local feature matrix of the temporal pattern of the operating state of the computer room, so as to capture the correlation information at the same feature space position of the two, and the interaction features between the chilled water supply temperature and the operating state of the computer room are weighted and aggregated with the attention weights generated in the above process, thereby generating a more representative cross-modal semantic interaction feature map of the operating state - chilled water supply temperature setting of the computer room, providing more accurate guidance for the subsequent temperature control of the air conditioning system.

[0070] In the above AI-based energy-saving control method for water-cooled central air conditioners, in step S5, based on the joint encoding feature map of the computer room operation state - chilled water outlet temperature setting response, a recommended value of the chilled water outlet temperature is determined. In a specific example of the present application, step S5 includes: inputting the joint encoding feature map of the computer room operation state - chilled water outlet temperature setting response into a chilled water outlet temperature dynamic optimization module based on a decoder to obtain a decoded value of the recommended value of the chilled water outlet temperature. In the present application, the decoder decodes the joint encoding feature map of the computer room operation state - chilled water outlet temperature setting response through a multi-layer neural network structure, and learns the non-linear correlation relationship between the chilled water outlet temperature and the computer room operation state through layer-by-layer feature transfer and non-linear transformation, so as to intelligently recommend an appropriate chilled water outlet temperature value according to the current computer room operation state and chilled water outlet temperature setting, so as to meet the real-time operation requirements and energy-saving goals of the computer room.

[0071] Specifically, the decoder is a neural network model that extracts and transforms the features of the joint encoding feature map of the computer room operation state - chilled water outlet temperature setting response through a series of neural network layers. Each layer contains a set of trainable weight and bias parameters, as well as a non-linear activation function, to simulate the complex relationships between data. As information is passed through each layer, the decoder can extract the key operation modes and rules of the air conditioning system. After multiple iterative trainings, the decoder can learn to accurately identify the correlation between the chilled water outlet temperature and the computer room operation state from the given input features, and adjust its output accordingly to generate a decoded value of the recommended value of the corresponding chilled water outlet temperature. It should be understood that the recommended value of the chilled water outlet temperature reflects the chilled water outlet temperature setting that is most conducive to maintaining efficient refrigeration while ensuring energy utilization efficiency under the current computer room operation state. By applying it to the control system of the water-cooled central air conditioner, it helps to achieve the dual goals of energy saving and performance improvement of the air conditioning system.

[0072] In a preferred example of the present application, inputting the joint encoding feature map of the computer room operation state - chilled water outlet temperature setting response into a chilled water outlet temperature dynamic optimization module based on a decoder to obtain a decoded value of the recommended value of the chilled water outlet temperature includes:

[0073] Performing feature clustering on the feature set of the joint encoding feature map of the computer room operation state - chilled water outlet temperature setting response to obtain a within-class feature set of the joint encoding of the computer room operation state - chilled water outlet temperature setting response and a between-class feature set of the joint encoding of the computer room operation state - chilled water outlet temperature setting response, that is:

[0074]

[0075] Wherein, is the set of intra-class feature values of the joint encoding of the operating state of the computer room - chilled water outlet temperature setting response, f 1i is the feature value at the i-th position in the set of intra-class feature values of the joint encoding of the operating state of the computer room - chilled water outlet temperature setting response, f 2j is the feature value at the j-th position in the set of extra-class feature values of the joint encoding of the operating state of the computer room - chilled water outlet temperature setting response.

[0076] Calculate the ratio of the number of feature values in the set of intra-class feature values of the joint encoding of the operating state of the computer room - chilled water outlet temperature setting response to the number of feature values in the set of feature values of the joint encoding feature map of the operating state of the computer room - chilled water outlet temperature setting response, that is:

[0077]

[0078] where k is the number of feature values in the set of intra-class feature values of the joint encoding of the operating state of the computer room - chilled water outlet temperature setting response, m is the number of feature values in the set of feature values of the joint encoding feature map of the operating state of the computer room - chilled water outlet temperature setting response, and λ is the ratio of k to m.

[0079] Calculate the ratio of the λ-th power of the sum of the absolute values of all feature values in the set of intra-class feature values of the joint encoding of the operating state of the computer room - chilled water outlet temperature setting response to the λ-th power of the sum of the absolute values of all feature values in the set of feature values of the joint encoding feature map of the operating state of the computer room - chilled water outlet temperature setting response to obtain the joint encoding modulation weight of the operating state of the computer room - chilled water outlet temperature setting response, that is:

[0080]

[0081] where f p is the feature value at the p-th position in the set of feature values of the joint encoding feature map of the operating state of the computer room - chilled water outlet temperature setting response, and w1 is the joint encoding modulation weight of the operating state of the computer room - chilled water outlet temperature setting response.

[0082] Calculate the ratio of the (λ / 2)-th power of the sum of the squares of all feature values in the set of intra-class feature values of the joint encoding of the operating state of the computer room - chilled water outlet temperature setting response to the (λ / 2)-th power of the sum of the squares of all feature values in the set of feature values of the joint encoding feature map of the operating state of the computer room - chilled water outlet temperature setting response to obtain the joint encoding harmonic weight of the operating state of the computer room - chilled water outlet temperature setting response, that is:

[0083]

[0084] where w2 is the joint encoding harmonic weight of the operating state of the computer room - chilled water outlet temperature setting response.

[0085] For each eigenvalue in the within-class feature set of the joint encoding of the computer room operating status - chilled water outlet temperature setting response, calculate the product of it and the harmonic weight of the joint encoding of the computer room operating status - chilled water outlet temperature setting response, and then add the modulation weight of the joint encoding of the computer room operating status - chilled water outlet temperature setting response to obtain an optimized eigenvalue, that is:

[0086] f’ 1i = w2 × f 1i + w1

[0087] where f' 1i is the i-th optimized eigenvalue in the within-class feature set of the joint encoding of the computer room operating status - chilled water outlet temperature setting response.

[0088] For each eigenvalue in the out-of-class feature set of the joint encoding of the computer room operating status - chilled water outlet temperature setting response, calculate the product of it and the tuning weight of the joint encoding of the computer room operating status - chilled water outlet temperature setting response to obtain an optimized eigenvalue, that is:

[0089] f 2j ' = w1 × f 2j

[0090] where f 2j ' is the j-th optimized eigenvalue in the out-of-class feature set of the joint encoding of the computer room operating status - chilled water outlet temperature setting response.

[0091] Input the optimized joint encoding feature map of the computer room operating status - chilled water outlet temperature setting response based on the within-class feature set of the joint encoding of the computer room operating status - chilled water outlet temperature setting response and the out-of-class feature set of the joint encoding of the computer room operating status - chilled water outlet temperature setting response into the decoder-based chilled water outlet temperature dynamic optimization module to obtain the decoded value of the recommended value of the chilled water outlet temperature.

[0092] Here, when the one-hot encoding vector of the chilled water outlet temperature setting and the semantic encoding feature map of the computer room operating status time series mode represent the one-hot encoding feature of the chilled water outlet temperature setting and the semantic association feature of the computer room operating status time series multi-parameters respectively, considering the differences between their cross-modal prompt gates, it is expected to improve the semantic consistent aggregation expression effect of the joint encoding feature map of the computer room operating status - chilled water outlet temperature setting response obtained by optimizing the encoder through cross-modal interaction.

[0093] Based on this, while clustering the joint coding feature map of the computer room operation status - chilled water outlet temperature setting response in this application, for the interactive description of the key feature information of the joint coding feature map of the computer room operation status - chilled water outlet temperature setting response during the clustering process, a geometric equivariant topology of the feature is constructed through feature low-rank harmonic modulation based on the equivariance of the clustering features and the overall features of the joint coding feature map of the computer room operation status - chilled water outlet temperature setting response, so as to obtain the translational and rotational symmetry of the schematic distribution of the clustering features of the joint coding feature map of the computer room operation status - chilled water outlet temperature setting response relative to the overall features. Thus, on the basis of introducing geometric message passing in the feature expression of the joint coding feature map of the computer room operation status - chilled water outlet temperature setting response, the clustering mapping symmetry of the joint coding feature map of the computer room operation status - chilled water outlet temperature setting response is realized through the manipulation of irreducible low-rank order coefficients, enhancing the consistency of the clustering-based feature representation of the joint coding feature map of the computer room operation status - chilled water outlet temperature setting response, and thereby improving the accuracy of the decoded value of the recommended value of the chilled water outlet temperature obtained by inputting the joint coding feature map of the computer room operation status - chilled water outlet temperature setting response into the chilled water outlet temperature dynamic optimization module based on the decoder.

[0094] In the above AI-based energy-saving control method for water-cooled central air conditioners, in step S6, the set value of the chilled water outlet temperature of the chiller is updated to the recommended value of the chilled water outlet temperature. That is, by taking the recommended value of the chilled water outlet temperature as the new set value and inputting it into the chiller control system, the real-time dynamic adjustment of the chilled water outlet temperature can be realized. In this way, not only can it ensure that the chiller achieves the optimal energy efficiency ratio while meeting the cooling requirements of the computer room, realizing the goal of energy conservation and consumption reduction, but also it can automatically adjust the set value of the chilled water outlet temperature according to the changes in the computer room operation status, ensuring the stability of the computer room environment and the long-term operation efficiency of the equipment, thereby effectively reducing energy waste, lowering the operation cost, and having a positive impact on the environment.

[0095] The specific implementation manner of this process will be introduced in detail below:

[0096] First, it is necessary to select a suitable communication interface for data exchange with the chiller control system in the machine room. Common communication protocols include Modbus TCP / IP, OPC UA, BACnet, LonWorks, etc. These protocols have their own characteristics and are suitable for different application scenarios. For example, Modbus TCP / IP is widely used in the industrial automation field due to its simplicity, ease of use, and strong openness; OPC UA is favored for its powerful interoperability and security; BACnet is designed specifically for building automation and is suitable for the HVAC systems of large buildings; LonWorks is known for its distributed network architecture and flexible node configuration.

[0097] After selecting the communication protocol, it is necessary to configure the chiller control system accordingly to ensure that it can receive instructions from external systems. This usually includes setting the communication port, configuring communication parameters (such as baud rate, data bits, stop bits, etc.), setting the network address, etc. In addition, it is also necessary to ensure that the chiller control system supports the required command set so that it can receive and execute the new chilled water outlet temperature setpoint.

[0098] After the communication interface configuration is completed, it is necessary to define the data format and protocol parsing rules to ensure that the recommended values can be correctly transmitted to the chiller control system. This usually involves the following aspects:

[0099] Data format definition: It is necessary to define the format of the transmitted data, including data type, data length, data unit, etc. For example, the chilled water outlet temperature setpoint may be transmitted in the form of a floating-point number, with the unit being Celsius or Fahrenheit. The data format definition needs to match the input requirements of the chiller control system to ensure that the data can be correctly parsed and applied.

[0100] Protocol parsing rules: It is necessary to write protocol parsing code to convert the generated recommended values into a packet format that conforms to the communication protocol. This usually involves data packing and unpacking operations to ensure that the data is not lost or damaged during transmission. For example, when using the Modbus TCP / IP protocol, it is necessary to encapsulate the recommended values into a Modbus message format, including fields such as function code, register address, data length, and data value.

[0101] Once the data format and protocol parsing rules are defined, the recommended values can be sent to the chiller control system through the communication interface. This process generally consists of the following steps: Establishing a connection: Establish a connection with the chiller control system through a network or serial communication interface. If it is a network connection, a Socket connection needs to be established using the TCP / IP protocol stack; if it is a serial communication interface, the serial port needs to be opened and communication parameters set. Data transmission: After the connection is established, the system sends the generated data packet through the communication interface to the chiller control system. The data packet contains the recommended chilled water outlet temperature setting value and other necessary control information. During the transmission process, issues such as network latency and packet loss that may occur need to be handled to ensure the reliable transmission of data. Command execution: After receiving the data packet, the chiller control system performs protocol parsing and extracts the recommended chilled water outlet temperature setting value. The control system adjusts the operating state of the chiller according to the new setting value, such as adjusting the operating frequency of the compressor and changing the rotational speed of the chilled water pump, etc., to achieve the purpose of adjusting the chilled water outlet temperature.

[0102] To ensure the correct application of the recommended values, the system also needs to establish a feedback mechanism to monitor the response of the chiller. This generally includes the following aspects: Status feedback: After receiving the new setting value, the chiller control system will return a confirmation message to inform the system that the new setting value has been successfully applied. The system can receive these feedback messages through the communication interface to verify whether the recommended values are correctly executed. Exception handling: If the system detects that the chiller fails to respond correctly to the new setting value or other abnormal situations occur, immediate measures need to be taken to handle them. For example, the system can resend the recommended values or revert to the previous setting value to ensure the stable operation of the system. In addition, the system can record the abnormal situations and generate log files for subsequent analysis and troubleshooting.

[0103] After applying the recommended value of the chilled water outlet temperature, the system continues to monitor the operating states of the chiller and the entire refrigeration system and collect new operating data. These data are used to evaluate whether the new setting value has achieved the expected effects, such as whether it has effectively improved energy efficiency and whether it has maintained a stable indoor temperature, etc. Based on this feedback information, the system can further optimize the setting value of the chilled water outlet temperature, forming a continuous improvement process.

[0104] Although the system can automatically adjust the setting value of the chilled water outlet temperature, user intervention and confirmation are sometimes still required. The system can provide a user interface that allows operators to view the current setting value, operating state, and performance indicators and make manual adjustments. User intervention can serve as a supplement to the system's adaptive adjustment to ensure the more flexible and reliable operation of the system.

[0105] During the entire update process, the security and reliability of the system are of utmost importance. To ensure the stable operation of the system, a variety of security measures need to be taken, including but not limited to: Data Encryption: During data transmission, use encryption protocols such as SSL / TLS to protect the security of data and prevent data from being stolen or tampered with. Access Control: Implement strict access control for all devices accessing the network to ensure that only authorized users can access sensitive information. The system can authenticate users through methods such as usernames and passwords, digital certificates, etc. Redundant Backup: The system can be configured with redundant communication paths and data backup mechanisms to ensure that data transmission and system operation are not affected in the event of network failures or device failures. Fault Recovery: The system needs to have an automatic fault recovery function that can automatically switch to a backup system or restore to the previous set value when abnormal situations are detected to ensure the continuous operation of the system.

[0106] Furthermore, an AI-based energy-saving control system for water-cooled central air conditioners is also provided.

[0107] Figure 6 It is a block diagram of an AI-based energy-saving control system for water-cooled central air conditioners according to an embodiment of the present application. As Figure 6 shown, the AI-based energy-saving control system 100 for water-cooled central air conditioners according to an embodiment of the present application includes: a data acquisition module 110 for acquiring the operation data of the refrigeration machine room of the water-cooled central air conditioner and simultaneously obtaining the set value of the chilled water outlet temperature of the chiller; a feature extraction module 120 for extracting the temporal pattern features of the operation data of the refrigeration machine room of the water-cooled central air conditioner to obtain a semantic encoding feature map of the temporal pattern of the machine room operation state; a one-hot encoding module 130 for performing one-hot encoding on the set value of the chilled water outlet temperature of the chiller to obtain a one-hot encoding vector of the chilled water outlet temperature setting; a cross-modal interaction optimization module 140 for performing cross-modal interaction optimization based on prompt information on the one-hot encoding vector of the chilled water outlet temperature setting and the semantic encoding feature map of the temporal pattern of the machine room operation state to obtain a joint encoding feature map of the machine room operation state - chilled water outlet temperature setting response; a temperature value decision module 150 for determining a recommended value of the chilled water outlet temperature based on the joint encoding feature map of the machine room operation state - chilled water outlet temperature setting response; and a temperature value update module 160 for updating the set value of the chilled water outlet temperature of the chiller to the recommended value of the chilled water outlet temperature.

[0108] Here, those skilled in the art can understand that the specific operations of each module in the above AI-based energy-saving control system for water-cooled central air conditioners have been described in detail in the description of the AI-based energy-saving control method for water-cooled central air conditioners above, and therefore, the repeated description thereof will be omitted. Figures 1 to 5 of the above

[0109] The basic principles of the present invention have been described in conjunction with specific embodiments. However, it should be noted that the advantages, benefits, effects, etc. mentioned in the present invention are merely examples and not limitations, and it cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of the present invention. Additionally, the specific details of the above embodiments are only for the purpose of illustration and facilitating understanding, rather than limitations. The above details do not limit the present invention to necessarily adopt the above specific details for implementation.

Claims

1. An AI-based energy-saving control method for water-cooled central air conditioners, characterized in that, Including: Collecting the operation data of the water-cooled central air-conditioning refrigeration machine room, and simultaneously obtaining the set value of the chilled water outlet temperature of the chiller; Extracting the temporal pattern features of the operation data of the water-cooled central air-conditioning refrigeration machine room to obtain the semantic coding feature map of the temporal pattern of the machine room operation state; Performing one-hot encoding on the set value of the chilled water outlet temperature of the chiller to obtain the one-hot encoding vector of the chilled water outlet temperature setting; Performing cross-modal interaction optimization based on the hint information on the one-hot encoding vector of the chilled water outlet temperature setting and the semantic coding feature map of the temporal pattern of the machine room operation state to obtain the joint coding feature map of the machine room operation state - chilled water outlet temperature setting response; Based on the joint coding feature map of the machine room operation state - chilled water outlet temperature setting response, determining the recommended value of the chilled water outlet temperature; Updating the set value of the chilled water outlet temperature of the chiller to the recommended value of the chilled water outlet temperature; Performing cross-modal interaction optimization based on the hint information on the one-hot encoding vector of the chilled water outlet temperature setting and the semantic coding feature map of the temporal pattern of the machine room operation state to obtain the joint coding feature map of the machine room operation state - chilled water outlet temperature setting response, including: extracting the local fine-grained interaction features between the one-hot encoding vector of the chilled water outlet temperature setting and the semantic coding feature map of the temporal pattern of the machine room operation state as hint information to obtain a set of semantic coding matrices of the cross-modal hint information of the machine room operation state - chilled water outlet temperature setting; based on the set of semantic coding matrices of the cross-modal hint information of the machine room operation state - chilled water outlet temperature setting, guiding the one-hot encoding vector of the chilled water outlet temperature setting and the semantic coding feature map of the temporal pattern of the machine room operation state to perform cross-modal interaction optimization to obtain the joint coding feature map of the machine room operation state - chilled water outlet temperature setting response.

2. The AI-based energy-saving control method for water-cooled central air conditioners according to claim 1, wherein The operation data includes: the time queue of the outdoor ambient temperature; the time queue of the supply and return water temperatures, supply and return water pressures, and supply and return water flows of the chilled water main pipe; the time series of the on / off states and the time queue of the power of the chiller, chilled water pump, cooling water pump, and cooling water tower; the time queue of the inlet and outlet water temperatures, evaporation temperature, condensation temperature, evaporation pressure, and condensation pressure of the chiller; the time queue of the frequencies and inlet and outlet water pressures of the cooling water pump and the cooling water pump.

3. The AI-based energy-saving control method for water-cooled central air conditioners according to claim 2, wherein, Extracting the temporal pattern features of the operation data of the water-cooled central air-conditioning refrigeration machine room to obtain the semantic coding feature map of the temporal pattern of the machine room operation state, including: After arranging the operation data of the water-cooled central air-conditioning refrigeration machine room into a temporal aggregation matrix of the operation state of the water-cooled central air-conditioning refrigeration machine room according to the time dimension and the parameter sample dimension, inputting the temporal aggregation matrix of the operation state of the water-cooled central air-conditioning refrigeration machine room into the temporal pattern encoder of the machine room operation state based on the RNN-LSTM hybrid model to obtain the semantic coding feature map of the temporal pattern of the machine room operation state.

4. The AI-based energy-saving control method for water-cooled central air conditioners according to claim 3, wherein, Extracting the local fine-grained interaction features between the one-hot encoding vector of the chilled water outlet temperature setting and the semantic coding feature map of the temporal pattern of the machine room operation state as hint information to obtain a set of semantic coding matrices of the cross-modal hint information of the machine room operation state - chilled water outlet temperature setting, including: Calculate the product between the one-hot encoded vector of the chilled water outlet temperature setting and the transposed vector of the one-hot encoded vector of the chilled water outlet temperature setting to obtain the self-correlation encoding matrix of the chilled water outlet temperature setting; Perform feature decoupling on the semantic encoding feature map of the timing pattern of the computer room operation status to obtain a set of local feature matrices of the timing pattern of the computer room operation status; Perform cross-modal correlation interaction based on the attention mechanism between the self-correlation encoding matrix of the chilled water outlet temperature setting and each local feature matrix of the timing pattern of the computer room operation status in the set of local feature matrices of the timing pattern of the computer room operation status to obtain a set of semantic encoding matrices of the cross-modal prompt information of the computer room operation status - chilled water outlet temperature setting; 5. The AI-based energy-saving control method for water-cooled central air conditioners according to claim 4, wherein Performing cross-modal correlation interaction based on the attention mechanism between the self-correlation encoding matrix of the chilled water outlet temperature setting and each local feature matrix of the timing pattern of the computer room operation status in the set of local feature matrices of the timing pattern of the computer room operation status to obtain a set of semantic encoding matrices of the cross-modal prompt information of the computer room operation status - chilled water outlet temperature setting, including: Perform a linear transformation on the self-correlation encoding matrix of the chilled water outlet temperature setting to obtain a query encoding matrix of the chilled water outlet temperature setting and a value encoding matrix of the chilled water outlet temperature setting; Input the query encoding matrix of the chilled water outlet temperature setting, the value encoding matrix of the chilled water outlet temperature setting, and each local feature matrix of the timing pattern of the computer room operation status in the set of local feature matrices of the timing pattern of the computer room operation status into a cross-modal prompt information encoder based on the Transformer structure to obtain a set of semantic encoding matrices of the cross-modal prompt information of the computer room operation status - chilled water outlet temperature setting; 6. The AI-based energy-saving control method for water-cooled central air conditioners according to claim 5, characterized in that, Based on the set of semantic encoding matrices of the cross-modal prompt information of the computer room operation status - chilled water outlet temperature setting, guide the one-hot encoded vector of the chilled water outlet temperature setting and the semantic encoding feature map of the timing pattern of the computer room operation status to perform cross-modal interaction optimization to obtain the joint encoding feature map of the response of the computer room operation status - chilled water outlet temperature setting, including: Input the set of semantic encoding matrices of the cross-modal prompt information of the computer room operation status - chilled water outlet temperature setting into an information gating unit based on a decoder to obtain a set of cross-modal semantic interaction attention weights of the computer room operation status - chilled water outlet temperature setting; Input the self-correlation encoding matrix of the chilled water outlet temperature setting and the set of local feature matrices of the timing pattern of the computer room operation status into a cross-modal interaction unit to obtain a set of cross-modal interaction local feature matrices of the computer room operation status - chilled water outlet temperature setting; Input the set of cross-modal semantic interaction attention weights of the computer room operation status - chilled water outlet temperature setting and the set of cross-modal interaction local feature matrices of the computer room operation status - chilled water outlet temperature setting into a cross-modal interaction optimization unit to obtain the joint encoding feature map of the response of the computer room operation status - chilled water outlet temperature setting; 7. The AI-based energy-saving control method for water-cooled central air conditioners according to claim 6, characterized in that, Based on the joint encoding feature map of the response of the computer room operation status - chilled water outlet temperature setting, determine the recommended value of the chilled water outlet temperature, including: Input the joint encoding feature map of the computer room operating status - chilled water outlet temperature setting response into the chilled water outlet temperature dynamic optimization module based on a decoder to obtain the decoded value of the recommended chilled water outlet temperature.

8. An AI-based energy-saving control system for water-cooled central air conditioners, characterized in that, It includes: A data acquisition module, which is used to collect the operating data of the water-cooled central air-conditioning refrigeration computer room and simultaneously obtain the set value of the chilled water outlet temperature of the chiller; A feature extraction module, which is used to extract the time-series pattern features of the operating data of the water-cooled central air-conditioning refrigeration computer room to obtain the time-series pattern semantic encoding feature map of the computer room operating status; A one-hot encoding module, which is used to perform one-hot encoding on the set value of the chilled water outlet temperature of the chiller to obtain the one-hot encoding vector of the chilled water outlet temperature setting; A cross-modal interaction optimization module, which is used to perform cross-modal interaction optimization based on hint information on the one-hot encoding vector of the chilled water outlet temperature setting and the time-series pattern semantic encoding feature map of the computer room operating status to obtain the joint encoding feature map of the computer room operating status - chilled water outlet temperature setting response; A temperature value decision module, which is used to determine the recommended value of the chilled water outlet temperature based on the joint encoding feature map of the computer room operating status - chilled water outlet temperature setting response; A temperature value update module, which is used to update the set value of the chilled water outlet temperature of the chiller to the recommended value of the chilled water outlet temperature; Among them, performing cross-modal interaction optimization based on hint information on the one-hot encoding vector of the chilled water outlet temperature setting and the time-series pattern semantic encoding feature map of the computer room operating status to obtain the joint encoding feature map of the computer room operating status - chilled water outlet temperature setting response includes: extracting the local fine-grained interaction features between the one-hot encoding vector of the chilled water outlet temperature setting and the time-series pattern semantic encoding feature map of the computer room operating status as hint information to obtain a set of cross-modal hint information semantic encoding matrices of the computer room operating status - chilled water outlet temperature setting; based on the set of cross-modal hint information semantic encoding matrices of the computer room operating status - chilled water outlet temperature setting, guiding the one-hot encoding vector of the chilled water outlet temperature setting and the time-series pattern semantic encoding feature map of the computer room operating status to perform cross-modal interaction optimization to obtain the joint encoding feature map of the computer room operating status - chilled water outlet temperature setting response.

Images