Control method, device and equipment of air conditioner water pump system and storage medium

Abstract

The invention discloses a control method, device and equipment for an air conditioner water pump system and a storage medium. Operation data of the air conditioner water pump system are collected in real time, and feature information of the operation data is extracted; constructing a physical model based on the air conditioner water pump system, establishing a target function and setting constraint conditions; inputting the feature information into a physical model, and training the physical model by taking target function maximization as a target under the constraint of a constraint condition to obtain a first model; an optimization code corresponding to the air conditioner water pump system is generated; calling the first model for simulation through the optimization code, and driving parameter optimization of the first model to determine an optimal parameter; and applying the optimal parameters to the air conditioner water pump system. According to the method, control is conducted through the features to improve the accuracy, the first model is obtained through training, the optimal parameters of the first model are determined through optimization code simulation, accurate and optimized control over the air conditioner water pump system can be achieved, the error rate is greatly reduced, and normal operation of the whole air conditioner system is guaranteed.

Description

Technical Field

[0001] This application relates to the field of control technology for air conditioning water pump systems, specifically to a control method, device, equipment, and storage medium for an air conditioning water pump system. Background Technology

[0002] Air conditioning water pump systems are a core component of modern building HVAC systems. They are mainly used to circulate chilled water or cooling water to achieve heat exchange and temperature regulation. Therefore, precise and flexible control of air conditioning water pump systems is crucial. Traditional control methods for air conditioning water pump systems usually rely on human experience.

[0003] However, relying on human experience is very prone to errors and cannot achieve precise control of the air conditioning water pump system, which will affect the normal operation of the entire air conditioning system. Summary of the Invention

[0004] In view of this, this application provides a control method, device, equipment and storage medium for an air conditioning water pump system, which solves the problem that human experience is very prone to error and cannot achieve precise control of the air conditioning water pump system, thus affecting the normal operation of the entire air conditioning system.

[0005] To achieve the above objectives, the following solution is proposed:

[0006] Firstly, a control method for an air conditioning water pump system includes:

[0007] Real-time acquisition of operating data from the air conditioning water pump system, and extraction of feature information from the operating data;

[0008] A physical model is constructed based on the air conditioning water pump system, an objective function is established, and constraints are set.

[0009] The feature information is input into the physical model, and the physical model is trained under the constraints of the constraints with the objective function as the goal, to obtain the first model;

[0010] Generate optimization code corresponding to the air conditioning water pump system;

[0011] The optimization code calls the first model to perform simulation, driving the parameter optimization of the first model to determine the optimal parameters;

[0012] The optimal parameters are applied to the air conditioning water pump system.

[0013] Preferably, the extraction of feature information from the operational data includes:

[0014] The running data is separated into multiple different types of data, and each type of data is normalized to obtain normalized data corresponding to each type of data.

[0015] For each type of normalized data, extract the feature information of that normalized data.

[0016] Preferably, the step of separating multiple different types of data from the operational data, and then normalizing each type of data to obtain normalized data corresponding to each type of data, includes:

[0017] Separate the continuous sensor data, discrete device status codes, operation logs, and timing alarm data from the operational data;

[0018] The continuous sensor data is subjected to adaptive quantile normalization to obtain the corresponding normalized data;

[0019] The discrete device status codes are encoded using probability distribution to obtain the corresponding normalized data;

[0020] The operation logs are hierarchically embedded to obtain the corresponding normalized data;

[0021] The time-series alarm data is dynamically time-normalized and aligned to obtain the corresponding normalized data.

[0022] Preferably, the physical model constructed based on the air conditioning water pump system includes:

[0023] The feature information is subjected to hidden pattern recognition to obtain multiple hidden patterns at different time scales corresponding to the air conditioning water pump system;

[0024] The feature information is used to mine relationships to generate a causal relationship graph;

[0025] The hidden patterns and causal graphs are used to define the various state variables and input data to form a physical model.

[0026] Preferably, the step of performing hidden pattern recognition on the feature information to obtain multiple hidden patterns at different time scales corresponding to the air conditioning water pump system includes:

[0027] Extract multivariate time-series data from the feature information;

[0028] Cross-time step correlation modeling is performed on the multivariate time series data to obtain a dependency model;

[0029] Set up each short-time window and each long-time window;

[0030] Based on the dependency model, the hidden modes corresponding to each of the short-term and long-term windows are separated.

[0031] Preferably, the step of performing correlation mining on the feature information to generate a causal relationship graph includes:

[0032] From the aforementioned feature information, each sub-feature data is determined;

[0033] Each of the aforementioned sub-feature data is treated as a node;

[0034] For each pair of sub-feature data, determine whether there is a direct influence between these two sub-feature data;

[0035] If so, determine the direction of influence and set a directed edge between the nodes corresponding to these two sub-feature data according to the direction of influence;

[0036] Set the preset first threshold as the weight of the directed edge;

[0037] This forms a causal relationship graph composed of each node and each weighted directed edge.

[0038] Preferably, the objective function is:

[0039] ;

[0040] in, Describe the objective function. Indicates the energy efficiency ratio of the air conditioning water pump system. This indicates the operating cost of the air conditioning water pump system. This represents the maximum reference value for operating costs. , , All are weighting coefficients, and , This indicates the comfort level of the air conditioning water pump system.

[0041] Secondly, a control device for an air conditioning water pump system includes:

[0042] The feature information determination module is used to collect the operating data of the air conditioning water pump system in real time and extract the feature information of the operating data;

[0043] The physical model construction module is used to construct a physical model based on the air conditioning water pump system, establish an objective function, and set constraints.

[0044] The first model acquisition module is used to input the feature information into the physical model, and train the physical model with the objective function maximization as the objective under the constraints of the constraints to obtain the first model;

[0045] An optimization code generation module is used to generate optimization codes corresponding to the air conditioning water pump system.

[0046] The optimal parameter determination module is used to call the first model for simulation through the optimization code, drive the parameter optimization of the first model, and determine the optimal parameters.

[0047] The optimal parameter application module is used to apply the optimal parameters to the air conditioning water pump system.

[0048] Thirdly, a control device for an air conditioning water pump system includes a memory and a processor;

[0049] The memory is used to store programs;

[0050] The processor is configured to execute the program to implement the various steps of the control method for the air conditioning water pump system as described in any of the first aspects.

[0051] Fourthly, a storage medium having a computer program stored thereon, which, when executed by a processor, implements the various steps of the control method for an air conditioning water pump system as described in any of the first aspects.

[0052] As can be seen from the above technical solution, this application collects real-time operating data of the air conditioning water pump system and extracts the feature information of the operating data; constructs a physical model based on the air conditioning water pump system, establishes an objective function and sets constraints; inputs the feature information into the physical model, and trains the physical model with the objective function maximization as the objective under the constraints to obtain a first model; generates optimization code corresponding to the air conditioning water pump system; calls the first model for simulation through the optimization code to drive the parameter optimization of the first model to determine the optimal parameters; and applies the optimal parameters to the air conditioning water pump system. This application first collects real-time operating data of the air conditioning water pump system, thereby extracting feature information containing relevant features of the air conditioning water pump system. Subsequent control using these features can improve accuracy. A physical model, objective function, and constraints are established, thereby training a first model. Then, optimization code is generated, and the first model is simulated using the optimization code to determine the optimal parameters of the first model. Finally, the optimal parameters are applied to the air conditioning water pump system, thus achieving precise and optimized control of the air conditioning water pump system, greatly reducing the error rate and ensuring the normal operation of the entire air conditioning system. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0054] Figure 1 An optional flowchart of a control method for an air conditioning water pump system provided in an embodiment of this application;

[0055] Figure 2 A schematic diagram of the structure of a control device for an air conditioning water pump system provided in an embodiment of this application;

[0056] Figure 3 This is a schematic diagram of the structure of a control device for an air conditioning water pump system provided in an embodiment of this application. Detailed Implementation

[0057] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0058] This invention can be used in a wide variety of general-purpose or special-purpose computing environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor devices, distributed computing environments including any of the above devices, etc.

[0059] This invention provides a control method for an air conditioning water pump system. This method can be applied to various computer terminals or smart terminals, and its execution entity can be the processor or server of the computer terminal or smart terminal. The method flowchart is shown below. Figure 1 As shown, it specifically includes:

[0060] S1: Collect real-time operating data of the air conditioning water pump system and extract the feature information of the operating data.

[0061] Real-time data can be collected from industrial control systems related to air conditioning water pumps, covering various types of data to ensure comprehensiveness.

[0062] After obtaining the operational data, data preprocessing can be performed, such as data cleaning to remove missing values, duplicate values, and obviously erroneous outliers. For example, filtering out values ​​from the water pump flow sensor that are far outside the normal range due to a malfunction can be done. The data can also be organized into matrices, and robust principal component analysis can be used to classify the matrices into low-rank matrices. (Key features) and sparse matrix (Noise / Outliers), the formula is:

[0063] ;

[0064] This represents the total matrix consisting of a low-rank matrix and a coefficient matrix.

[0065] It can also normalize the running data, such as labeling each data source in the running data with modal type, then building a parallel management channel, with each processor corresponding to a specific data type, and then ensuring the spatiotemporal alignment of different modal data through comparative learning loss.

[0066] In the process of extracting feature information, a feature hierarchical architecture can be adopted to extract bottom-level features, middle-level features, and high-level features. Bottom-level features include the operating characteristics of air conditioning water pump equipment, such as the hydraulic performance curve parameters of the water pump, such as the head-flow curve parameters, the efficiency curve parameters of the motor, and the heat transfer coefficient of the air conditioning heat exchanger. Middle-level features are process constraints, such as the cooling and heating load balance equation of the air conditioning system, the flow rate and head relationship equation of the water pump, and the correlation equation between the cooling and heating capacity of the air conditioning unit and its operating parameters. High-level features are physical condition constraints, such as the cooling and heating load balance equation of the air conditioning system, the flow rate and head relationship equation of the water pump, and the correlation equation between the cooling and heating capacity of the air conditioning unit and its operating parameters. High-level features are the operating objectives of the air conditioning system, such as the energy efficiency ratio (COP) of the air conditioning system, the energy consumption ratio of the water pump, and the comfort index of indoor temperature and humidity.

[0067] This step acquires operational data and extracts feature information, transforming the raw data into indicators that are easier for the physical model to understand and process. This lays the foundation for subsequent analysis and optimization, provides a data basis for subsequent modeling and optimization, and the real-time acquisition of operational data can improve the flexibility of control.

[0068] S2: Construct a physical model based on the air conditioning water pump system, establish the objective function, and set the constraints.

[0069] Establishing a physical model can enhance the interpretability of control. Compared with purely data-driven methods, it is more physically meaningful, easier to debug and optimize, and the objective function can clearly define the optimization direction, ensuring that the control process meets actual needs. Constraints can be set by setting flow rate, pressure, temperature, equipment safety limits, process requirements, etc., which can prevent over-limit operation, protect the safety of the air conditioning water pump system, and extend its service life.

[0070] The objective function can be determined based on the optimization objective, such as maximizing energy efficiency or production efficiency. The physical model, objective function, and constraints can all be set according to the air conditioning water pump system, resulting in a higher degree of fit.

[0071] The physical model can also automatically select appropriate optimization objectives and constraints based on different scenarios and problem characteristics, and design scientific and reasonable optimization problems, overcoming the limitations of traditional manual design methods when facing complex data and problems.

[0072] S3: Input the feature information into the physical model, and train the physical model with the objective function maximization as the objective under the constraints to obtain the first model.

[0073] By using actual operating data to optimize the parameters of the physical model, making it more closely resemble the real system and reducing the deviation between the theoretical model and the actual working conditions, compared with pure black box models (such as deep learning), the physical model + data training method can reduce the risk of overfitting and improve the generalization ability while ensuring accuracy. The first model after training can provide a reliable basis for parameter optimization.

[0074] S4: Generate optimization code corresponding to the air conditioning water pump system.

[0075] This application automatically invokes models and optimization algorithms through code, reducing manual intervention and improving computational efficiency. Furthermore, the optimization code can adjust the optimization strategy according to different air conditioning water pump systems, which can enhance the versatility of the method.

[0076] You can set an optimization problem, and based on the definition of the optimization problem, use a large model to call the automatic code generation function to generate the corresponding optimization code based on Python, genetic algorithm or simulated annealing algorithm, etc.

[0077] For example, in optimizing the energy efficiency of air conditioning water pump systems, genetic algorithms are used to write code to find the optimal combination of variable frequency speed control parameters for the water pump and operating parameters for the air conditioning unit to maximize the energy efficiency ratio. The data generated by executing the optimization code is then used to fine-tune the large model using LoRA, further enhancing its code generation and optimization capabilities in the field of air conditioning water pump system optimization, and continuously improving the quality and accuracy of the optimization code. This not only greatly improves the efficiency and accuracy of code writing but also avoids errors and defects that may occur with manually written code. Furthermore, the large model can flexibly generate diverse code structures and logic according to different optimization problems and algorithm requirements, adapting to various complex optimization scenarios. This solves the problem of dependence on professional programmers in traditional methods, reducing labor costs and the risk of errors.

[0078] S5: The optimization code calls the first model to perform simulation, driving the parameter optimization of the first model to determine the optimal parameters.

[0079] The optimization code is executed in a sandbox environment that simulates the actual operating environment of the air conditioning water pump system, including equipment parameters, operating conditions, control logic, etc., to ensure the safety and reliability of code execution and avoid interference and impact on the normal operation of the actual air conditioning water pump system. At the same time, the optimization code is fully tested and verified to identify and correct potential problems in a timely manner.

[0080] After the optimization code is executed, the patrol results are analyzed and interpreted in depth to extract key information, namely the optimal parameters, including the variable frequency of the water pump, the cooling capacity setting of the air conditioning unit, the fan speed of the cooling tower, etc., and the performance improvement (such as the percentage increase in energy efficiency ratio, the specific value of energy consumption reduction, the degree of reduction in indoor temperature fluctuation, etc.). Information can be extracted from the preceding large model analysis nodes by setting certain parameters for the large model Agent node. The control commands are organized into / params, with the following structure: [

[0081] {

[0082] "linkId": "xxx",

[0083] "paramId": "xxx:xxxxxx",

[0084] value: "xx"

[0085] } ]

[0087] Executing the optimization code in a sandbox environment not only ensures the security and reliability of the code operation, but also allows for the full testing and verification of the effectiveness and feasibility of the optimization scheme without affecting the actual air conditioning water pump control system. The large model's ability to interpret the optimization results further enhances the practicality and user-friendliness of the invention. It can transform the complex optimization process and results into easy-to-understand language and knowledge, helping users better understand the meaning and value of optimization, providing strong support for subsequent decision-making and optimization, and overcoming the shortcomings of traditional optimization systems in terms of result interpretation and user communication.

[0088] S6: Apply the optimal parameters to the air conditioning water pump system.

[0089] Finally, by applying the optimal parameters to the air conditioning water pump system, precise control of the air conditioning water pump system can be achieved.

[0090] In controlling the air conditioning water pump system using optimal parameters, the control results can be presented to users in an intuitive way, such as generating charts, like energy efficiency change curves and indoor temperature control effect diagrams; detailed reports, such as a visual interface explaining the optimization scheme, expected benefits, and implementation steps, to help users quickly understand the optimization effect and key information; at the same time, the feedback results are transmitted to the industrial control system of the air conditioning water pump system for real-time adjustment of the control strategy, optimization of the operation process of the air conditioning water pump system, and realization of efficient, stable and economical operation of the system.

[0091] As can be seen from the above technical solution, this application collects real-time operating data of the air conditioning water pump system and extracts the feature information of the operating data; constructs a physical model based on the air conditioning water pump system, establishes an objective function and sets constraints; inputs the feature information into the physical model, and trains the physical model with the objective function maximization as the objective under the constraints to obtain a first model; generates optimization code corresponding to the air conditioning water pump system; calls the first model for simulation through the optimization code to drive the parameter optimization of the first model to determine the optimal parameters; and applies the optimal parameters to the air conditioning water pump system. This application first collects real-time operating data of the air conditioning water pump system, thereby extracting feature information containing relevant features of the air conditioning water pump system. Subsequent control using these features can improve accuracy. A physical model, objective function, and constraints are established, thereby training a first model. Then, optimization code is generated, and the first model is simulated using the optimization code to determine the optimal parameters of the first model. Finally, the optimal parameters are applied to the air conditioning water pump system, thus achieving precise and optimized control of the air conditioning water pump system, greatly reducing the error rate and ensuring the normal operation of the entire air conditioning system.

[0092] Among existing optimization methods based on system control, parameter optimization based on model predictive control (MPC) is a model-based closed-loop optimization control strategy widely used in complex process control in industries such as oil refining, chemical engineering, metallurgy, central air conditioning, and power generation. However, the selection of parameters such as penalty coefficients and prediction time domain length in MPC controllers has a significant impact on controller performance. Traditional manual trial-and-error methods are not only cumbersome but also difficult to achieve optimization results. Another example is the parameter optimization method for intelligent manufacturing based on machine learning and the Industrial Internet of Things (IIoT). This method constructs a digital and intelligent production line model, utilizes the IIoT to centrally collect and process equipment data, and then uses machine learning models to calculate the optimal value for each part, providing engineers with adjustment benchmarks to reduce debugging workload and improve adjustment accuracy. However, this method mainly focuses on optimizing the overall efficiency of the production line and does not fully consider the professional knowledge and special needs of the control field for optimizing industrial control parameters.

[0093] Therefore, in traditional industrial production, the optimization of control parameters mainly relies on manual experience and trial-and-error methods. However, with the increasing complexity of industrial systems and the growing number of parameters, manual debugging methods are no longer sufficient to meet the demands for efficient and accurate optimization. Furthermore, various interference factors and uncertainties in industrial production processes further increase the difficulty of parameter optimization. Consequently, large-scale models are gaining increasing attention in the industrial field, especially in air conditioning, as their powerful data processing and knowledge discovery capabilities provide new ideas and methods for solving industrial problems. For example, large-scale models have demonstrated good performance and application prospects in fault diagnosis, quality inspection, and production scheduling. Meanwhile, automatic code generation technology is maturing in software development and engineering applications. By defining code generation rules and templates, the function of automatically generating code according to specific needs can be achieved. Therefore, this application combines the knowledge and logical reasoning capabilities of large-scale models to realize the function of automatically writing optimization code, further improving the automation level and efficiency of control parameter optimization in air conditioning water pump systems.

[0094] With the development of artificial intelligence technology, large models have demonstrated powerful capabilities in data processing, knowledge extraction, and problem solving. Therefore, this application, based on these advantages of large models, introduces them into the field of control parameter optimization for air conditioning water pump systems. Large models can organize and analyze massive amounts of industrial field data, uncover potential patterns and knowledge within the data, and provide a more accurate basis for constructing optimization problems. At the same time, large models also have the ability to automatically generate code, which can automatically generate corresponding optimization code according to the requirements of the optimization problem, thereby improving the efficiency and accuracy of optimization.

[0095] This application addresses the problem that manual experience is prone to errors and cannot achieve precise control of the air conditioning water pump system, which can affect the normal operation of the entire air conditioning system. It also solves the shortcomings of existing automatic optimization systems that lack intelligent code writing and a safe testing environment. Through the design and optimization of the physical model, the quality and efficiency of optimization are improved. At the same time, the sandbox environment ensures the safety of code execution and avoids potential risks to the air conditioning water pump system.

[0096] The process for extracting feature information from the running data in the method provided by this embodiment of the invention is described in detail below:

[0097] The running data is separated into multiple different types of data, and each type of data is normalized to obtain normalized data corresponding to each type of data.

[0098] For each type of normalized data, extract the feature information of that normalized data.

[0099] Specifically, this step employs an innovative hierarchical normalization strategy, performing normalization processing separately for different types of data, which improves the accuracy of normalization.

[0100] Specifically, the step of separating multiple different types of data from the operational data, and normalizing each type of data to obtain normalized data corresponding to each type of data, may include:

[0101] Separate the continuous sensor data, discrete device status codes, operation logs, and timing alarm data from the operational data;

[0102] The continuous sensor data is subjected to adaptive quantile normalization to obtain the corresponding normalized data;

[0103] The discrete device status codes are encoded using probability distribution to obtain the corresponding normalized data;

[0104] The operation logs are hierarchically embedded to obtain the corresponding normalized data;

[0105] The time-series alarm data is dynamically time-normalized and aligned to obtain the corresponding normalized data.

[0106] Specifically, continuous sensor data includes data from flow rate, head, and water temperature sensors of the air conditioning water pump system, and data from temperature and humidity sensors of the air conditioning system; discrete equipment status codes include speed setpoints, start-stop control parameters, operating mode parameters of the air conditioning unit, or operating status of the air conditioning water pump system; operation logs include equipment start-stop times, maintenance operation records, parameter adjustment logs, etc.

[0107] Specifically, for physical quantity data such as flow rate and head, linear normalization is performed based on their dimensions and normal operating range; for environmental parameter data such as temperature and humidity, normalization is performed in conjunction with the comfort requirements of the air conditioning system; for discrete data in control parameters and operation logs, one-heat encoding and other processing are performed to bring their values ​​to the same scale for easier subsequent analysis.

[0108] Therefore, based on the above ideas, the normalization methods for the above four types of data are summarized as adaptive quantile normalization, probability distribution coding, hierarchical embedding, and dynamic time warping alignment, thereby generating their respective normalized data.

[0109] Among them, adaptive quantile normalization processing The calculation formula is:

[0110] ;

[0111] Taking the flow rate of the air conditioning water pump system as an example, This represents the 90th percentile flow rate value in the dataset corresponding to the flow rate of the air conditioning water pump system, meaning that 90% of the observations are less than or equal to this flow rate value. This represents the flow rate value at the 10th percentile in the dataset, meaning that 10% of the observations are less than or equal to this flow rate value. This represents the maximum traffic value in the dataset; This represents the minimum flow value in the dataset.

[0112] The formula for calculating probability distribution coding is:

[0113] ;

[0114] Taking the operating status of an air conditioning water pump system as an example, Indicates the first of the air conditioning water pump system Each running status; This represents the total number of operating states of the air conditioning water pump system. Indicates the first The frequency of occurrence of each running state A set representing the running state, which may include loading, unloading, stopping, running, etc., but this embodiment does not limit this.

[0115] Hierarchical embedding The calculation formula is:

[0116] ;

[0117] Represents the operation log The model encoding process, for example, encodes "the frequency of the No. 1 air conditioner water pump system jumps to 45Hz" as... Eigenvalues; Represents coded symbols, This indicates the specific parameter value.

[0118] The formula for calculating dynamic time warping alignment is:

[0119] ;

[0120] Taking the low performance alarm of the air conditioning water pump system as an example, This represents the frequency data matrix of the air conditioning water pump system. This represents the alarm information matrix of the air conditioning water pump system. Represents the correlation coefficient. express Data in the time-frequency data matrix, express The data in the real-time alarm information matrix is ​​obtained through... The calculation can align the timing of frequency data and alarm information, and then study the correlation between the two.

[0121] The process of constructing a physical model based on the air conditioning water pump system in this application will be described in detail below.

[0122] The feature information is subjected to hidden pattern recognition to obtain multiple hidden patterns at different time scales corresponding to the air conditioning water pump system;

[0123] The feature information is used to mine relationships to generate a causal relationship graph;

[0124] The hidden patterns and causal graphs are used to define the various state variables and input data to form a physical model.

[0125] This application identifies key factors affecting the control performance of air conditioning water pump systems by mining hidden patterns and relationships in feature information, thereby establishing a physical model of the equipment.

[0126] Dynamic physical models can be constructed using the concept of neural ODEs (neural ordinary differential equations).

[0127] ;

[0128] in, It is a neural network that learns the derivative (rate of change) of a state. State variables representing the air conditioning water pump system, including flow rate, head, and speed; The control inputs, including the inverter output frequency, are represented. Through order reduction processing, the state dimension of the air conditioning water pump system is compressed from the original complex multi-dimensional (including many parameters such as motor winding temperature and bearing temperature) to three main dimensions that reflect the operating performance of the air conditioning water pump system: flow rate, head, and efficiency. While ensuring the accuracy of the physical model, the simulation calculation speed is greatly improved, and the efficiency of subsequent optimization analysis is enhanced.

[0129] Specifically, the process of performing hidden pattern recognition on the aforementioned feature information to obtain multiple hidden patterns at different time scales corresponding to the air conditioning water pump system may include:

[0130] Extract multivariate time-series data from the feature information;

[0131] Cross-time step correlation modeling is performed on the multivariate time series data to obtain a dependency model;

[0132] Set up each short-time window and each long-time window;

[0133] Based on the dependency model, the hidden modes corresponding to each of the short-term and long-term windows are separated.

[0134] Specifically, the Self-Attention mechanism of Transformer can be used to model the correlation across time steps of multivariate time-series data of air conditioning water pump systems, including real-time flow rate, head, and motor power of water pumps, and indoor temperature and humidity of air conditioning systems. This yields a dependency model, and different attention heads can be set as key-value dimensions to correspond to various short-term windows, such as the last 10 time steps, which can capture the correlation between water pump flow fluctuations and motor power changes within a few minutes. For various long-term windows, such as 100 time steps, the correlation between the cooling and heating load change cycles of the air conditioning system and the water pump operation mode over a span of several hours or even longer can be analyzed. This allows for the separation of hidden patterns at different time scales, such as the transient flow change pattern caused by short-term water pump start-stop and the regular influence pattern of long-term diurnal load changes of the air conditioning system on water pump operation.

[0135] Specifically, the process of mining the relationships between the aforementioned feature information to generate a causal relationship graph may include:

[0136] From the aforementioned feature information, each sub-feature data is determined;

[0137] Each of the aforementioned sub-feature data is treated as a node;

[0138] For each pair of sub-feature data, determine whether there is a direct influence between these two sub-feature data;

[0139] If so, determine the direction of influence and set a directed edge between the nodes corresponding to these two sub-feature data according to the direction of influence;

[0140] Set the preset first threshold as the weight of the directed edge;

[0141] This forms a causal relationship graph composed of each node and each weighted directed edge.

[0142] Specifically, the Peter-Clark Algorithm (PC) is used in conjunction with the domain constraints of the air conditioning water pump system to construct a causal graph. For example, it is found that the cooling water supply temperature in the air conditioning water pump system has a direct impact on the chiller's cooling efficiency, and this path is controlled by the cooling tower fan speed as a common parent node. Since the cooling tower fan speed affects the cooling water's heat dissipation effect, it changes the degree to which the supply water temperature affects the cooling efficiency. Therefore, the cooling water supply temperature and the chiller's cooling efficiency are treated as two nodes, and a directed edge is established between them. The direction of the edge is the direction of influence, i.e., from the cooling water supply temperature to the chiller's cooling efficiency, and the edge weight is set to the first threshold, such as 0.78.

[0143] The following embodiments provide a detailed explanation of the objective function in this application. Since multiple optimization objectives often need to be considered in practical applications, such as high energy efficiency, low operating costs, and ensuring indoor comfort, this application establishes the following objective function considering multiple aspects:

[0144] ;

[0145] in, Describe the objective function. Indicates the energy efficiency ratio of the air conditioning water pump system. This indicates the operating cost of the air conditioning water pump system. This represents the maximum reference value for operating costs. , , All are weighting coefficients, and , This indicates the comfort level of the air conditioning water pump system.

[0146] The specific value of the weighting coefficient can be set according to the actual priority and importance. For example, if the energy efficiency ratio is the most important, the weighting coefficient can be set to a value that reflects the actual priority and importance of the energy efficiency ratio. Set the value to 0.4~0.5, then the operating cost to 0.3~0.4, and the comfort level to around 0.2~0.3.

[0147] Specifically, energy efficiency ratio (COP, Coefficient of Performance) is one of the important indicators for measuring the performance of an air conditioning system. For cooling mode, It can be represented as:

[0148] ;

[0149] in, It represents the heat absorbed from the cooled space (such as the indoor environment), and the unit is kilowatt (kW). It is the part of the heat that is released to the outside after the evaporator absorbs indoor heat and transfers it to the condenser through the refrigeration cycle of the air conditioning system. In actual calculations, it can be derived from the cooling capacity parameters of the air conditioning system, the rate of change of indoor temperature, and relevant thermodynamic formulas. This indicates the electrical power consumed by equipment such as air conditioning compressors, measured in kilowatts (kW). It includes the compressor's input power as well as the power consumption of other auxiliary equipment, such as fans and air conditioning water pump systems. Compressor power can be determined based on its performance curve and operating parameters (such as refrigerant flow rate, condensing pressure, and evaporating pressure). The power of fans and air conditioning water pump systems, however, is related to factors such as flow rate, head, and efficiency, and can be calculated using corresponding power formulas. For example, the power calculation formula for an air conditioning water pump system is:

[0150] ;

[0151] in, This indicates the power of the air conditioning water pump system, measured in kilowatts (kW). This indicates the flow rate of the air conditioning water pump system, which is the volume of liquid delivered by the air conditioning water pump system per unit time. The unit is usually cubic meters per hour (m³ / h). The head of the air conditioning water pump system is the height that the air conditioning water pump system can lift the liquid to, and the unit is meters (m). This indicates the density of the pumped liquid, expressed in kilograms per cubic meter (kg / m³). This is the acceleration due to gravity, usually taken as 9.81 m / s². The efficiency of an air conditioning water pump system is a dimensionless ratio, representing the ratio of the system's output power to its input power.

[0152] By setting the above objective function, we can comprehensively measure each key factor and find the operating parameters of the air conditioning water pump system that optimize the overall comprehensive benefits under certain constraints.

[0153] This application can efficiently organize and deeply analyze massive and complex field data of air conditioning water pump systems, and uncover hidden patterns and key information in the data. This provides a precise basis for the design of optimization problems, ensuring that the designed optimization problems can accurately reflect the actual control needs and operating conditions of the air conditioning water pump system.

[0154] and Figure 1 Corresponding to the method described above, embodiments of the present invention also provide a control device for an air conditioning water pump system, used for controlling... Figure 1 In a specific implementation of the method, the control device for the air conditioning water pump system provided in this embodiment of the invention can be located in a computer terminal or various mobile devices, combined with... Figure 2 The control device of the air conditioning water pump system is introduced, such as... Figure 2 As shown, the device may include:

[0155] The feature information determination module 10 is used to collect the operating data of the air conditioning water pump system in real time and extract the feature information of the operating data;

[0156] The physical model construction module 20 is used to construct a physical model based on the air conditioning water pump system, establish an objective function and set constraints.

[0157] The first model acquisition module 30 is used to input the feature information into the physical model, and train the physical model with the objective function maximization as the objective under the constraints of the constraints to obtain the first model;

[0158] The optimization code generation module 40 is used to generate optimization codes corresponding to the air conditioning water pump system;

[0159] The optimal parameter determination module 50 is used to call the first model for simulation through the optimization code, drive the parameter optimization of the first model, and determine the optimal parameters.

[0160] The optimal parameter application module 60 is used to apply the optimal parameters to the air conditioning water pump system.

[0161] As can be seen from the above technical solution, this application collects real-time operating data of the air conditioning water pump system and extracts the feature information of the operating data; constructs a physical model based on the air conditioning water pump system, establishes an objective function and sets constraints; inputs the feature information into the physical model, and trains the physical model with the objective function maximization as the objective under the constraints to obtain a first model; generates optimization code corresponding to the air conditioning water pump system; calls the first model for simulation through the optimization code to drive the parameter optimization of the first model to determine the optimal parameters; and applies the optimal parameters to the air conditioning water pump system. This application first collects real-time operating data of the air conditioning water pump system, thereby extracting feature information containing relevant features of the air conditioning water pump system. Subsequent control using these features can improve accuracy. A physical model, objective function, and constraints are established, thereby training a first model. Then, optimization code is generated, and the first model is simulated using the optimization code to determine the optimal parameters of the first model. Finally, the optimal parameters are applied to the air conditioning water pump system, thus achieving precise and optimized control of the air conditioning water pump system, greatly reducing the error rate and ensuring the normal operation of the entire air conditioning system.

[0162] Furthermore, embodiments of this application provide a control device for an air conditioning water pump system. Optionally, Figure 3 The hardware structure block diagram of the control equipment for the air conditioning water pump system is shown. (Refer to...) Figure 3 The hardware structure of the control equipment for the air conditioning water pump system may include: at least one processor 01, at least one communication interface 02, at least one memory 03 and at least one communication bus 04.

[0163] In this embodiment, the number of processor 01, communication interface 02, memory 03 and communication bus 04 is at least one, and processor 01, communication interface 02 and memory 03 communicate with each other through communication bus 04.

[0164] Processor 01 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0165] Memory 03 may include high-speed RAM, and may also include non-volatile memory, such as at least one disk storage device.

[0166] The memory stores a program, which the processor can call to execute the control method of the air conditioning water pump system, including:

[0167] Real-time acquisition of operating data from the air conditioning water pump system, and extraction of feature information from the operating data;

[0168] A physical model is constructed based on the air conditioning water pump system, an objective function is established, and constraints are set.

[0169] The feature information is input into the physical model, and the physical model is trained under the constraints of the constraints with the objective function as the goal, to obtain the first model;

[0170] Generate optimization code corresponding to the air conditioning water pump system;

[0171] The optimization code calls the first model to perform simulation, driving the parameter optimization of the first model to determine the optimal parameters;

[0172] The optimal parameters are applied to the air conditioning water pump system.

[0173] Optionally, the refined and extended functions of the program can be found in the description of the control method for the air conditioning water pump system in the method embodiments.

[0174] This application embodiment also provides a storage medium that can store a program suitable for execution by a processor. When the program runs, it controls the device where the storage medium is located to execute the following control method for an air conditioning water pump system, including:

[0175] Real-time acquisition of operating data from the air conditioning water pump system, and extraction of feature information from the operating data;

[0176] A physical model is constructed based on the air conditioning water pump system, an objective function is established, and constraints are set.

[0177] The feature information is input into the physical model, and the physical model is trained under the constraints of the constraints with the objective function as the goal, to obtain the first model;

[0178] Generate optimization code corresponding to the air conditioning water pump system;

[0179] The optimization code calls the first model to perform simulation, driving the parameter optimization of the first model to determine the optimal parameters;

[0180] The optimal parameters are applied to the air conditioning water pump system.

[0181] Specifically, the storage medium can be a computer-readable storage medium, which can be an electronic storage device such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM.

[0182] Optionally, the refined and extended functions of the program can be found in the description of the control method for the air conditioning water pump system in the method embodiments.

[0183] Furthermore, the functional modules in the various embodiments of this disclosure can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part. If the function is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a live streaming device, or a network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this disclosure.

[0184] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0185] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0186] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A control method for an air conditioning water pump system, characterized in that, include: Real-time acquisition of operating data from the air conditioning water pump system, and extraction of feature information from the operating data; A physical model is constructed based on the air conditioning water pump system, an objective function is established, and constraints are set. The feature information is input into the physical model, and the physical model is trained under the constraints of the constraints with the objective function as the goal, to obtain the first model; Generate optimization code corresponding to the air conditioning water pump system; The optimization code calls the first model to perform simulation, driving the parameter optimization of the first model to determine the optimal parameters; The optimal parameters are applied to the air conditioning water pump system.

2. The method according to claim 1, characterized in that, The extraction of feature information from the operational data includes: The running data is separated into multiple different types of data, and each type of data is normalized to obtain normalized data corresponding to each type of data. For each type of normalized data, extract the feature information of that normalized data.

3. The method according to claim 2, characterized in that, The process of separating multiple different types of data from the operational data, and then normalizing each type of data to obtain normalized data corresponding to each type of data, includes: Separate the continuous sensor data, discrete device status codes, operation logs, and timing alarm data from the operational data; The continuous sensor data is subjected to adaptive quantile normalization to obtain the corresponding normalized data; The discrete device status codes are encoded using probability distribution to obtain the corresponding normalized data; The operation logs are hierarchically embedded to obtain the corresponding normalized data; The time-series alarm data is dynamically time-normalized and aligned to obtain the corresponding normalized data.

4. The method according to claim 1, characterized in that, The physical model constructed based on the air conditioning water pump system includes: The feature information is subjected to hidden pattern recognition to obtain multiple hidden patterns at different time scales corresponding to the air conditioning water pump system; The feature information is used to mine relationships to generate a causal relationship graph; The hidden patterns and causal graphs are used to define the various state variables and input data to form a physical model.

5. The method according to claim 4, characterized in that, The process of performing hidden pattern recognition on the feature information yields multiple hidden patterns at different time scales corresponding to the air conditioning water pump system, including: Extract multivariate time-series data from the feature information; Cross-time step correlation modeling is performed on the multivariate time series data to obtain a dependency model; Set up each short-time window and each long-time window; Based on the dependency model, the hidden modes corresponding to each of the short-term and long-term windows are separated.

6. The method according to claim 4, characterized in that, The step of mining the correlations of the feature information to generate a causal relationship graph includes: From the aforementioned feature information, each sub-feature data is determined; Each of the aforementioned sub-feature data is treated as a node; For each pair of sub-feature data, determine whether there is a direct influence between these two sub-feature data; If so, determine the direction of influence and set a directed edge between the nodes corresponding to these two sub-feature data according to the direction of influence; Set the preset first threshold as the weight of the directed edge; This forms a causal relationship graph composed of each node and each weighted directed edge.

7. The method according to any one of claims 1 to 6, characterized in that, The objective function is: ； in, Describe the objective function. Indicates the energy efficiency ratio of the air conditioning water pump system. This indicates the operating cost of the air conditioning water pump system. This represents the maximum reference value for operating costs. , , All are weighting coefficients, and , This indicates the comfort level of the air conditioning water pump system during operation.

8. A control device for an air conditioning water pump system, characterized in that, include: The feature information determination module is used to collect the operating data of the air conditioning water pump system in real time and extract the feature information of the operating data; The physical model construction module is used to construct a physical model based on the air conditioning water pump system, establish an objective function, and set constraints. The first model acquisition module is used to input the feature information into the physical model, and train the physical model with the objective function maximization as the objective under the constraints of the constraints to obtain the first model; An optimization code generation module is used to generate optimization codes corresponding to the air conditioning water pump system; The optimal parameter determination module is used to call the first model for simulation through the optimization code, drive the parameter optimization of the first model, and determine the optimal parameters. The optimal parameter application module is used to apply the optimal parameters to the air conditioning water pump system.

9. A control device for an air conditioning water pump system, characterized in that, Including memory and processor; The memory is used to store programs; The processor is used to execute the program to implement the various steps of the control method for the air conditioning water pump system as described in any one of claims 1-7.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the various steps of the control method for the air conditioning water pump system as described in any one of claims 1-7.

Images