An air conditioner distributed regulation system based on neural network and a method thereof

Abstract

The application relates to the technical field of intelligent control of a building environment, and provides an air conditioner distributed regulation and control system and method based on a neural network, which is suitable for energy efficiency optimization and comfort guarantee of a multi-zone building. The system comprises a coordination control module, an optimization decomposition module, a regional control module, a model evolution module and a strategy self-learning module, a multi-objective optimization model is constructed, decomposition is carried out based on an alternating direction multiplier method, regional-level scheduling and parallel solution of a control task are realized, a control model is constructed by using a hypergraph neural network and online updating, migration deployment and adaptive adjustment are supported, strategy self-learning is realized through a strategy evaluation and aggregation mechanism driven by a control error, and a regulation and control closed loop is formed. The system improves the precision, flexibility and energy efficiency performance of air conditioner regulation and control, and is suitable for building scenes such as intelligent buildings and green parks.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control of building environment, and particularly relates to an air conditioner distributed regulation and control system based on neural network and a method thereof. BACKGROUND

[0002] With the continuous improvement of the demand for building intelligence and fine energy consumption management, as the main component of building energy consumption, air conditioning systems urgently need to realize more efficient, flexible and intelligent regulation and control capabilities. In multi-zone building scenarios, due to the significant differences in orientation, volume, thermal parameters, personnel activity intensity, etc., the cold load demand, comfort setting and energy consumption constraints of each region present high heterogeneity and time variability. Such differences pose the following challenges to air conditioning control systems:

[0003] Strong regional coupling, low centralized control efficiency: Most current building air conditioning control systems use centralized architecture, and it is difficult for unified control strategies to adapt to the individualized needs of different regions in terms of thermal inertia, thermal coupling degree and load variation, which easily causes regulation and control error accumulation and comfort deviation. Control model is static and lacks evolution mechanism: In existing systems, the control model is usually established based on initial building parameters, and lacks dynamic updating mechanism, which makes it difficult to adapt to seasonal changes, thermal performance aging or space use changes, etc., resulting in continuous decline of control effect. Control task instruction is coarse-grained and lacks deployability: Traditional scheduling output is mostly global set value, lacking structured control instructions for regional allocation and time step mapping, which cannot effectively decouple with regional execution modules, affecting the flexibility and implementation efficiency of regulation and control. Control strategy update lags behind, feedback loop is missing: Control errors cannot effectively drive strategy adjustment, and the system usually relies on manual parameter adjustment or periodic update, and cannot realize adaptive strategy optimization according to operation data. Difficulty in decoupling regional optimization, high computational cost: The regional state has a natural coupling relationship in the control task, and solving the global optimization problem uniformly is computationally complex and has poor scalability, which is not conducive to real-time protection and parallel execution.

[0004] In view of the above problems, it is urgent to build an air conditioner distributed regulation and control system for multi-zone building scenarios to realize a precise, adjustable and evolving building air conditioning intelligent control system. SUMMARY

[0005] The present application aims to overcome the problems of control centralization, model staticity, strategy lag and optimization decoupling in existing building air conditioning systems in multi-zone distributed scenarios, and provides an air conditioner distributed regulation and control system based on neural network with local adaptive modeling capability, global coordinated control capability, dynamic strategy evolution capability and feedback driven learning capability, and a regulation and control method thereof, to improve the overall regulation and control precision, energy efficiency and deployment flexibility of the system.

[0006] In order to achieve the above object, the present application is realized by the following technical scheme:

[0007] A neural network-based air conditioner distributed regulation and control system, comprising:

[0008] A coordination control module, configured to acquire environmental state information, energy consumption constraint parameters and user comfort degree settings of multiple building areas within a control period, construct a multi-objective optimization model containing a thermal dynamics state equation, a power boundary condition and a user comfort degree interval, generate control task instructions matched with area states and regulation and control constraints based on a solution result of the multi-objective optimization model, and distribute the control task instructions to area control modules to complete distribution and scheduling of the control task instructions;

[0009] An optimization decomposition module integrated in the coordination control module, configured to decompose the multi-objective optimization model into multiple area control sub-problems based on an alternating direction multiplier method, and perform regulation and control efficiency matching adjustment according to a regulation and control error mechanism;

[0010] An area control module deployed in each building area, configured to receive the control task instructions distributed by the coordination control module, and execute air conditioner equipment regulation and control operations in combination with environmental state information of the area, and feed back control error information to the coordination control module;

[0011] A model evolution module, configured to construct a neural network control model using a hypergraph neural network structure according to building characteristic parameters and control task requirements of each area, and update, migrate, deploy and adaptively adjust the neural network control model during operation;

[0012] A strategy self-learning module, configured to execute control strategy updating according to the control error information after the control period ends, and update the neural network control model through a strategy evaluation mechanism and a strategy aggregation mechanism.

[0013] Further improvement of the present application is that the coordination control module comprises:

[0014] A state aggregation unit, configured to perform dimension reduction mapping and discretization processing based on state vectors fed back by the area control modules, and construct a unified system state space;

[0015] An optimization modeling unit, configured to construct a multi-objective optimization model in the following form based on the unified system state space, energy consumption constraint parameters and user comfort degree intervals:

[0016]

[0017] Wherein, E(u) is an energy consumption function, D(u,r) is a comfort degree deviation function, C(u) is a control scheduling cost function, and α1, α2 and α3 are weighting coefficients; u is a control instruction vector; and r is a user expected value vector.

[0018] The control task generation unit is used to generate control task instructions for multiple regions based on the solution results of the multi-objective optimization model.

[0019] The instruction allocation and scheduling unit is used to combine the constraint status and boundary power conditions of each region to execute the task mapping and scheduling allocation of the control task instructions, form a scheduling table matched by region number and time step, and send it to the region control module.

[0020] A further improvement of the present invention is that the control task generation unit in the coordination control module is defined as: generating structured control task instructions based on the output results of the optimization modeling unit and the optimization decomposition module, and establishing a time series matching relationship with the control equipment in each region, so that the control task instructions have the functions of region identification, control target expression and state feedback reservation, and are adapted to the scheduling and execution of the regional control module.

[0021] A further improvement of the present invention is that the optimization decomposition module includes:

[0022] The regional variable decomposition unit is used to divide the global control vector in the multi-objective optimization model constructed by the coordination control module into several regional execution vectors a. j And introduce a synchronization variable g j With multiplier variable ξ j Construct a regional optimization problem of the following form:

[0023]

[0024] Where: φ j (a j ) represents the local cost function of the j-th region; a j For the air conditioning settings (such as airflow and set temperature) of this area; g j ξ is the system synchronization target vector; j η is the multiplier estimation vector; η is the coordination factor, which adjusts the local-global coupling strength; N is the number of regions participating in parallel optimization computation in the control system;

[0025] The alternating update unit is used to perform the following steps within the control cycle:

[0026]

[0027] Feedback coupling unit, used to calculate the convergence error for each region:

[0028]

[0029] Where: k is the iteration counter in the optimization algorithm, used to identify the state of the current k-th iteration; and These are the original residual and the dual residual, respectively, reflecting the degree of deviation between the current regional control output and the synchronization reference, serving as an evaluation index for control coordination performance; and the above residuals are fed back to the coordination control module to update the coordination factor η or adjust the control task priority configuration table for the next cycle.

[0030] A further improvement of the present invention is that the area control module includes:

[0031] The instruction parsing unit is used to receive and parse the control task instructions issued by the coordination and control module, and extract the area identifier field, execution time period field, control target field and constraint status field therein;

[0032] The execution control unit is used to combine the real-time environmental status information of the area with the control objective to determine whether the execution conditions are met, including but not limited to equipment availability and boundary constraint feasibility. If the conditions are met, the corresponding air conditioning equipment is driven to perform control operations.

[0033] The feedback generation unit is used to calculate control error information based on the deviation between the control result and the target after the control operation is completed, and to feed back the control error information and the control task execution status to the coordination control module.

[0034] A further improvement of the present invention is that the model evolution module includes:

[0035] The feature encoding unit is used to extract building feature parameters for each region, including building orientation, thermal resistance of the building envelope, window-to-wall ratio, volumetric heat capacity, and ventilation rate, and encode them into structural vectors of regional topological nodes.

[0036] The hypergraph generation unit is used to construct a hypergraph structure based on the above-mentioned regional node structure vectors. Each hyperedge in the hypergraph structure connects multiple building region nodes to represent the thermal coupling relationship, load linkage relationship or energy consumption similarity between regions.

[0037] The control model construction unit is used to input the region regulation task description based on the hypergraph structure, including historical control variables, target temperature sequence, and disturbance boundary, and to train the hypergraph neural network model and output the initial neural network control model parameter set required for each control cycle.

[0038] The model update unit is used to perform online parameter updates, cross-regional migration deployments, or structural adjustments on the neural network control model during system operation, based on the control error information and task completion status fed back by the regional control module, through continuous learning mechanisms including distillation migration, adaptive regression, or retraining.

[0039] A further improvement of the present invention is that the model update unit is defined as including an update trigger judgment unit, used to determine, within each control cycle, whether any of the following conditions are met based on the control error information and state trajectory fed back by the area control module:

[0040] The control error exceeds the set drift threshold for multiple consecutive cycles;

[0041] There is a time lag between the control model's predictions and the actual equipment response;

[0042] The regional building thermal parameters change beyond the set boundary values;

[0043] When any condition is met, the parameter fine-tuning strategy of the neural network control model is executed, and the adaptive regression method is used to perform incremental training on the original model within a finite number of steps with the current error information as the optimization target.

[0044] A further improvement of the present invention is that the model update unit further includes:

[0045] The migration alignment module is used to select reference control models with thermal characteristic similarity higher than a set threshold from the pre-trained control model library based on the building feature parameters of the target area after the update is triggered.

[0046] Based on the current control input state sequence and the output prediction results of the reference control model, a parameter alignment training process for the target region control model is constructed, and its loss function is defined as:

[0047]

[0048] in: The control quantity predicted by the target area control model; The target value for optimized control in the current cycle; The reference control model outputs the same inputs; β∈(0,1) are the weighting coefficients of the loss function; θ T The target region control model is a set of adjustable parameters; KL represents the Kullback-Leibler divergence; after training, the target region control model is deployed to the corresponding region control module to replace the original model and perform subsequent control tasks.

[0049] A further improvement of the present invention is that the policy self-learning module includes:

[0050] The strategy evaluation unit is used to quantitatively evaluate the performance of the current control strategy based on the control error information fed back by the regional control module after the control cycle ends. The evaluation indicators include at least control accuracy, user comfort achievement rate, equipment energy consumption deviation value and response latency.

[0051] The strategy update triggering unit is used to trigger the update operation of the control strategy when any evaluation indicator exceeds the set deviation threshold.

[0052] The strategy aggregation unit is used to perform a weighted fusion operation based on the control strategy parameters in multiple historical control cycles and the current strategy parameters to generate updated control strategy parameters, which are used to replace or optimize the corresponding strategy control parameter set in the currently used neural network control model.

[0053] A neural network-based distributed control method for air conditioning, applying any of the neural network-based distributed control systems for air conditioning described above, includes the following steps:

[0054] Step 1: Obtain environmental status information, energy consumption constraint parameters, and user comfort settings for multiple building areas as input for the current control cycle;

[0055] Step 2: Construct a multi-objective optimization model based on the thermodynamic equation of state, power boundary conditions, and user comfort range;

[0056] Step 3: Generate control task instructions based on the multi-objective optimization model, and distribute the control task instructions to the control modules corresponding to each region;

[0057] Step 4: Perform optimization decomposition processing based on the alternating direction multiplier method on the multi-objective optimization model, decompose it into multiple regional control sub-problems, and coordinate the synchronous optimization of each region;

[0058] Step 5: Each area control module receives the control task instruction, performs air conditioning equipment adjustment operations based on the environmental conditions of its area, and feeds back control error information to the coordination control module;

[0059] Step 6: Based on the building characteristic parameters and control task requirements of each region, construct a neural network control model using a hypergraph neural network structure, and update, migrate, deploy, or adaptively adjust the neural network control model during system operation;

[0060] Step 7: After the control cycle ends, evaluate the performance of the control strategy based on the control error information, and update the control strategy parameters used by the neural network control model through the strategy evaluation mechanism and the strategy aggregation mechanism.

[0061] The beneficial effects of this invention are as follows: This invention integrates environmental state information, energy consumption boundary conditions, and user comfort settings of each region within the control cycle to establish a multi-objective optimization model that includes thermodynamic state equations, power boundary conditions, and comfort ranges. A coordinated control module generates deployable and executable control task instructions, enabling targeted scheduling and distribution of control tasks. The optimization decomposition module integrated within the coordinated control module uses the alternating direction multiplier method to decompose the global optimization model into regional control sub-problems, supporting collaborative optimization of local control efficiency, thereby improving the decoupling between regions and the overall system response speed. Regional control modules deployed in each region autonomously execute control operations based on the received task instructions and feed back real-time error information to the central module, ensuring the system operates with stable state perception and feedback adjustment mechanisms. By introducing a hypergraph neural network structure through the model evolution module to construct the regional control model, the system can not only accurately characterize the building structural features and thermal coupling relationships between regions but also possess the ability to continuously optimize, migrate, and adaptively adjust the control model during actual operation, ensuring the adaptability and robustness of the control strategy to dynamic building conditions. Furthermore, the strategy self-learning module, based on control error-driven self-updating of the control strategy, integrates and updates strategy parameters after each control cycle using a strategy evaluation and aggregation mechanism. This enables the coordinated evolution of the control strategy and the neural network control model, giving the system continuous optimization and intelligent evolution capabilities over long-term operation. Overall, while maintaining control accuracy and user comfort, the system achieves distributed control, online model evolution, and strategy self-learning for building air conditioning systems. It possesses significant technical advantages such as structural integrity, logical closed-loop, high control efficiency, and flexible deployment, making it suitable for various application scenarios including smart buildings and green, low-carbon energy efficiency management. Attached Figure Description

[0062] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] in:

[0064] Figure 1 This is a modular diagram of the system of the present invention;

[0065] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0067] like Figure 1 As shown, this is an embodiment of the present invention, which provides a distributed air conditioning control system based on a neural network, comprising:

[0068] (1) Coordination and Control Module

[0069] It is used to acquire environmental status information, energy consumption constraint parameters and user comfort settings of multiple building areas within the control cycle, construct a multi-objective optimization model including thermodynamic state equation, power boundary conditions and user comfort range, generate control task instructions that match the area status and regulation constraints based on the solution results of the multi-objective optimization model, and send them to the area control module to complete the allocation and scheduling of control task instructions;

[0070] The coordination control module includes:

[0071] The state aggregation unit is used to perform dimensionality reduction mapping and discretization processing based on the state vector fed back by the region control module, and to construct a unified system state space.

[0072] The optimization modeling unit is used to construct a multi-objective optimization model of the following form based on a unified system state space, energy consumption constraint parameters, and user comfort range:

[0073]

[0074] Where: E(u) is the energy consumption function; D(u,r) is the comfort deviation function; C(u) is the control and scheduling cost function; α1, α2, α3 are weighting coefficients; u is the control command vector; r is the user expectation vector;

[0075] The control task generation unit is used to generate control task instructions for multiple regions based on the solution results of the multi-objective optimization model.

[0076] The instruction allocation and scheduling unit is used to combine the constraint status and boundary power conditions of each region to execute the task mapping and scheduling allocation of the control task instructions, form a scheduling table matched by region number and time step, and send it to the region control module.

[0077] The construction steps in this embodiment are as follows:

[0078] I. Steps and Formulas for Generating Multi-Objective Optimization Models

[0079] The coordination control module receives the standardized state vectors from each region, expressed as:

[0080] S i,t =[T in,i,t ,q i,t ,v i,t ,R t ,...];

[0081] Wherein: T in,i,t Let q be the indoor temperature of region i at time t; i,t v is the heat transfer flux of region i; i,t R represents the wind speed in region i; t Environmental or solar radiation intensity (systematic external disturbance); s i,t This is the overall system state matrix, composed of the states of all regions.

[0082] The unified processing is represented by the system state matrix S. t for:

[0083]

[0084] Construct the following multi-objective optimization function:

[0085]

[0086] Where: u i,t P represents the control decision variables (such as airflow and temperature settings) for the i-th region at time t; i For the power consumption model of the i-th region; T in,i,t T set,i,t These represent the actual room temperature and the target set temperature, respectively; α1, α2, and α3 are the target weights, configured according to preference or scenario; P i (u i,t ) is the control input u i,t The corresponding energy consumption function or power consumption function; ||u i,t -u i,t-1 || 2 It represents the magnitude of change of the control variable between adjacent time steps and is used to measure volatility;

[0087] This invention constructs a weighted multi-objective optimization function that comprehensively considers three types of objectives: minimizing control energy consumption, minimizing user comfort deviation, and smoothing control command changes. The resulting optimization objective function supports independent optimization for each region and uses weighting coefficients to achieve a balance between different objectives, adapting to the dual-objective control needs of energy saving and comfort in various building scenarios.

[0088] II. Mechanism Logic for Generating Control Task Instructions

[0089] Solving the above optimization model yields the global control vector u. t Then, each control variable is analyzed into the following structure:

[0090]

[0091] This refers to the control task instruction for the i-th region at that moment, including the target value of the control quantity and the execution timing.

[0092] III. Allocation and Scheduling of Control Task Instructions

[0093] To ensure that the task is executable within the constraints, a boundary scheduling matrix B is introduced. t =[b i,j,t ], Execute according to equipment capabilities and control task objectives:

[0094] Check constraints: whether the maximum / minimum power and response time limits are met;

[0095] Dynamic adjustment: If there is a conflict, adjust based on the minimum adjustment cost min‖Δu i,t || 2 Prioritize scheduling controllable tasks;

[0096] Output scheduling table:

[0097]

[0098] Among them: flag i,t Indicates whether the scheduling was successful (0 - failure, 1 - success, 2 - dispatched after adjustment).

[0099] The scheduling table is the set of tasks that are ultimately sent to the regional control module.

[0100] The control task generation unit in the coordination control module is defined as follows: it generates structured control task instructions based on the output results of the optimization modeling unit and the optimization decomposition module, and establishes a time series matching relationship with the control equipment in each region, so that the control task instructions have the functions of region identification, control target expression and status feedback reservation, and are adapted to the scheduling and execution of the regional control module.

[0101] In a preferred embodiment of the present invention, after the coordination control module completes the optimization modeling and variable solving in each control cycle, it constructs a structured control task instruction object through the control task generation unit, which is used to drive the control modules of each region to perform air conditioning equipment regulation operations.

[0102] The control task instructions have a structured data organization format, which is suitable for standardized interface transmission and regional module parsing. An example structure is shown in the table below:

[0103] Field Name (example) Meaning Explanation Zone_ID Building zone number where the control instruction applies Time_Slot Target execution timestamp (or time slot index) of the control instruction Actuation_Type Control action type (e.g. supply air volume adjustment, set temperature modification, etc.) Actuation_Target Specific target value (e.g. supply air volume = 500 L / s) Local_Constraint Current zone's operational constraints (e.g. maximum air volume, allowed temperature range, etc.) Priority_Level Priority level of the current task in the overall scheduling system (e.g. 0-3 levels) Execution_Flag Task execution status flag (not executed / executed successfully / failed / delayed)

[0104] The above fields can be encapsulated through structures, dictionaries, JSON objects, or other data representation methods, and distributed and managed in the system as control command objects (ControlTaskObject).

[0105] In addition, this structure supports matching with the task scheduling table of the region control module to achieve the following functions:

[0106] Tasks are distributed by region based on Zone_ID;

[0107] Alignment of the scheduling schedule with the regional control module based on Time_Slot;

[0108] Based on Priority_Level, participate in conflict resolution and resource scheduling ordering;

[0109] Feedback is written back to the coordination and control module based on Execution_Flag to support closed-loop optimization.

[0110] By using the structured instruction format described above, this invention avoids the problem of relying on fixed command interfaces in traditional control systems, and enhances the system's control flexibility and task traceability under multi-region and multi-strategy conditions.

[0111] (2) Optimize the decomposition module

[0112] Integrated into the coordination and control module, it is used to decompose the multi-objective optimization model into multiple regional control sub-problems based on the alternating direction multiplier method, and to perform control efficiency matching adjustment according to the control error mechanism;

[0113] The optimization decomposition module includes:

[0114] The regional variable decomposition unit is used to divide the global control vector in the multi-objective optimization model constructed by the coordination control module into several regional execution vectors a. j And introduce a synchronization variable g j With multiplier variable ξ j Construct a regional optimization problem of the following form:

[0115]

[0116] Where: φ j (a j ) represents the local cost function of the j-th region; a jFor the air conditioning settings (such as airflow and set temperature) of this area; g j ξ is the system synchronization target vector; j η is the multiplier estimation vector; η is the coordination factor, which adjusts the local-global coupling strength; N is the number of regions participating in parallel optimization computation in the control system;

[0117] The alternating update unit is used to perform the following steps within the control cycle:

[0118]

[0119] Feedback coupling unit, used to calculate the convergence error for each region:

[0120]

[0121] Where: k is the iteration counter in the optimization algorithm, used to identify the state of the current k-th iteration; and These are the original residual and the dual residual, respectively, reflecting the degree of deviation between the current regional control output and the synchronization reference, serving as an evaluation index for control coordination performance; and the above residuals are fed back to the coordination control module to update the coordination factor η or adjust the control task priority configuration table for the next cycle.

[0122] To differentiate itself from existing technologies that rely on a centralized global optimization function for unified solution, the optimization decomposition module of this invention introduces regional synchronous control objectives and multiplier estimation variables. Through a distributed optimization structure, it decouples the local control tasks of each region while maintaining coordination with the system's global control objective. The alternating direction multiplier method employed in this module constructs a dynamic coordination mechanism between regional and global objectives. This allows each building region to independently update its control parameters locally to the optimal level while ensuring overall objective consistency, significantly improving the parallel computing efficiency and dynamic response capability of the multi-region system.

[0123] In practical implementation, the system can dynamically adjust the coordination intensity between regions based on the error information between the control results of each region and the target reference, thereby effectively controlling the convergence rate of regional optimization and ensuring the overall control stability during system operation. Particularly noteworthy is that this invention innovatively introduces the original residual and dual residual as explicit feedback indicators of control accuracy, and transmits this feedback information back to the coordination control module. This enables dynamic ranking of cross-regional control accuracy and adjustment of control task priorities, thereby enhancing the system's robustness in complex scenarios such as load fluctuations and significant regional differences.

[0124] Furthermore, the regional variable update mechanism of this invention has good structural versatility and adaptability, supports the joint solution of various types of control parameters such as air volume, temperature setpoint, and fan on / off status, and is suitable for the joint scheduling of air conditioning equipment in building groups of different sizes. It has strong practical engineering deployability and promotion application value.

[0125] (3) Area control module

[0126] Deployed in various building areas, it is used to receive control task instructions issued by the coordination control module, and perform air conditioning equipment regulation operations in combination with the environmental status information of the area, and feed back control error information to the coordination control module;

[0127] The area control module includes:

[0128] The instruction parsing unit is used to receive and parse the control task instructions issued by the coordination and control module, and extract the area identifier field, execution time period field, control target field and constraint status field therein;

[0129] The execution control unit is used to combine the real-time environmental status information of the area with the control objective to determine whether the execution conditions are met, including but not limited to equipment availability and boundary constraint feasibility. If the conditions are met, the corresponding air conditioning equipment is driven to perform control operations.

[0130] The feedback generation unit is used to calculate control error information based on the deviation between the control result and the target after the control operation is completed, and to feed back the control error information and the control task execution status to the coordination control module.

[0131] (4) Model Evolution Module

[0132] This is used to construct a neural network control model using a hypergraph neural network structure based on the building characteristic parameters and control task requirements of each region, and to update, migrate, deploy, and adaptively adjust the neural network control model during operation.

[0133] The model evolution module includes:

[0134] The feature encoding unit is used to extract building feature parameters for each region, including building orientation, thermal resistance of the building envelope, window-to-wall ratio, volumetric heat capacity, and ventilation rate, and encode them into structural vectors of regional topological nodes.

[0135] The hypergraph generation unit is used to construct a hypergraph structure based on the above-mentioned regional node structure vectors. Each hyperedge in the hypergraph structure connects multiple building region nodes to represent the thermal coupling relationship, load linkage relationship or energy consumption similarity between regions.

[0136] The control model construction unit is used to input the region regulation task description based on the hypergraph structure, including historical control variables, target temperature sequence, and disturbance boundary, and to train the hypergraph neural network model and output the initial neural network control model parameter set required for each control cycle.

[0137] The model update unit is used to perform online parameter updates, cross-regional migration deployments, or structural adjustments on the neural network control model during system operation, based on the control error information and task completion status fed back by the regional control module, through continuous learning mechanisms including distillation migration, adaptive regression, or retraining.

[0138] The model update unit is defined as including an update trigger judgment unit, used to determine whether any of the following conditions are met within each control cycle, based on the control error information and state trajectory fed back by the area control module:

[0139] The control error exceeds the set drift threshold for multiple consecutive cycles;

[0140] There is a time lag between the control model's predictions and the actual equipment response;

[0141] The regional building thermal parameters change beyond the set boundary values;

[0142] When any condition is met, the parameter fine-tuning strategy of the neural network control model is executed, and the adaptive regression method is used to perform incremental training on the original model within a finite number of steps with the current error information as the optimization target.

[0143] The model update unit further includes:

[0144] The migration alignment module is used to select reference control models with thermal characteristic similarity higher than a set threshold from the pre-trained control model library based on the building feature parameters of the target area after the update is triggered.

[0145] Based on the current control input state sequence and the output prediction results of the reference control model, a parameter alignment training process for the target region control model is constructed, and its loss function is defined as:

[0146]

[0147] in: The control quantity predicted by the target area control model; The target value for optimized control in the current cycle; The reference control model outputs the same inputs; β∈(0,1) are the weighting coefficients of the loss function; θ TThe target region control model is a set of adjustable parameters; KL represents the Kullback-Leibler divergence; after training, the target region control model is deployed to the corresponding region control module to replace the original model and perform subsequent control tasks.

[0148] In traditional air conditioning control systems, the control model often adopts preset parameters or rule-driven methods, lacking the ability to sensitively express the heterogeneity of building spatial structures and the dynamic changes in the thermal environment, making it difficult to support the adaptive regulation requirements of complex control tasks in multi-zone building systems. To address these issues, this invention innovatively introduces a hypergraph neural network with graph structure modeling capabilities to characterize the coupling relationships between building features in different zones and the nonlinear mapping relationships between multidimensional regulation objectives, achieving generalized modeling of the control model in the spatial structure and continuous evolution in the time dimension.

[0149] The model evolution module establishes a complete model evolution chain through a four-level sub-mechanism of feature encoding, hypergraph generation, model construction, and parameter update. It can construct structural representations of spatial nodes based on complex thermal properties such as building volume, thermal resistance, window-to-wall ratio, and ventilation parameters. Furthermore, it generates a hypergraph structure reflecting regional thermal coupling characteristics by combining topological correlations, enabling the modeling and dependency representation of the propagation of thermal influences between regions. Further, by inputting historical control behavior and state feedback data, the trained neural network control model possesses cross-regional migration and deployment capabilities, adapting to task execution under different building conditions.

[0150] During model operation, the model evolution module not only supports adaptive incremental updates based on error feedback, but also performs policy transfer and parameter alignment on the reference model based on structural similarity and output behavior consistency. Specifically, this invention introduces a distillation training mechanism, using the original control policy output and the current optimization objective as joint supervision signals. This achieves a unified approach to model behavior constraints and feature representation compression during the transfer process, effectively improving the generalization performance and lightweight deployment capability of the control model, and ensuring that the system always possesses dynamic learning capabilities and regional customized control capabilities during long-term operation.

[0151] The model evolution module described in this invention has advantages such as high structural expression accuracy, flexible update mechanism, and lightweight deployment. It can adapt to various building types and operating scenarios, improve the interpretability, maintainability, and operating efficiency of the overall system control model, and embody the key evolution path of building intelligent control system from "static rule-driven" to "dynamic learning-driven".

[0152] (5) Strategy Self-Learning Module

[0153] This is used to update the control strategy based on the control error information after the control cycle ends, and to update the neural network control model through a strategy evaluation mechanism and a strategy aggregation mechanism.

[0154] The strategy self-learning module includes:

[0155] The strategy evaluation unit is used to quantitatively evaluate the performance of the current control strategy based on the control error information fed back by the regional control module after the control cycle ends. The evaluation indicators include at least control accuracy, user comfort achievement rate, equipment energy consumption deviation value and response latency.

[0156] The strategy update triggering unit is used to trigger the update operation of the control strategy when any evaluation indicator exceeds the set deviation threshold.

[0157] The strategy aggregation unit is used to perform a weighted fusion operation based on the control strategy parameters in multiple historical control cycles and the current strategy parameters to generate updated control strategy parameters, which are used to replace or optimize the corresponding strategy control parameter set in the currently used neural network control model.

[0158] like Figure 2 As shown, another embodiment of the present invention provides a distributed air conditioning control method based on a neural network, applying the distributed air conditioning control system based on a neural network as described above, including the following steps:

[0159] Step 1: Obtain environmental status information, energy consumption constraint parameters, and user comfort settings for multiple building areas as input for the current control cycle;

[0160] Step 2: Construct a multi-objective optimization model based on the thermodynamic equation of state, power boundary conditions, and user comfort range;

[0161] Step 3: Generate control task instructions based on the multi-objective optimization model, and distribute the control task instructions to the control modules corresponding to each region;

[0162] Step 4: Perform optimization decomposition processing based on the alternating direction multiplier method on the multi-objective optimization model, decompose it into multiple regional control sub-problems, and coordinate the synchronous optimization of each region;

[0163] Step 5: Each area control module receives the control task instruction, performs air conditioning equipment adjustment operations based on the environmental conditions of its area, and feeds back control error information to the coordination control module;

[0164] Step 6: Based on the building characteristic parameters and control task requirements of each region, construct a neural network control model using a hypergraph neural network structure, and update, migrate, deploy, or adaptively adjust the neural network control model during system operation;

[0165] Step 7: After the control cycle ends, evaluate the performance of the control strategy based on the control error information, and update the control strategy parameters used by the neural network control model through the strategy evaluation mechanism and the strategy aggregation mechanism.

[0166] In summary, this invention addresses key technical issues in existing air conditioning systems in multi-regional building scenarios by constructing a distributed air conditioning control system that integrates optimized modeling, task scheduling, regional execution, model evolution, and strategy self-learning. These issues include high control centralization, static models, lagging strategy updates, and missing local feedback loops. The system boasts significant advantages such as structural integrity, logical closed-loop, strong adaptability, and high scalability. It can dynamically optimize control strategies based on regional conditions and building characteristics, improving control accuracy and energy efficiency. It is suitable for energy-saving air conditioning control in smart buildings, green parks, and complex building complexes, possessing broad engineering application prospects and promotional value.

[0167] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0168] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process. Furthermore, the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functionality involved.

[0169] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A distributed air conditioning control system based on neural networks, characterized in that, include: The coordination and control module is used to acquire environmental status information, energy consumption constraint parameters and user comfort settings of multiple building areas within the control cycle, construct a multi-objective optimization model including thermodynamic state equations, power boundary conditions and user comfort ranges, generate control task instructions that match the regional status and regulation constraints based on the solution results of the multi-objective optimization model, and send them to the regional control module to complete the allocation and scheduling of control task instructions. The control task generation unit in the coordination control module is defined as follows: it generates structured control task instructions based on the output results of the optimization modeling unit and the optimization decomposition module, and establishes a time series matching relationship with the control equipment in each region, so that the control task instructions have the functions of region identification, control target expression and status feedback reservation, and are adapted to the scheduling and execution of the regional control module. The optimization decomposition module, integrated into the coordination control module, is used to decompose the multi-objective optimization model into multiple regional control sub-problems based on the alternating direction multiplier method, and to perform control efficiency matching adjustment according to the control error mechanism. The optimization decomposition module includes: The regional variable decomposition unit is used to divide the global control vector in the multi-objective optimization model constructed by the coordination control module into several regional execution vectors. And introduce synchronization variables With multiplier variables The region optimization problem can be constructed in the following form: ； in: Indicates the first Local cost function for each region; This refers to the air conditioning usage in the area. The target vector for system synchronization; The multiplier estimation vector; As a coordination factor, it adjusts the local-global coupling strength; To control the number of regions participating in parallel optimization computation in the system; The alternating update unit is used to perform the following steps within the control cycle, as shown in the formula: ； The feedback coupling unit is used to calculate the convergence error for each region, and the formula is: ； in: This is an iteration counter used in the optimization algorithm to identify the current iteration. The state of the next iteration; and These are the original residual and the dual residual, respectively, reflecting the degree of deviation between the current regional control output and the synchronous reference, serving as evaluation indicators of control coordination performance; the residuals are then fed back to the coordinated control module to update the coordination factor. Or adjust the control task priority configuration table for the next cycle; The area control module is deployed in each building area to receive control task instructions issued by the coordination control module, and to perform air conditioning equipment regulation operations in combination with the environmental status information of the area, and to feed back control error information to the coordination control module. The model evolution module is used to construct a neural network control model using a hypergraph neural network structure based on the building characteristic parameters and control task requirements of each region, and to update, migrate, deploy, and adaptively adjust the neural network control model during operation. The model update unit is defined as including an update trigger judgment unit, used to determine whether any of the following conditions are met within each control cycle, based on the control error information and state trajectory fed back by the area control module: The control error exceeds the set drift threshold for multiple consecutive cycles; There is a time lag between the control model's predictions and the actual equipment response; The regional building thermal parameters change beyond the set boundary values; And when any condition is met, the parameter fine-tuning strategy of the neural network control model is executed, and the adaptive regression method is used to perform incremental training on the original model within a finite number of steps with the current error information as the optimization target. The model update unit also includes: The migration alignment module is used to select reference control models with thermal characteristic similarity higher than a set threshold from the pre-trained control model library based on the building feature parameters of the target area after the update is triggered. Based on the current control input state sequence and the output prediction results of the reference control model, a parameter alignment training process for the target region control model is constructed, and the loss function is defined as: ； in: The control quantity predicted by the target area control model; The target value for optimized control in the current cycle; The output of the reference control model for the same input; These are the weighting coefficients for the loss function; The adjustable parameter set for the target area control model; This represents the Kullback-Leibler divergence; after training, the target region control model is deployed to the corresponding region control module to replace the original model and perform subsequent regulation tasks. The strategy self-learning module is used to update the control strategy based on the control error information after the control cycle ends, and to update the neural network control model through the strategy evaluation mechanism and the strategy aggregation mechanism.

2. The distributed air conditioning control system based on a neural network according to claim 1, characterized in that, The coordination control module includes: The state aggregation unit is used to perform dimensionality reduction mapping and discretization processing based on the state vector fed back by the region control module, and to construct a unified system state space. The optimization modeling unit is used to construct a multi-objective optimization model of the following form based on a unified system state space, energy consumption constraint parameters, and user comfort range: ； in: It is an energy consumption function; This is the comfort deviation function; To control the scheduling cost function; , , These are weighting coefficients; For control command vectors; A vector of user expected values; The control task generation unit is used to generate control task instructions for multiple regions based on the solution results of the multi-objective optimization model. The instruction allocation and scheduling unit is used to combine the constraint status and boundary power conditions of each region to execute the task mapping and scheduling allocation of the control task instructions, form a scheduling table matched by region number and time step, and send it to the region control module.

3. The distributed air conditioning control system based on a neural network according to claim 1, characterized in that, The area control module includes: The instruction parsing unit is used to receive and parse the control task instructions issued by the coordination and control module, and extract the area identifier field, execution time period field, control target field and constraint status field therein; The execution control unit is used to combine the real-time environmental status information of the area with the control objective to determine whether the execution conditions are met, including but not limited to equipment availability and boundary constraint feasibility. If the conditions are met, the corresponding air conditioning equipment is driven to perform control operations. The feedback generation unit is used to calculate control error information based on the deviation between the control result and the target after the control operation is completed, and to feed back the control error information and the control task execution status to the coordination control module.

4. The distributed air conditioning control system based on a neural network according to claim 1, characterized in that, The model evolution module includes: The feature encoding unit is used to extract building feature parameters for each region, including building orientation, thermal resistance of the building envelope, window-to-wall ratio, volumetric heat capacity, and ventilation rate, and encode them into structural vectors of regional topological nodes. The hypergraph generation unit is used to construct a hypergraph structure based on the above-mentioned regional node structure vectors. Each hyperedge in the hypergraph structure connects multiple building region nodes to represent the thermal coupling relationship, load linkage relationship or energy consumption similarity between regions. The control model construction unit is used to input the region regulation task description based on the hypergraph structure, including historical control variables, target temperature sequence, and disturbance boundary, and to train the hypergraph neural network model and output the initial neural network control model parameter set required for each control cycle. The model update unit is used to perform online parameter updates, cross-regional migration deployments, or structural adjustments on the neural network control model during system operation, based on the control error information and task completion status fed back by the regional control module, through continuous learning mechanisms including distillation migration, adaptive regression, or retraining.

5. The distributed air conditioning control system based on a neural network according to claim 1, characterized in that, The strategy self-learning module includes: The strategy evaluation unit is used to quantitatively evaluate the performance of the current control strategy based on the control error information fed back by the regional control module after the control cycle ends. The evaluation indicators include at least control accuracy, user comfort achievement rate, equipment energy consumption deviation value and response latency. The strategy update triggering unit is used to trigger the update operation of the control strategy when any evaluation indicator exceeds the set deviation threshold. The strategy aggregation unit is used to perform a weighted fusion operation based on the control strategy parameters in multiple historical control cycles and the current strategy parameters to generate updated control strategy parameters, which are used to replace or optimize the corresponding strategy control parameter set in the currently used neural network control model.

6. A distributed control method for air conditioning based on neural networks, employing a distributed control system for air conditioning based on neural networks as described in any one of claims 1-5, characterized in that, Includes the following steps: Step 1: Obtain environmental status information, energy consumption constraint parameters, and user comfort settings for multiple building areas as input for the current control cycle; Step 2: Construct a multi-objective optimization model based on the thermodynamic equation of state, power boundary conditions, and user comfort range; Step 3: Generate control task instructions based on the multi-objective optimization model, and distribute the control task instructions to the control modules corresponding to each region; Step 4: Perform optimization decomposition processing based on the alternating direction multiplier method on the multi-objective optimization model, decompose it into multiple regional control sub-problems, and coordinate the synchronous optimization of each region; Step 5: Each area control module receives the control task instruction, performs air conditioning equipment adjustment operations based on the environmental conditions of its area, and feeds back control error information to the coordination control module; Step 6: Based on the building characteristic parameters and control task requirements of each region, construct a neural network control model using a hypergraph neural network structure, and update, migrate, deploy, or adaptively adjust the neural network control model during system operation; Step 7: After the control cycle ends, evaluate the performance of the control strategy based on the control error information, and update the control strategy parameters used by the neural network control model through the strategy evaluation mechanism and the strategy aggregation mechanism.

Images