Clustered temperature controller distributed collaborative control method and system

Abstract

The application discloses a cluster type temperature controller distributed collaborative control method and system, and relates to the technical field of temperature controller control. The method comprises the following steps: establishing temperature control consensus, constructing a temperature controller alliance chain, and achieving temperature control target consensus by using a practical Byzantine fault tolerance algorithm; sharing a prediction model, training a temperature prediction model locally based on a federal learning architecture, and uploading model parameter gradients to a cloud server for aggregation and updating; globally optimizing energy consumption, minimizing and optimizing global energy consumption by using a multi-agent deep deterministic policy gradient algorithm; and dynamically reorganizing a topology, supporting plug and play of equipment, and automatically reorganizing a cluster topology according to equipment states. The application solves the information island problem, realizes sharing and privacy protection of the temperature prediction model, and optimizes global energy consumption, so that the cluster temperature control consistency can be improved, energy utilization efficiency can be optimized, user comfort experience can be improved, hundreds of equipment extensions can be supported, and complex scene requirements can be met.

Description

Technical Field

[0001] This invention belongs to the field of thermostat control technology, and particularly relates to a distributed collaborative control method and system for clustered thermostats. Background Technology

[0002] As a control device for temperature regulation in temperature-controlled environments, a thermostat not only samples the temperature but also incorporates a built-in relay to directly control the operation of the fan coil unit for cooling or the water valve for heating. Therefore, users can control the temperature within their regulated environments using a thermostat.

[0003] However, traditional standalone thermostats suffer from information silos, and when multiple devices work together, control command conflicts can easily arise (such as conflicting temperature control targets in adjacent areas), leading to increased energy consumption and decreased comfort. This limitation makes it difficult for the system to achieve ideal overall energy efficiency and user experience in complex scenarios.

[0004] Therefore, there is an urgent need to propose a distributed collaborative control method and system for clustered thermostats. Through blockchain consensus mechanism, federated learning architecture, collaborative optimization algorithm and dynamic topology management, the system can improve the consistency rate of clustered temperature control and the overall energy efficiency of the system, support large-scale device expansion, thereby optimizing energy utilization efficiency and improving user comfort experience. Summary of the Invention

[0005] To address the aforementioned problems in existing technologies, this invention provides a distributed collaborative control method and system for clustered temperature controllers. By introducing a blockchain consensus mechanism, federated learning architecture, collaborative optimization algorithm, and dynamic topology management, it solves the information silo problem of traditional single-unit temperature controllers and the control command conflict problem when multiple devices collaborate, improving the consistency of clustered temperature control and system energy efficiency, and supporting the expansion of hundreds of devices.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] In a first aspect, embodiments of this application provide a distributed collaborative control method for clustered temperature controllers, comprising the following steps:

[0008] Establishing a consensus on temperature control: By building a temperature controller consortium blockchain, a practical Byzantine fault-tolerant algorithm is used to achieve a consensus on the temperature control target;

[0009] Shared prediction model: Based on the federated learning module, each thermostat node trains a temperature prediction model locally and uploads the model parameter gradients to the cloud server for aggregation and updating;

[0010] Global energy consumption optimization: A multi-agent deep deterministic policy gradient algorithm is used to minimize global energy consumption.

[0011] Dynamic topology reconfiguration: Supports plug-and-play devices via ZigBee Mesh networking and automatically reconfigures cluster topology based on device status.

[0012] Preferably, establishing a consensus on temperature control includes the following steps:

[0013] Building a consortium blockchain: Add all thermostat nodes to the blockchain network, with each thermostat node acting as a participant in the blockchain, recording temperature control targets and operating status;

[0014] Consensus algorithm execution: The practical Byzantine fault-tolerant algorithm is used to reach a consensus on the temperature control target among the temperature controller nodes, so that the temperature control target of adjacent areas is consistent;

[0015] The specific execution flow of the practical Byzantine fault-tolerant algorithm includes a pre-preparation phase, a preparation phase, and a submission phase. Each phase requires confirmation from more than two-thirds of the nodes before proceeding to the next phase.

[0016] Preferably, establishing a consensus on temperature control further includes the following steps:

[0017] Initialize nodes: Assign a unique identifier to each thermostat node and register it with the blockchain network;

[0018] Synchronous operation status: The operation status and temperature control target of each thermostat node are recorded through the blockchain ledger.

[0019] Preferably, the shared prediction model includes the following steps:

[0020] Local model training: Each thermostat node collects environmental data and user behavior data locally to train a temperature prediction model;

[0021] Gradient aggregation update: The parameter gradient of the temperature prediction model is uploaded to the cloud server. The cloud server generates a global model by weighted averaging of the parameter gradients and then distributes the updated model parameters to each temperature controller node.

[0022] Model Iterative Optimization: After receiving the model parameters, each temperature controller node continues to train locally and uploads new gradients until the model converges.

[0023] Preferably, the shared prediction model further includes the following steps:

[0024] Data privacy protection: The gradient data is encrypted before uploading the parameter gradient;

[0025] Anomaly detection: By comparing the parameter gradients uploaded by each temperature controller node with the deviation of the global model, the system detects the existence of abnormal nodes and removes abnormal data.

[0026] Preferably, the global energy consumption optimization includes the following steps:

[0027] Define the optimization objective: take the minimization of global energy consumption as the objective function, and construct the optimization problem by combining the operating status and environmental conditions of each temperature controller node;

[0028] Multi-agent collaboration: A multi-agent deep deterministic policy gradient algorithm is adopted, treating each thermostat node as an agent, and global optimization is performed through collaboration among agents;

[0029] Policy update: Each agent selects an action based on the current state and adjusts its policy through environmental feedback to gradually obtain the optimal solution.

[0030] Preferably, the global energy consumption optimization further includes the following steps:

[0031] Environmental perception: Real-time collection of ambient temperature and humidity data through sensors serves as input for optimization problems;

[0032] User preference integration: Transform users' comfort needs into constraints to ensure that the optimization results meet user expectations.

[0033] Preferably, the dynamic topology reorganization includes the following steps:

[0034] Device access detection: Monitors the access and disconnection status of devices in real time through the ZigBee Mesh network;

[0035] Topology adjustment: When a new device is connected or an old device is disconnected, the network automatically adjusts its topology to maintain the connectivity of communication links;

[0036] Load balancing: Dynamically allocate tasks based on the operating load of the equipment.

[0037] Preferably, the dynamic topology reorganization further includes the following steps:

[0038] Signal strength assessment: The communication quality between devices is evaluated using RSSI values, and links with high signal strength are prioritized.

[0039] Fault recovery: When a link is interrupted, the network automatically switches to a backup link.

[0040] Secondly, this application provides a distributed collaborative control system for clustered thermostats, applied to the distributed collaborative control method for clustered thermostats as described above, including a blockchain module, a federated learning module, an optimization calculation module, and a network management module that are connected in sequence.

[0041] The blockchain module is used to reach a consensus on the temperature control target through a practical Byzantine fault-tolerant algorithm, and to record the operating status and temperature control target of each temperature controller node.

[0042] The federated learning module is used for each thermostat node to train a temperature prediction model locally and to aggregate and update the model parameters through a cloud server.

[0043] The optimization calculation module is used to minimize global energy consumption using a multi-agent deep deterministic policy gradient algorithm.

[0044] The network management module is used to support plug-and-play devices through the ZigBee Mesh network and automatically reorganize the cluster topology based on the device status.

[0045] The beneficial effects of this invention are as follows:

[0046] This invention solves the information silo problem of traditional single-unit temperature controllers by using a blockchain consensus mechanism, ensuring consistent temperature control targets in adjacent areas and avoiding control command conflicts. A federated learning architecture enables the sharing and optimization of temperature prediction models among nodes, improving prediction accuracy while protecting user privacy. A collaborative optimization algorithm minimizes global energy consumption, significantly improving overall system energy efficiency. Dynamic topology management supports plug-and-play devices and automatically reorganizes the network topology, enhancing system scalability and robustness. Through the comprehensive application of these technologies, this invention improves the consistency rate of cluster temperature control, optimizes energy utilization efficiency, enhances user comfort, and supports expansion to hundreds of devices, adapting to diverse needs in complex scenarios. Attached Figure Description

[0047] To better understand and implement this application, the technical solution is described in detail below with reference to the accompanying drawings.

[0048] Figure 1 A flowchart illustrating the steps of the distributed collaborative control method for a clustered temperature controller provided in this application embodiment;

[0049] Figure 2 A flowchart illustrating the structure of the federated learning module provided in this application embodiment;

[0050] Figure 3 A flowchart illustrating the steps of the optimization calculation module in this application embodiment to implement global energy consumption optimization;

[0051] Figure 4 This is a schematic diagram of the structure of the clustered temperature controller distributed collaborative control system provided in the embodiments of this application. Detailed Implementation

[0052] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, exemplary embodiments will be described in detail below, examples of which are illustrated in the accompanying drawings. In the following description relating to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of methods and systems consistent with some aspects of this application as detailed in the appended claims.

[0053] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to any and all possible combinations comprising one or more of the associated listed items.

[0054] The following detailed description of the specific implementation methods, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided in detail.

[0055] Example 1

[0056] Please see Figure 1 This application provides a distributed collaborative control method for clustered temperature controllers, comprising the following steps:

[0057] Establishing a consensus on temperature control: By building a temperature controller consortium blockchain, a practical Byzantine fault-tolerant algorithm is used to achieve a consensus on the temperature control target;

[0058] Shared prediction model: Based on the federated learning module, each thermostat node trains a temperature prediction model locally and uploads the model parameter gradients to the cloud server for aggregation and updating;

[0059] Global energy consumption optimization: A multi-agent deep deterministic policy gradient algorithm is used to minimize global energy consumption.

[0060] Dynamic topology reconfiguration: Supports plug-and-play devices via ZigBee Mesh networking and automatically reconfigures cluster topology based on device status.

[0061] Specifically, the aforementioned distributed collaborative control method for clustered thermostats comprises four main steps: establishing temperature control consensus, sharing prediction models, global energy consumption optimization, and dynamic topology reorganization. These steps are implemented through sequential communication connections between the blockchain module, federated learning module, optimization computing module, and network management module. Specifically, the thermostat node, as the core component, works collaboratively with other modules. Each thermostat node is equipped with a processing unit, a storage unit, and a temperature control unit to perform local computation, data storage, and temperature regulation functions. Thermostat nodes communicate with each other via a ZigBee Mesh network to ensure efficient information transmission.

[0062] In the temperature control consensus establishment phase, the nodes are first initialized: all temperature controller nodes are initialized and registered in the blockchain network, and each temperature controller node is assigned a unique identity as a participant in the blockchain. This process is completed through the blockchain module, which records the operating status and temperature control target of each temperature controller node, ensuring information transparency and immutability. The temperature control target consensus is reached among the temperature controller nodes through the PBFT (Practical Byzantine Fault Tolerance) algorithm, that is: the blockchain ledger records the operating status and temperature control target of each temperature controller node, ensuring that the temperature control target is consistent in adjacent areas; the PBFT consensus process includes a pre-preparation phase, a preparation phase, and a commit phase. In the pre-preparation phase, the master node broadcasts the temperature control target proposal to other nodes, the proposal including the temperature control target value and the current operating status. After entering the preparation phase, all nodes verify and vote on the proposal; only after more than 2 / 3 of the nodes confirm can the commit phase begin. In the commit phase, all nodes confirm the proposal again, finally reaching a consensus on the temperature control target. Through this process, temperature controller nodes in adjacent areas can maintain a consistent temperature control target, avoiding control command conflicts.

[0063] The next stage is the shared prediction model phase, such as... Figure 2The diagram illustrates the architecture of the federated learning module, including a local model training component, a cloud aggregation server, and a global model component. Each thermostat node collects environmental and user behavior data, such as indoor and outdoor temperatures, humidity, and user-set temperatures, through its local model training module to train a temperature prediction model. After training, the gradient parameters of the local model are uploaded to the cloud aggregation server. The cloud aggregation server performs a weighted average of the received gradient data to generate a global model and distributes the updated model parameters to each thermostat node. After receiving the global model parameters, each thermostat node continues training locally and uploads new gradients until the model converges. To protect user privacy, the data is encrypted before uploading gradients to ensure that user privacy is not leaked. Simultaneously, the deviations between the gradients uploaded by each thermostat node and the global model are compared to detect abnormal nodes and remove abnormal data. This anomaly detection mechanism removes abnormal data that deviates from the global model. Through this process, each thermostat node can share the optimized temperature prediction model, improving prediction accuracy.

[0064] The next stage is the global energy consumption optimization phase, implemented through an optimization calculation module. This module uses minimizing global energy consumption as the objective function, constructing an optimization problem based on the operating status and environmental conditions of each thermostat node. The module employs a Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm, treating each thermostat node as an agent. Each agent selects actions based on its current state and adjusts its strategy through environmental feedback, gradually approaching the optimal solution. During this process, sensors collect real-time data on ambient temperature and humidity as input to the optimization problem, while user comfort requirements are transformed into constraints to ensure the optimization result meets user expectations. To improve training efficiency, the optimization calculation module uses an experience replay mechanism to store historical interaction data and accelerates model convergence through batch sampling. Through this process, the system can minimize global energy consumption while meeting user comfort requirements. Figure 3 As shown, the process of global energy consumption optimization implemented by the optimization calculation module is as follows: First, the operating status and environmental conditions of each thermostat node are initialized; then, the optimization problem is constructed: the objective function of minimizing global energy consumption is constructed; next, the optimization algorithm is called to solve the optimal control strategy; then, the control parameters of each thermostat node are updated, and then it is judged whether the optimization result meets the convergence condition. If it converges, the optimization ends; otherwise, it returns to update the environmental conditions and operating status.

[0065] Finally, the dynamic topology reconfiguration phase is implemented through the network management module. The network management module supports plug-and-play devices via the ZigBeeMesh network and automatically reconfigures the cluster topology based on device status. Specifically, the network management module includes a signal strength assessment unit, a topology adjustment unit, and a fault recovery link. The operational mechanism of the network management module, including device access detection, topology adjustment, and fault recovery, follows this process:

[0066] First, the system detects incoming devices to determine whether a new device is connecting or an old device is disconnecting. Then, the signal strength assessment unit evaluates the communication quality between devices using RSSI values ​​and prioritizes links with higher signal strength. When a new device connects or an old device disconnects, the topology adjustment unit automatically adjusts the network topology to ensure communication link connectivity. Furthermore, the network management module has a load balancing function, dynamically allocating tasks based on device workload to prevent overload of any single node. The system also checks for communication link interruptions. If a link is interrupted, the fault recovery link automatically switches to a backup link to ensure communication continuity. If the communication link is not interrupted, the existing topology is maintained. Through this process, the system can adapt to changes in the number of devices and maintain stable communication performance.

[0067] In the aforementioned stages, the collaborative relationships between the modules are as follows: The blockchain module is responsible for recording the operating status and temperature control targets of each thermostat node, providing a data foundation for subsequent stages; the federated learning module achieves model parameter sharing and optimization through a cloud aggregation server, improving predictive capabilities; the optimization computation module performs energy consumption optimization based on the global model provided by the federated learning module, ensuring maximum system energy efficiency; the network management module supports plug-and-play devices through a ZigBee Mesh network and automatically adjusts the topology according to device status, providing stable communication guarantees for the entire system. Communication between modules is achieved through standard interface protocols, ensuring efficient and reliable data transmission.

[0068] Furthermore, to further enhance the system's scalability and robustness, this application can be optimized for specific scenarios in practical applications. For example, in commercial building scenarios, the system scale can be expanded by increasing the number of thermostat nodes, while utilizing dynamic topology reconfiguration to handle frequent device connection and disconnection. In home scenarios, the computing module can be optimized to integrate user-specific preferences, providing more precise temperature control services. Regardless of the scenario, this invention can achieve consistency in cluster temperature control and improve system energy efficiency through the aforementioned technical means.

[0069] The specific embodiments of the present invention have been described in detail above, covering the complete process from hardware composition to software operation. The connection relationships, positional relationships, and mutual cooperation relationships between the various components have been explained with reference to the accompanying drawings. Through the above embodiments, those skilled in the art can clearly understand and implement the technical solution of the present invention.

[0070] To enable those skilled in the art to fully understand and implement this invention, the specific implementation principle of this invention will be further explained below in conjunction with a specific application scenario.

[0071] In a commercial building scenario, suppose an office building has multiple thermostat nodes installed, distributed across different floors and areas, used to control the cooling and heating equipment of the air conditioning system. The clustered thermostat distributed collaborative control method and system provided by this invention can achieve efficient management of the temperature throughout the building.

[0072] First, during the temperature control consensus establishment phase, all temperature controller nodes are initialized and registered in the blockchain network, with each node assigned a unique identity. The PBFT consensus process then begins, with the master node broadcasting the proposed temperature control target for the current area to other temperature controller nodes, for example, setting it to 24℃. In the pre-preparation phase, each node receives the proposal, verifies its rationality, and provides confirmation. Once in the preparation phase, if more than two-thirds of the nodes vote in favor of the proposal, it enters the submission phase. Finally, all nodes reconfirm the proposal content, reaching a consensus on the temperature control target. This process ensures that temperature controller nodes in adjacent areas do not waste energy due to target conflicts, improving overall energy efficiency.

[0073] Next, in the shared prediction model phase, each thermostat node collects environmental data, such as indoor and outdoor temperature and humidity, through its local model training module, and combines this data with user behavior data (e.g., temperature preferences over a specific time period), to train its local temperature prediction model. After training, the parameter gradients of the local model are encrypted and uploaded to the cloud aggregation server. The cloud server performs a weighted average of the received gradient data to generate a global model and distributes the updated model parameters to each thermostat node. Each node receives the global model and continues to iterate and optimize it locally. In this way, each node can share the optimized prediction model, thereby improving the accuracy of temperature prediction and reducing energy waste caused by prediction errors.

[0074] Subsequently, in the global energy consumption optimization phase, the optimization calculation module constructs an optimization problem with minimizing global energy consumption as the objective function. The module employs the MADDPG algorithm, treating each thermostat node as an agent. For example, when a sensor in a certain area detects a current temperature of 26℃, and the user-set target temperature is 24℃, the agent selects an action based on the current state (such as adjusting the cooling equipment's operating power) and adjusts its strategy based on environmental feedback. During this process, the user's comfort requirements are transformed into constraints, such as allowing temperature fluctuations within ±1℃. The optimization calculation module also utilizes an experience replay mechanism to store historical interaction data and accelerates model convergence through batch sampling. Through this mechanism, the system can minimize global energy consumption while meeting user comfort requirements.

[0075] Finally, during the dynamic topology reconfiguration phase, the network management module monitors the connection and disconnection status of devices in real time through the ZigBee Mesh network. The signal strength assessment unit evaluates the communication quality between devices using RSSI values ​​and prioritizes links with high signal strength. For example, when a thermostat node is added to a floor, the topology adjustment unit automatically adjusts the network topology to ensure the new node can connect smoothly and maintain communication link connectivity. Simultaneously, the network management module has load balancing capabilities, dynamically allocating tasks based on device workload to prevent overload of any single node. If a link is interrupted due to a fault, the fault recovery link automatically switches to the backup link to ensure communication continuity. This dynamic topology reconfiguration mechanism allows the system to adapt to changes in the number of devices while maintaining stable communication performance.

[0076] Throughout these stages, the modules collaborate closely. For example, the temperature control targets and operational status recorded by the blockchain module provide the foundational data for the federated learning module; the global model generated by the federated learning module provides accurate predictive capabilities for the energy consumption optimization of the optimization computing module; and the network management module ensures efficient collaboration between the modules through stable communication guarantees. Furthermore, the system supports scalability optimization; for instance, in a home setting, the optimization computing module can integrate users' personalized preferences to provide more precise temperature control services.

[0077] The foregoing has described in detail the operating principle and implementation steps of the present invention in specific application scenarios, covering the complete process from hardware composition to software operation. The connection relationships, positional relationships, and mutual cooperation relationships between the various components have been explained with reference to the accompanying drawings. Through the above embodiments, those skilled in the art can clearly understand and implement the technical solution of the present invention.

[0078] Example 2

[0079] Please see Figure 4This application provides a clustered temperature controller distributed collaborative control system, including a blockchain module, a federated learning module, an optimization computing module and a network management module that are connected in sequence.

[0080] The blockchain module is used to reach a consensus on the temperature control target through a practical Byzantine fault-tolerant algorithm, and to record the operating status and temperature control target of each temperature controller node.

[0081] The federated learning module is used for each thermostat node to train a temperature prediction model locally and to aggregate and update the model parameters through a cloud server.

[0082] The optimization calculation module is used to minimize global energy consumption using a multi-agent deep deterministic policy gradient algorithm.

[0083] The network management module is used to support plug-and-play devices through the ZigBee Mesh network and automatically reorganize the cluster topology based on the device status.

[0084] The data management strategy of the blockchain module includes: initial synchronization phase: full synchronization of historical data of each node to ensure ledger consistency; incremental update phase: only synchronization of newly added transaction data to reduce communication overhead.

[0085] As a preferred embodiment of the present invention, the federated learning module is used to protect user data privacy by employing differential privacy technology when training the model locally; at the same time, it manages the version of the global model through a cloud server to ensure that each node always uses the latest model.

[0086] The optimization calculation module is used to store historical interaction data using an experience replay mechanism and improve training efficiency through batch sampling when executing the MADDPG algorithm.

[0087] The network management module is used to optimize communication paths and reduce latency by employing distributed routing algorithms during network topology reconfiguration.

[0088] Example 3

[0089] This application embodiment also provides a temperature control device, including a temperature control unit, a storage unit, a processing unit, and a temperature control device control program stored in the storage unit and executable on the processing unit, wherein:

[0090] The temperature control unit is used to regulate the on / off state of the refrigeration or heating circuit;

[0091] When the temperature control device control program is executed by the processing unit, it implements the steps of the clustered temperature controller distributed collaborative control method as described above.

[0092] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0093] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0094] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0095] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A cluster temperature controller distributed collaborative control method, characterized in that: The method comprises the following steps: Establish temperature control consensus: by constructing a temperature controller alliance chain, a practical Byzantine fault tolerance algorithm is used to reach a temperature control target consensus; Share prediction model: based on a federal learning module, each temperature controller node locally trains a temperature prediction model, and uploads the model parameter gradient to a cloud server for aggregation and updating; Global energy consumption optimization: a multi-agent deep deterministic policy gradient algorithm is used to minimize and optimize global energy consumption; Dynamic topology reorganization: ZigBee Mesh network supports plug and play, and automatically reorganizes the cluster topology according to the device state; The establishment of the temperature control consensus comprises the following steps: Constructing an alliance chain: all temperature controller nodes are added to the blockchain network, and each temperature controller node is a participant in the blockchain, recording the temperature control target and the running state; Consensus algorithm execution: the practical Byzantine fault tolerance algorithm is used to reach a temperature control target consensus among temperature controller nodes, so that the temperature control targets of adjacent areas are consistent; The specific execution process of the practical Byzantine fault tolerance algorithm includes a pre-preparation stage, a preparation stage and a submission stage, and each stage must be confirmed by more than two-thirds of the nodes before entering the next stage.

2. The cluster thermostat distributed collaborative control method of claim 1, wherein: The establishment of the temperature control consensus further comprises the following steps: Initialization of nodes: each temperature controller node is assigned a unique identity and registered in the blockchain network; Synchronization of running state: the running state and temperature control target of each temperature controller node are recorded in the blockchain ledger.

3. The cluster thermostat distributed collaborative control method of claim 1, wherein: The shared prediction model comprises the following steps: Local model training: each temperature controller node locally collects environmental data and user behavior data to train a temperature prediction model; Gradient aggregation and updating: the parameter gradient of the temperature prediction model is uploaded to the cloud server, the cloud server performs weighted averaging on the parameter gradient to generate a global model, and the updated model parameters are distributed to each temperature controller node; Model iteration optimization: after receiving the model parameters, each temperature controller node continues to train locally and uploads new gradients until the model converges.

4. The cluster thermostat distributed collaborative control method of claim 3, wherein: The shared prediction model further comprises the following steps: Data privacy protection: the gradient data is encrypted before uploading the parameter gradient; Abnormality detection: by comparing the parameter gradient uploaded by each temperature controller node with the deviation of the global model, it is detected whether there is an abnormal node, and the abnormal data is excluded.

5. The cluster thermostat distributed collaborative control method of claim 1, wherein: The global energy consumption optimization comprises the following steps: Defining the optimization target: taking the minimization of global energy consumption as the objective function, combining the running state of each temperature controller node and the environmental conditions to construct an optimization problem; Multi-agent collaboration: using a multi-agent deep deterministic policy gradient algorithm, each temperature controller node is regarded as an agent, and global optimization is performed through collaboration between agents; Strategy update: each agent selects an action according to the current state, and adjusts the strategy through environmental feedback to gradually obtain the optimal solution.

6. The cluster thermostat distributed collaborative control method of claim 5, wherein: The global energy consumption optimization further comprises the following steps: Environmental perception: real-time collection of environmental temperature and humidity data through sensors as input conditions for the optimization problem; Integration of user preferences: the user's comfort requirements are converted into constraint conditions, so that the optimization result meets the user's expectations.

7. The cluster thermostat distributed collaborative control method of claim 1, wherein: The dynamic topology reorganization comprises the following steps: Device access detection: Real-time monitoring of device access and disconnection status through ZigBee Mesh network; Topology adjustment: When new devices are connected or old devices are disconnected, the network automatically adjusts the topology to maintain the connectivity of the communication link; Load balancing: Dynamically allocate tasks according to the running load of the device.

8. The cluster thermostat distributed collaborative control method of claim 7, wherein: The dynamic topology reorganization further comprises the following steps: Signal strength evaluation: Evaluate the communication quality between devices through RSSI value and prefer the link with high signal strength; Fault recovery: When a link is interrupted, the network automatically switches to the backup link.

9. A distributed collaborative control system for cluster thermostats, applying the distributed collaborative control method for cluster thermostats according to any of claims 1-8, characterized in that: It includes a blockchain module, a federated learning module, an optimization calculation module and a network management module connected in sequence; The blockchain module is used to reach the temperature control target consensus through the practical Byzantine fault tolerance algorithm, and record the running state and temperature control target of each temperature controller node; The federated learning module is used for each temperature controller node to train a temperature prediction model locally, and update the model parameters through a cloud server; The optimization calculation module is used to minimize the global energy consumption by using the multi-agent deep deterministic policy gradient algorithm; The network management module is used to support plug-and-play through ZigBee Mesh network, and automatically reorganize the cluster topology according to the device state.

Images