Building cluster cloud-side collaborative optimization control method and device and application architecture

Abstract

The invention discloses a building cluster cloud-edge collaborative optimization control method and device and an application architecture. The architecture comprises an edge device, a radio frequency communication LoRa module and a cloud platform. The edge device and the cloud platform realize cloud edge two-way communication through the LoRa module; the edge device comprises a sensing layer, an edge calculation core and a control layer and is used for collecting real-time building data, executing a local model predictive control algorithm MPC based on a building thermodynamics RC model and generating an optimal control instruction to drive air conditioning equipment; the cloud platform comprises a network service interface module, an optimization algorithm module and a data storage system and is used for operating a global coordination algorithm based on an alternating direction multiplier method (ADMM) and generating differentiated price signals; the edge device uploads data such as power requirements to the cloud platform, the cloud platform issues differential excitation signals to the edge device, and the edge device and the cloud platform perform collaborative iteration. According to the invention, distributed optimization control of the building cluster is realized through cloud-side cooperation, and the value of the building cluster as a power grid flexible resource is significantly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wireless communication system simulation and channel modeling, in particular to a building cluster cloud-edge collaborative optimization control method, device and application architecture. BACKGROUND

[0002] With the acceleration of new power system construction, buildings, as core distributed resources on the user side, have gradually become an important participant in grid coordination and interaction due to their large energy consumption base and flexible regulation characteristics. To solve the problems of transaction volume increase, low transaction efficiency and high operation and maintenance cost caused by the direct participation of a large number of dispersed building users in the electricity market, the building aggregator model has emerged. By integrating the regulation capabilities of multiple buildings, the model enables the scale of the subject to participate in electricity market transactions and grid regulation services, while optimizing the distribution of interests among itself, building users and power companies, and has both economic benefits and service value. Under this background, building energy management technology is developing rapidly. For example, in China, building energy efficiency has been improving year by year, and technologies such as energy consumption monitoring and single building energy-saving control in the field of intelligent buildings have gradually been implemented. Abroad, a general evaluation framework for building energy flexibility has been established, and theoretical research on distributed optimization algorithms in the field of multi-building coordination and microgrid scheduling has been continuously deepened, laying a technical foundation for the coordination of building clusters and power grids.

[0003] However, the existing technology still has significant limitations in practical application, making it difficult to meet the needs of efficient operation of building aggregators and grid coordination. On the one hand, existing building energy management systems mostly focus on the energy efficiency optimization of single buildings, lacking overall consideration of the coordination and interaction between building clusters and power grids. Even though some studies involve multi-building coordination, they do not fully integrate electricity market trading rules and distribution network operation constraints, making it difficult for building flexible resources to be efficiently integrated into grid scheduling. On the other hand, there are bottlenecks in core technologies: algorithm level. Most distributed optimization and precise control algorithms are still at the theoretical or simulation stage, and have not been verified in real building cluster scenarios for low-bandwidth adaptability and stable convergence. Moreover, it is difficult to balance user privacy protection and data interaction needs. At the hardware level, there is a lack of integrated devices deeply coupled with algorithms. Existing edge devices mostly only have data acquisition functions and cannot implement local algorithm solving and precise control, resulting in insufficient coordination between "algorithm-hardware-scenario" and restricting the scale-up and operational efficiency improvement of building aggregators.

[0004] Therefore, there is an urgent need for a building cluster cloud-edge collaborative optimization control method to solve the above problems, enabling building flexible resources to be efficiently integrated into grid scheduling, realizing local algorithm solving and precise control, and enabling building aggregators to be scaled up and operational efficiency to be improved. SUMMARY

[0005] To this end, the application provides a building cluster cloud-edge collaborative optimization control method, device and application architecture, solves the problems that the flexible resources of the prior art building are difficult to be efficiently integrated into the power grid and the core algorithm of the prior art is mostly based on theory and has not been verified for real scene adaptability and convergence, and realizes local solution algorithm and precise control, so that the building aggregator can be promoted in scale and the operation efficiency can be improved.

[0006] In order to achieve the above-mentioned purpose, the application provides the following technical scheme: a building cluster cloud-edge collaborative optimization control method, comprising: Based on the characteristics of building envelope structure and the physical law of heat transfer, a building thermodynamic RC virtual energy storage model is constructed; Based on the building thermal dynamic characteristics output by the building thermodynamic RC virtual energy storage model, a model predictive control algorithm MPC is constructed; Based on the adjustable potential parameters output by the building thermodynamic RC virtual energy storage model and the control constraint boundary of the model predictive control algorithm MPC, an alternating direction multiplier method distributed optimization model is constructed; the alternating direction multiplier method distributed optimization model comprises a user side sub-problem optimization model and an aggregator side sub-problem optimization model; The virtual variables, dual variables and iteration number of the alternating direction multiplier method distributed optimization model are initialized; The user side edge control module collects building real-time operation data through a sensor; based on the building real-time operation data, the building thermodynamic RC virtual energy storage model is calculated to obtain the current adjustable potential of the building; Based on the current adjustable potential and the current incentive price, the user side sub-problem optimization model is solved to obtain user response quantity data, and the user response quantity data is uploaded to the cloud; The cloud processing module solves the aggregator side sub-problem optimization model based on the user response quantity data and the microgrid operation constraint to obtain a differentiated incentive price, and the differentiated incentive price is issued to the user side; Based on the user response quantity data and the differentiated incentive price, the virtual variables and dual variables of the alternating direction multiplier method distributed optimization model are updated to obtain updated variables; Based on the updated variables, residual error and dual residual error are calculated; based on the residual error and the dual residual error, the alternating direction multiplier method distributed optimization model is judged for convergence; if convergence is achieved, the optimal user response quantity and the optimal differentiated incentive price are output; if convergence is not achieved, the user side sub-problem optimization model and the aggregator side sub-problem optimization model are iteratively optimized until the alternating direction multiplier method distributed optimization model converges; Based on the optimal user response quantity, the optimal differentiated incentive price and the disturbance prediction data of the future setting time length, an optimal control strategy is generated through the model predictive control algorithm MPC processing. The control time domain is moved forward by one period, the building real-time operation data is re-acquired, the next round of optimization is performed, and rolling optimization control is realized.

[0007] As a preferred scheme of the building cluster cloud-edge collaborative optimization control method, the process of constructing the building thermodynamics RC virtual energy storage model is: The building nodes are divided by a thermal resistance-thermal capacity network strategy; the building nodes include wall nodes and indoor air nodes; Based on the building nodes, a node heat balance relationship is constructed; Based on the node heat balance relationship, the thermal resistance and thermal capacity parameters are determined by the least square method combined with building characteristic data; Based on the thermal resistance and the thermal capacity parameters, the building thermal inertia is converted into the "virtual battery" characteristics, the building energy storage state and the up and down regulation capacity are quantified, and the model construction is completed.

[0008] As a preferred scheme of the building cluster cloud-edge collaborative optimization control method, the process of constructing the model predictive control algorithm MPC is: The control period, the prediction time domain and the data sampling interval are set to obtain time parameters; Based on the time parameters, a target function including temperature deviation and device energy consumption is constructed; Based on the target function, combined with the up and down regulation capacity output by the building thermodynamics RC virtual energy storage model, the constraint conditions of temperature comfort interval and device power limit are set, the optimization feasible region is delineated, and the construction is completed.

[0009] As a preferred scheme of the building cluster cloud-edge collaborative optimization control method, the user-side sub-problem optimization model takes the minimum energy consumption of a single building as the target; the expression of the target function is: ; In the formula, N is the total number of heating areas; n is a heating area; T is the number of daily regulation cycles; t is the number of daily regulation times; is the time interval; is the user's electricity purchase price; is the electricity purchase power of a single heating area to the microgrid; The constraint conditions of the user-side sub-problem optimization model include: heat balance constraint, indoor temperature comfort range constraint, HVAC system constraint, real-time electricity price constraint and average electricity price constraint.

[0010] As a preferred scheme of a building cluster cloud-edge collaborative optimization control method, the aggregator side sub-problem optimization model maximizes the micro-grid operation revenue of the building cluster as an objective; the objective function expression is: ; In the formula, A is the total number of building clusters; a is the building cluster number; is the electricity selling price of the micro-grid to the user; is the electricity selling power of the micro-grid to a single heating area; is the electricity purchasing price of the micro-grid system from the power distribution network; is the electricity purchasing power of the micro-grid from the power distribution network; is the economic cost of the power distribution network imposed on the building aggregator when congestion occurs; The constraint conditions of the aggregator side sub-problem optimization model include: micro-grid power balance constraint, electricity purchase and sale price constraint and power distribution network node voltage constraint.

[0011] The application also provides a building cluster cloud-edge collaborative optimization control device based on the above building cluster cloud-edge collaborative optimization control method, comprising: A building thermodynamics RC virtual energy storage model construction unit is configured to construct a building thermodynamics RC virtual energy storage model based on building envelope characteristics and heat transfer physical laws; A model predictive control algorithm MPC construction unit is configured to construct a model predictive control algorithm MPC based on building thermal dynamic characteristics output by the building thermodynamics RC virtual energy storage model; An alternating direction multiplier method distributed optimization model construction unit is configured to construct an alternating direction multiplier method distributed optimization model based on adjustable potential parameters output by the building thermodynamics RC virtual energy storage model and control constraint boundaries of the model predictive control algorithm MPC; the alternating direction multiplier method distributed optimization model comprises a user side sub-problem optimization model and an aggregator side sub-problem optimization model; An alternating direction multiplier method distributed optimization model initialization unit is configured to initialize virtual variables, dual variables and iteration numbers of the alternating direction multiplier method distributed optimization model; A current adjustable potential calculation unit is configured to acquire building real-time operation data by a user side edge control module through a sensor; based on the building real-time operation data, the building thermodynamics RC virtual energy storage model is used for calculation to obtain current adjustable potential of the building; A user response quantity data acquisition and uploading unit is configured to solve the user side sub-problem optimization model based on the current adjustable potential and current incentive price, acquire user response quantity data, and upload the user response quantity data to the cloud; The differentiated incentive price acquisition and distribution unit is used by the cloud processing module to solve the optimization model of the aggregator-side sub-problem based on the user response volume data and microgrid operation constraints, obtain the differentiated incentive price, and distribute the differentiated incentive price to the user side. The variable update unit is used to update the dummy variables and dual variables of the alternating direction multiplier method distributed optimization model based on the user response volume data and the differentiated incentive price, so as to obtain the updated variables; The model convergence judgment and processing unit is used to calculate the residuals and dual residuals based on the updated variables; to perform convergence judgment on the alternating direction multiplier method distributed optimization model based on the residuals and dual residuals; if converged, the optimal user response quantity and the optimal differentiated incentive price are output; if not converged, the user-side subproblem optimization model and the aggregator-side subproblem optimization model are iteratively optimized until the alternating direction multiplier method distributed optimization model converges. The optimal control strategy acquisition unit is used to generate the optimal control strategy based on the optimal user response volume, the optimal differentiated incentive price, and disturbance prediction data for a future set duration, through the Model Predictive Control (MPC) algorithm. The rolling optimization control unit is used to move the control time domain forward by one cycle, re-collect the real-time operating data of the building, and perform the next round of optimization to achieve rolling optimization control.

[0012] As a preferred embodiment of a building cluster cloud-edge collaborative optimization control device, the process of constructing the building thermodynamic RC virtual energy storage model in the building thermodynamic RC virtual energy storage model construction unit is as follows: Building nodes are divided using a thermal resistance-thermal capacity network strategy; the building nodes include wall nodes and indoor air nodes. Based on the aforementioned building nodes, a node thermal balance relationship is constructed; Based on the nodal thermal balance relationship and combined with building characteristic data, the thermal resistance and heat capacity parameters are determined by the least squares method. Based on the thermal resistance and thermal capacity parameters, the building's thermal inertia is transformed into "virtual battery" characteristics, and the building's energy storage status and vertical adjustment capacity are quantitatively output to complete the model construction.

[0013] As a preferred embodiment of a cloud-edge collaborative optimization control device for building clusters, the process of constructing the Model Predictive Control Algorithm (MPC) in the MPC construction unit is as follows: The control period, prediction time domain, and data sampling interval are set to obtain the time parameters; Based on the time parameters, an objective function is constructed that includes temperature deviation and equipment energy consumption; Based on the target function, the up and down adjustment capacity output by the building thermodynamics RC virtual energy storage model is combined, the constraint conditions of temperature comfort interval and equipment power limit are set, the optimization feasible region is delimited, and the construction is completed.

[0014] As a preferred scheme of the building cluster cloud-edge collaborative optimization control device, in the alternating direction multiplier method distributed optimization model construction unit, the user side sub-problem optimization model takes the minimum energy consumption cost of a single building as the target; the expression of the target function is: ; In the formula, N is the total number of heating areas; n is a heating area; T is the number of daily regulation cycles; t is the number of daily regulation times; is a time interval; is the electricity purchase price of the user; is the electricity purchase power of a single heating area to the microgrid; The constraint conditions of the user side sub-problem optimization model include: heat balance constraint, indoor temperature comfort range constraint, HVAC system constraint, real-time electricity price constraint and average electricity price constraint.

[0015] As a preferred scheme of the building cluster cloud-edge collaborative optimization control device, in the alternating direction multiplier method distributed optimization model construction unit, the aggregator side sub-problem optimization model takes the maximum microgrid operation income of the building cluster as the target; the expression of the target function is: ; In the formula, A is the total number of building clusters; a is the building cluster number; is the electricity selling price of the microgrid to the user; is the electricity selling power of the microgrid to a single heating area; is the electricity purchase price of the microgrid system to the power distribution network; is the electricity purchase power of the microgrid to the power distribution network; is the economic cost of the power distribution network imposed on the building aggregator when congestion occurs; The constraint conditions of the aggregator side sub-problem optimization model include: microgrid power balance constraint, electricity purchase and sale price constraint and power distribution network node voltage constraint.

[0016] The application also provides an application architecture of a building cluster cloud-edge collaborative optimization control method, which adopts a hierarchical architecture design of a perception layer, a control execution layer, a communication layer and an algorithm layer. The hardware integration of the perception layer is based on an edge computing device, and the infrared perception device, the light intensity detection device and the temperature and humidity acquisition device for monitoring the building environment are mounted on the edge computing device, so that the real-time data of the indoor and outdoor temperature, the light intensity and the environmental humidity of the building can be comprehensively collected. The design of the control execution layer is to build a bidirectional communication and control link through the infrared signal transmitting component and the infrared signal receiving component, realize remote regulation and control of the air conditioning system in the building, the infrared receiving component converts the infrared remote control signal of the set protocol into a digital pulse signal for the edge computing device to analyze and store, and the infrared transmitting component outputs a pulse signal with a specific frequency and duty cycle according to the analyzed control instruction, and sends it to the air conditioning equipment after modulation, dynamically adjusts the air conditioning operation parameters. The architecture and security mechanism of the communication layer are to build a communication link using a low-power long-range radio frequency communication module combined with a data encryption algorithm; the communication module is adapted to the communication needs of multi-node and distributed building environment; the communication module is connected with the edge computing core device and the cloud processing device through a serial interface to realize bidirectional data interaction between the edge side and the cloud side. The software module integration and collaboration process of the algorithm layer is that the edge side embedded software has the functions of collecting data from the perception layer, preprocessing the collected data, generating control instructions based on the data, and calling the optimization solver to execute the model predictive control algorithm, combining the building thermal dynamic model and real-time environmental data to generate optimal operation instructions for air conditioning, and using the infrared protocol analysis tool to realize the encoding and sending of infrared control signals; the cloud software platform integrates distributed optimization algorithm modules, data storage systems and network service interfaces, executes distributed optimization operations based on the optimization solver, receives power demand data uploaded from the edge side, combines the microgrid operation model for global coordination, generates differentiated electricity price signals and sends them to the edge side, and saves historical operation data through the data storage system; The hardware platform building process is to take the edge computing core device as the center, and sequentially complete the selection and assembly of various collection devices in the perception layer, the deployment and function debugging of the communication module in the communication layer, and the connection and operation verification of the infrared control component in the control execution layer. The communication protocol and edge-cloud data interaction mechanism is to develop a system communication protocol based on a custom instruction set, build a radio frequency communication link relying on a long-range radio frequency communication module, and meet the data interaction needs in the edge-cloud collaboration scenario.

[0017] The application has the following advantages: the application is based on the characteristics of building envelope and the physical law of heat transfer to construct a building thermodynamic RC virtual energy storage model; based on the building thermal dynamic characteristics output by the building thermodynamic RC virtual energy storage model, a model predictive control algorithm MPC is constructed; based on the adjustable potential parameters output by the building thermodynamic RC virtual energy storage model and the control constraint boundary of the model predictive control algorithm MPC, an alternating direction multiplier method distributed optimization model is constructed; the alternating direction multiplier method distributed optimization model includes a user side sub-problem optimization model and an aggregator side sub-problem optimization model; the virtual variables, dual variables and iteration number of the alternating direction multiplier method distributed optimization model are initialized; a user side edge control module collects real-time operation data of the building through a sensor; based on the real-time operation data of the building, the current adjustable potential of the building is obtained through calculation by the building thermodynamic RC virtual energy storage model; based on the current adjustable potential and the current incentive price, the user side sub-problem optimization model is solved to obtain user response quantity data, and the user response quantity data is uploaded to the cloud; a cloud processing module solves the aggregator side sub-problem optimization model based on the user response quantity data and micro-grid operation constraints to obtain a differentiated incentive price, and the differentiated incentive price is issued to the user side; based on the user response quantity data and the differentiated incentive price, the virtual variables and dual variables of the alternating direction multiplier method distributed optimization model are updated to obtain updated variables; based on the updated variables, residual error and dual residual error are calculated; based on the residual error and the dual residual error, convergence judgment is performed on the alternating direction multiplier method distributed optimization model; if convergence is achieved, the optimal user response quantity and the optimal differentiated incentive price are output; if convergence is not achieved, the user side sub-problem optimization model and the aggregator side sub-problem optimization model are iteratively optimized until the alternating direction multiplier method distributed optimization model converges; based on the optimal user response quantity, the optimal differentiated incentive price and disturbance prediction data of a future set time length, the optimal control strategy is generated through the model predictive control algorithm MPC processing; the control time domain is moved forward by one period, the real-time operation data of the building is re-collected, the next round of optimization is performed, and rolling optimization control is realized. The application breaks through the limitation of single building energy efficiency optimization, links building clusters and power grid scheduling through the ADMM distributed model and edge-cloud collaborative architecture, takes into account the reduction of user energy cost, the improvement of aggregator revenue and the balance of power grid power, and realizes the win-win of all parties. The core algorithms of the application are designed in combination with real scenarios, the MPC algorithm is adapted to the building thermal dynamic characteristics, the ADMM algorithm verifies the low bandwidth adaptability and stable convergence, and the local solution at the edge protects user privacy, solving the problem of disconnection between algorithm theory and application.The application integrates multiple types of sensing, edge computing and infrared control modules, and the edge device can synchronously realize data acquisition, local algorithm solving and HVAC precise control, avoids the disconnection of "algorithm-hardware-scene", and supports large-scale promotion. The application quantifies the building regulation potential based on the RC virtual energy storage model, and generates a rolling control sequence by combining the MPC algorithm with future disturbance prediction, which can not only ensure that the indoor temperature is stable in the comfortable interval, but also flexibly respond to the grid regulation demand and adapt to complex working conditions. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only exemplary, and for those skilled in the art, other embodiments can be derived from the provided drawings without creative labor.

[0019] The structures, proportions, sizes, etc. shown in the specification are only used to cooperate with the content disclosed in the specification, to be understood and read by those skilled in the art, and are not used to limit the limiting conditions that the present application can be implemented, so they do not have technical significance. Any modification of structure, change of proportion relationship or adjustment of size, without affecting the effect and purpose that the present application can produce, should still fall within the scope of the technical content disclosed by the present application.

[0020] Figure 1 A flowchart of a building cluster cloud-edge collaborative optimization control method provided in embodiment 1 of the present application; Figure 2 A general idea diagram of a building cluster cloud-edge collaborative optimization control method provided in embodiment 1 of the present application; Figure 3 A building multi-element data acquisition diagram in a building cluster cloud-edge collaborative optimization control method provided in embodiment 1 of the present application; Figure 4 A single heating area RC network model diagram in a building cluster cloud-edge collaborative optimization control method provided in embodiment 1 of the present application; Figure 5 A single-line radial topology structure diagram of a power distribution network in a building cluster cloud-edge collaborative optimization control method provided in embodiment 1 of the present application; Figure 6 An ADMM-MPC coupling algorithm diagram in a building cluster cloud-edge collaborative optimization control method provided in embodiment 1 of the present application; Figure 7 An application architecture diagram of a building cluster cloud-edge collaborative optimization control method provided in embodiment 1 of the present application; Figure 8 Fig. 1 is a schematic diagram of temperature variation and comfort interval of each building in a possible embodiment provided in Embodiment 1 of the present application; Figure 9 Fig. 2 is a schematic diagram of total electricity cost reduction of each building in a possible embodiment provided in Embodiment 1 of the present application; Figure 10 Fig. 3 is a schematic diagram of negotiated electricity price and power of each building in a possible embodiment provided in Embodiment 1 of the present application; Figure 11 Fig. 4 is a schematic diagram of economic benefit analysis in a possible embodiment provided in Embodiment 1 of the present application; Figure 12 Fig. 5 is a schematic diagram of architecture of a building cluster cloud-edge collaborative optimization control device provided in Embodiment 2 of the present application. DETAILED DESCRIPTION

[0021] The embodiments of the present application will be described in detail by specific embodiments, and those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in the specification. Obviously, the described embodiments are part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0022] Embodiment 1

[0023] Reference Figure 1 , Embodiment 1 of the present application provides a building cluster cloud-edge collaborative optimization control method, comprising the following steps: S1, based on the characteristics of building envelope structure and the physical law of heat transfer, a building thermodynamic RC virtual energy storage model is constructed; S2, based on the building thermal dynamic characteristics output by the building thermodynamic RC virtual energy storage model, a model predictive control algorithm MPC is constructed; S3, based on the adjustable potential parameters output by the building thermodynamic RC virtual energy storage model and the control constraint boundary of the model predictive control algorithm MPC, an alternating direction multiplier method distributed optimization model is constructed; the alternating direction multiplier method distributed optimization model includes a user-side sub-problem optimization model and an aggregator-side sub-problem optimization model; S4, the virtual variables, dual variables and iteration number of the alternating direction multiplier method distributed optimization model are initialized; S5, the user-side edge control module collects real-time operation data of the building through sensors; based on the real-time operation data of the building, the current adjustable potential of the building is obtained through calculation by the building thermodynamic RC virtual energy storage model; S6, based on the current adjustable potential and the current incentive price, solving the user-side sub-problem optimization model to obtain user response data, and uploading the user response data to the cloud; S7, the cloud processing module based on the user response data and the micro-grid operation constraints, solving the aggregator-side sub-problem optimization model to obtain a differentiated incentive price, and issuing the differentiated incentive price to the user side; S8, updating the virtual variables and dual variables of the alternating direction multiplier method distributed optimization model based on the user response data and the differentiated incentive price to obtain updated variables; S9, based on the updated variables, calculating the residual and the dual residual; based on the residual and the dual residual, making a convergence judgment on the alternating direction multiplier method distributed optimization model; if converged, outputting the optimal user response and the optimal differentiated incentive price; if not converged, iteratively optimizing the user-side sub-problem optimization model and the aggregator-side sub-problem optimization model until the alternating direction multiplier method distributed optimization model converges; S10, based on the optimal user response, the optimal differentiated incentive price and the disturbance prediction data of the future set time length, generating an optimal control strategy through the model predictive control algorithm MPC processing; S11, moving the control time domain forward by one cycle, reacquiring the building real-time operation data, and performing the next round of optimization to realize rolling optimization control.

[0024] In this embodiment, as Figure 2As shown, starting from the logical chain of "real demand-algorithm solution-hardware implementation-research target": First, in view of the high complexity of building thermal inertia and user behavior, which makes it difficult to quantify and evaluate the flexibility boundary, the thermal RC model is established to convert the building thermal inertia into a "virtual battery", forming a measurable energy storage state and up and down adjustment capacity, and realizing the visualization and quantitative evaluation of flexibility potential. Second, in view of the strong coupling and randomness of building environment and load, which makes it difficult for traditional control methods to ensure high precision response under comfort constraints, the model predictive control (MPC) is used to optimize the load trajectory in time sequence and dynamically correct it, which significantly improves the adjustment accuracy and resists the uncertainty of external disturbance. Third, in view of the uncertainty and difference of response cost, benefit distribution and incentive contract in building cluster involving multiple subjects, the alternating direction method of multipliers (ADMM) is used to realize the distributed coordination between the aggregator and multiple building users, which can quickly converge under low communication bandwidth and take into account the individualized constraints and benefit distribution of different users, realizing fair interaction and stable response at the cluster level. Fourth, the "edge-management-cloud" integrated hardware support is realized, that is, local calculation is performed on the edge using Raspberry Pi microprocessor, low-power long-distance transmission is realized on the communication layer using LoRa module, and global coordination is performed on the cloud building aggregation platform. Finally, the cloud-edge collaborative intelligent system and device for building HVAC cluster are formed, realizing the unified closed loop of flexibility evaluation, control and interaction.

[0025] In this embodiment, in step S1, based on the characteristics of building envelope and the physical law of heat transfer, a building thermodynamic RC virtual energy storage model is constructed; Specifically, the building is divided into wall and air nodes by the thermal resistance-capacity network method, the node heat balance relationship is established, the characteristics data of building materials, area, orientation, etc. are combined, the least squares method is used to identify the thermal resistance and capacity parameters, the inherent thermal inertia of the building is converted into quantifiable "virtual battery" characteristics, and finally the building energy storage state and up and down adjustment capacity are output, providing the basis for building physical constraints for subsequent optimization.

[0026] In this embodiment, a perfect building cluster side energy consumption analysis model is established, which comprehensively considers the influence factors such as environment, behavior and internal architecture, and dynamically formulates the accurate prediction and energy management strategy of building cluster energy consumption. The required data types are divided into four categories: Building cluster itself characteristic data, such as: orientation, area, envelope material and heat transfer coefficient, etc.; External environment data, including: light intensity and outdoor temperature; Electrical quantity data, such as: air conditioner power, air conditioner energy efficiency ratio and total room power; User behavior data: human flow and air conditioner running state, etc.

[0027] In this embodiment, in step S2, a model predictive control algorithm (MPC) is constructed based on the building thermal dynamic characteristics output by the building thermodynamic RC virtual energy storage model. Specifically, based on the building thermal dynamics reflected by the RC model, a 1-hour control cycle and a 24-hour prediction time domain are set, and an objective function including temperature deviation and equipment energy consumption is constructed. At the same time, combined with the adjustable capacity output by the RC model, constraints such as indoor temperature comfort range and air conditioning power limit are set to form a rolling optimization control framework that can cope with future disturbances, providing algorithmic support for the precise regulation of the air conditioning system.

[0028] In this embodiment, in step S3, based on the adjustable potential parameters output by the building thermodynamics (RC) virtual energy storage model and the control constraint boundary of the model predictive control algorithm (MPC), an alternating direction multiplier method distributed optimization model is constructed; the alternating direction multiplier method distributed optimization model includes a user-side sub-problem optimization model and an aggregation quotient-side sub-problem optimization model. Specifically, by integrating the adjustable potential of the RC model with the control constraints of MPC, a two-layer sub-problem architecture is established: the user-side sub-problem aims to minimize energy costs, with the upper and lower limits of decision variables set based on the adjustable potential of the RC model; the aggregator-side sub-problem aims to maximize microgrid revenue and incorporates power balance constraints of the distribution network; by introducing dummy variables and augmented Lagrangian functions, decoupling between the two sides is achieved, and linear relaxation is used to handle nonlinear terms, forming a distributed optimization mechanism that adapts to edge-cloud collaboration.

[0029] The user-side sub-problem optimization model aims to minimize the energy cost of a single building; the objective function is expressed as follows: ; In the formula, N is the total number of heating areas; n is the number of heating areas; T is the number of daily control cycles; and t is the number of daily control operations. For time intervals; The electricity price for users; The power purchased from the microgrid by a single heating zone; The constraints of the user-side sub-problem optimization model include: thermal balance constraints, indoor temperature comfort range constraints, HVAC system constraints, real-time electricity price constraints, and average electricity price constraints.

[0030] Specifically, Thermal balance constraints: An equivalent thermal parameter model of the building is constructed using the thermal resistance-thermal capacity RC network method to describe the heat transfer and storage process of the building envelope and internal air. For example... Figure 3 As shown, taking a single heating zone 1 as an example, the thermal balance constraints are defined for both the hot spots on the wall and the hot spots of the indoor air, based on the law of conservation of energy: ; ; In the formula, the subscript 'a' represents the building cluster number; , and These are heating zone 1 and air node, respectively. The heat capacity, midpoint temperature, and thermal resistance of the partition wall; , and These are the heat absorption rate, surface area, and light intensity of the walls between heating zone 1 and the air node, respectively. This is a binary variable, representing a value of 1 when the wall receives sunlight and a value of 0 otherwise. , and These are the window's thermal resistance, area, and transmittance, respectively. air node Temperature; The indoor air temperature of heating zone 1; , , and These are the indoor air heat capacity, air specific heat capacity, HVAC airflow rate, and air supply temperature for heating zone 1, respectively. and These represent the random heat gain of indoor air and the heat gain from solar radiation in heating zone 1, respectively.

[0031] The above equations take into account conduction, convection, solar radiation heat gain, internal random heat gain, and the heating and cooling output of the HVAC system.

[0032] Indoor temperature comfort range constraints: like Figure 4 As shown, to ensure user experience, a comfortable indoor temperature range is set, i.e., the building's indoor temperature... It must be kept within the preset comfort range.

[0033] ; In the formula, , These represent the minimum and maximum indoor temperatures of the building, respectively.

[0034] HVAC system constraints: For controllable load models, the strategy of adjusting the supply air outlet temperature is used to control HVAC: ; ; ; In the formula, and The airflow temperature at the HVAC supply outlet is respectively The lower limit; , and These represent the electricity purchased from the microgrid by a single heating zone, the electricity consumed by HVAC systems, and the electricity consumed by uncontrollable electrical equipment, respectively. This refers to the energy efficiency ratio of HVAC equipment.

[0035] Real-time electricity price constraints: ; In the formula, The electricity price for users; and These are the lower and upper limits for microgrid electricity sales prices, respectively. Average electricity price constraint: ; In the formula, This represents the upper limit of the average electricity sales price for microgrids; This refers to the number of intraday adjustment cycles.

[0036] In this embodiment, the optimization model for the aggregation merchant sub-problem aims to maximize the microgrid operation benefits of the building cluster; the objective function expression is: ; In the formula, A represents the total number of building clusters; a represents the building cluster number. The electricity price for microgrids to supply power to users; The power output of the microgrid sold to a single heating area; The price at which the microgrid system purchases electricity from the distribution network; This refers to the power purchased by the microgrid from the distribution network. The economic cost imposed on building aggregators when a blockage occurs in the distribution network; The constraints of the optimization model for the aggregation business side problem include: microgrid power balance constraints, electricity purchase and sale price constraints, and distribution network node voltage constraints.

[0037] Specifically, Microgrid power balance constraints: ; In the formula, , These represent the power purchased by the microgrid from the distribution network and the total power generation of the distributed PV system within the microgrid, respectively. A represents the power output of the microgrid to a single heating area; A represents the total number of building clusters.

[0038] Electricity purchase and sale price constraints: Electricity purchase and sale prices must be subject to both real-time electricity prices and average electricity prices: ; ; In the formula, The electricity price charged to users for microgrids.

[0039] Distribution network node voltage constraints: The power flow network of the distribution network adopts a single-line radial topology, such as... Figure 5 As shown. For any two adjacent nodes m and m+1, the power flow at any time must satisfy the constraints shown, and the node voltages must satisfy: ; ; ; ; In the formula, and Injection nodes The active and reactive power; and They are nodes The active and reactive power consumed by the load; For nodes The voltage; and They are nodes and The resistance and reactance between them; This represents the offset margin of the per-unit value of the node voltage.

[0040] In this embodiment, in step S4, the dummy variables, dual variables, and iteration count of the alternating direction multiplier method distributed optimization model are initialized; Specifically, such as Figure 6 As shown, initial conditions are set for the iterative solution of the ADMM algorithm, including setting the dummy variable as the mean of the initial response of each user, setting the dual variable as the zero vector, setting the iteration number k to 0, and setting a convergence accuracy threshold to ensure that the subsequent edge-cloud iterative optimization has a clear starting point and termination criteria.

[0041] In this embodiment, in step S5, the user-side edge control module collects real-time building operation data through sensors; based on the real-time building operation data, it calculates the current adjustable potential of the building through the building thermodynamic RC virtual energy storage model. Specifically, on the user side, multiple types of sensors, such as infrared thermal imagers and temperature and humidity sensors, are used to collect real-time data on indoor and outdoor temperature and light intensity. This data is then input into the RC virtual energy storage model. The model calculates the adjustable power range and temperature tolerance range of the building at the current moment, reflecting the building's flexible adjustment capability while ensuring comfort. This provides a dynamic constraint basis for solving sub-problems on the user side.

[0042] In this embodiment, in step S6, the user-side sub-problem optimization model is solved based on the current adjustable potential and the current incentive price to obtain user response data, and the user response data is uploaded to the cloud. Specifically, the edge device uses the current adjustable potential as the boundary and combines it with the latest incentive price issued by the cloud to solve the subproblem of minimizing user-side costs, thereby obtaining the optimal electricity purchase / regulation plan within the future regulation cycle, i.e., the user response volume. Only the optimization results, rather than the original data, are uploaded to the cloud, thus protecting user privacy while participating in global optimization.

[0043] In this embodiment, in step S7, the cloud processing module solves the optimization model of the aggregator-side sub-problem based on the user response volume data and microgrid operation constraints to obtain the differentiated incentive price, and then sends the differentiated incentive price to the user side. Specifically, the cloud integrates all user-uploaded response data, and solves the aggregator revenue maximization sub-problem under the premise of satisfying global constraints such as distribution network voltage constraints, power balance and congestion costs. It generates time-sharing incentive prices for different buildings, and guides users to adjust their response strategies through price signals, forming a closed-loop interaction between edge and cloud.

[0044] In this embodiment, in step S8, the dummy variables and dual variables of the alternating direction multiplier method distributed optimization model are updated based on the user response volume data and the differentiated incentive price to obtain the updated variables; Specifically, based on the interaction between the user-side response volume and the aggregator-side incentive price, the dummy variables and dual variables are adjusted according to the ADMM update rules, so that the solutions to the two subproblems gradually converge, driving the global optimization towards convergence.

[0045] In this embodiment, in step S9, based on the updated variables, residuals and dual residuals are calculated; based on the residuals and dual residuals, the convergence of the alternating direction multiplier method distributed optimization model is determined; if converged, the optimal user response amount and the optimal differentiated incentive price are output; if not converged, the user-side subproblem optimization model and the aggregator-side subproblem optimization model are iteratively optimized until the alternating direction multiplier method distributed optimization model converges. Specifically, by calculating the residuals of the user response quantity and the dummy variable, as well as the dual residuals of the dual variable, it is determined whether the preset convergence threshold is met; if the threshold is met, the final optimization result is output; otherwise, the process returns to step S6 to resolve the two sub-problems, and through multiple rounds of iteration, the consistency between the local optimum on the user side and the global optimum on the aggregate quotient side is achieved.

[0046] In this embodiment, in step S10, based on the optimal user response volume, the optimal differentiated incentive price, and the disturbance prediction data of the future set duration, the optimal control strategy is generated through the model predictive control algorithm MPC. Specifically, after the ADMM algorithm converges, the optimal response quantity and incentive price are used as inputs. Combined with the predicted data of outdoor temperature, solar radiation and other disturbances for the next 12 hours, the MPC algorithm is called to generate the optimal control sequence of the air conditioning system, such as temperature setting and operating power, to ensure that the indoor temperature is kept stable in the comfort range while meeting economic objectives.

[0047] In this embodiment, in step S11, the control time domain is moved forward by one cycle, the real-time operation data of the building is collected again, and the next round of optimization is performed to achieve rolling optimization control.

[0048] Specifically, after executing the first action of the MPC control sequence, the control time domain is advanced by 1 hour, and the process returns to step S5 to re-collect real-time data and start a new round of optimization. Through periodic rolling updates, the system dynamically adapts to changes in building thermal characteristics and external disturbances, ensuring the stability and robustness of long-term optimization effects.

[0049] See Figure 7 In this embodiment, the application of a cloud-edge collaborative optimization control method for building clusters adopts a layered architecture design to achieve integration and collaboration from the perception layer to the control layer, communication layer, and algorithm layer, ensuring effective operation. The specific integration structure and implementation mechanism are as follows: Perception layer hardware integration: Using a Raspberry Pi 4B as the edge computing core, equipped with an MLX90640 infrared thermal imager, a GY-302 BH1750 light intensity sensor, a DHT22 temperature and humidity sensor, etc., to comprehensively perceive building environment data.

[0050] Control Execution Design: The control execution layer establishes a bidirectional communication and control channel between the infrared transmitting module (TSAL6200) and the infrared receiving module (HS-0038B) to achieve remote control of the air conditioning system. The infrared receiver uses a high-sensitivity infrared diode to convert the NEC protocol infrared remote control signal into a digital pulse signal for edge device parsing and storage. The transmitting end, based on the parsed control commands, outputs a pulse signal with a specific frequency and duty cycle through GPIO pins. This pulse signal is modulated by the infrared emitting diode and sent to the air conditioning equipment, enabling dynamic adjustment of the air conditioning operating parameters.

[0051] Communication Architecture and Security Mechanism: A low-power, long-distance radio frequency communication link is constructed using LoRa modules, combined with encryption algorithms to ensure data transmission security. It supports long-distance, low-power, and highly interference-resistant data transmission, suitable for stable communication in multi-node, distributed building environments. The LoRa modules are integrated between the Raspberry Pi 4B and the PC via a serial interface, enabling bidirectional data exchange.

[0052] Software module integration and collaboration mechanism: The edge-end embedded software is written in Python and has functions such as timed data acquisition, data preprocessing, model prediction, and control command generation. The system executes the MPC control algorithm by calling the OSQP optimization solver, combining the building thermal dynamics model and real-time environmental data to generate optimal air conditioning operation commands, and encodes and transmits infrared signals through the NEC protocol parsing library. The cloud software platform integrates the ADMM optimization algorithm module, database system, and possible Web service interfaces. The cloud performs ADMM distributed optimization based on the OSQP solver, receives power demand data uploaded by edge devices, performs global coordination in conjunction with the microgrid model, generates electricity price signals and sends them to the edge side, and stores historical operation data in the database.

[0053] Hardware platform setup process: The hardware platform of the edge intelligent device is based on Raspberry Pi 4B, and the selection and integration of sensor modules, deployment and debugging of communication modules, and connection and verification of infrared control modules are completed in sequence.

[0054] Communication Protocol and Edge-Cloud Data Interaction Mechanism: The system communication protocol is built on a custom instruction set and relies on the LoRa module ATK-MW1268D-1W to implement the radio frequency communication link. It has the characteristics of low power consumption, long distance and strong anti-interference, and is suitable for edge-cloud collaborative data interaction needs.

[0055] In one possible embodiment, a simulation verification example is provided as follows: Simulation verification mainly relies on the Python platform and the OSQP optimization solver to comprehensively test the constructed mathematical model and the designed optimization algorithm.

[0056] The simulation scenario is constructed as a building microgrid containing three buildings (A, B, and C) with different thermal characteristics, a distributed photovoltaic (PV) system, and a building aggregator. Key parameters are set as follows: Building characteristics: The three buildings have different thermal insulation performance, with building A being superior to building B, and building B being superior to building C. This differentiated setting aims to test the algorithm's adaptability and personalized optimization capabilities for buildings with different thermal inertia.

[0057] External conditions: The simulation environment is set as a typical winter day (24 hours), and the input data includes outdoor temperature and solar radiation intensity that vary over time.

[0058] Internal loads: Uncontrollable loads within each building (such as lighting and office equipment) and random factors such as human activity have also been incorporated into the model.

[0059] Optimization objective: The core objective of optimization is to minimize the overall energy cost of the entire building complex by coordinating and scheduling multiple intelligent agents, while ensuring that the indoor temperature of each building is always within the user-defined comfort range.

[0060] Simulation results analysis: like Figure 8 As shown, even while performing energy-saving optimization, the indoor temperatures of the three buildings were maintained between their respective preset comfort limits (as shown by the colored dashed lines in the figure), verifying the algorithm's ability to balance energy saving and comfort.

[0061] like Figure 9 As shown, the total electricity cost (blue bar) of each building (A, B, C) after optimization is significantly lower than that before optimization (pink bar), and building A, which has better thermal insulation performance, may have more significant energy-saving potential and cost savings.

[0062] like Figure 10 As shown, the dynamics of the electricity price (left axis) and exchange power (right axis) negotiated between each building and aggregator during the ADMM iteration process are displayed, intuitively reflecting the distributed collaborative optimization process.

[0063] like Figure 11 As shown, the upper subplot clearly displays the aggregator's cumulative electricity purchase cost (red dashed line), cumulative electricity sales revenue (green dashed line), and the resulting cumulative profit (blue solid line), proving the economic feasibility on the aggregator's side. The lower subplot shows the optimized cumulative energy costs for each building (A, B, C), further corroborating the economic benefits on the user side. This achieves a win-win equilibrium solution where the aggregator profits while user costs are reduced.

[0064] ADMM algorithm convergence: In the example, the ADMM algorithm satisfies the convergence condition after about 500 iterations. The obtained equilibrium solution is basically consistent with the result of the centralized algorithm, which verifies the effectiveness and convergence of the distributed algorithm.

[0065] The application scenarios of this invention are as follows: In urban commercial building clusters, the decentralized air conditioning system regulation capabilities can be integrated through a cloud-edge collaborative architecture to participate in grid demand response. When the grid load is at its peak, aggregators can guide each building to adjust its indoor temperature settings appropriately through differentiated incentive prices to release load potential; when the load is at its off-peak, buildings are encouraged to use low-cost electricity for heat storage, which reduces users' energy costs and helps the grid to smooth out peak and valley loads.

[0066] In smart parks or industrial parks, the RC model is used to accurately quantify the adjustment potential of each building based on the different energy consumption characteristics of various types of buildings in the park. Combined with the ADMM algorithm, the interaction between distributed photovoltaics, energy storage and building load is coordinated to achieve the optimal cost allocation between energy self-sufficiency within the park's microgrid and electricity purchase from the external grid, thereby improving the park's energy self-governance capabilities.

[0067] In residential community scenarios, relying on the local computing power of edge devices, the potential for regulating the air conditioning of thousands of households can be aggregated while protecting user privacy. This allows the system to participate in ancillary services in the electricity market. Through the MPC algorithm, it ensures that the regulation process does not affect residents' comfort, while allowing users to obtain economic benefits through flexible energy use, thus forming a virtuous cycle of interaction between the power grid, aggregators, and users.

[0068] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.

[0069] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0070] Example 2

[0071] See Figure 12 Embodiment 2 of the present invention also provides a building cluster cloud-edge collaborative optimization control device, comprising: Building thermodynamic RC virtual energy storage model construction unit 001 is used to construct a building thermodynamic RC virtual energy storage model based on the characteristics of building envelope and the physical laws of heat transfer. Model Predictive Control Algorithm (MPC) Construction Unit 002 is used to construct the Model Predictive Control Algorithm (MPC) based on the building thermal dynamic characteristics output by the building thermodynamic RC virtual energy storage model. The alternating direction multiplier method distributed optimization model construction unit 003 is used to construct an alternating direction multiplier method distributed optimization model based on the adjustable potential parameters output by the building thermodynamic RC virtual energy storage model and the control constraint boundary of the model predictive control algorithm MPC; the alternating direction multiplier method distributed optimization model includes a user-side sub-problem optimization model and an aggregation quotient-side sub-problem optimization model; The alternating direction multiplier method distributed optimization model initialization unit 004 is used to initialize the dummy variables, dual variables, and iteration number of the alternating direction multiplier method distributed optimization model; The adjustable potential calculation unit 005 is used by the user-side edge control module to collect real-time building operation data through sensors; based on the real-time building operation data, the current adjustable potential of the building is calculated through the building thermodynamic RC virtual energy storage model. The user response data acquisition and upload unit 006 is used to solve the user-side sub-problem optimization model based on the current adjustable potential and the current incentive price, obtain user response data, and upload the user response data to the cloud. The differentiated incentive price acquisition and distribution unit 007 is used by the cloud processing module to solve the optimization model of the aggregator-side sub-problem based on the user response volume data and microgrid operation constraints, obtain the differentiated incentive price, and distribute the differentiated incentive price to the user side. The variable update unit 008 is used to update the dummy variables and dual variables of the alternating direction multiplier method distributed optimization model based on the user response data and the differentiated incentive price, so as to obtain the updated variables; The model convergence judgment and processing unit 009 is used to calculate the residual and dual residual based on the updated variables; based on the residual and the dual residual, perform convergence judgment on the alternating direction multiplier method distributed optimization model; if converged, output the optimal user response amount and the optimal differentiated incentive price; if not converged, perform iterative optimization on the user-side subproblem optimization model and the aggregation quotient-side subproblem optimization model until the alternating direction multiplier method distributed optimization model converges. The optimal control strategy acquisition unit 010 is used to generate the optimal control strategy based on the optimal user response volume, the optimal differentiated incentive price, and disturbance prediction data of a future set duration, through the model predictive control algorithm MPC. The rolling optimization control unit 011 is used to move the control time domain forward by one cycle, re-collect the real-time operating data of the building, and perform the next round of optimization to achieve rolling optimization control.

[0072] In this embodiment, the process of constructing the building thermodynamic RC virtual energy storage model in the building thermodynamic RC virtual energy storage model construction unit 001 is as follows: Building nodes are divided using a thermal resistance-thermal capacity network strategy; the building nodes include wall nodes and indoor air nodes. Based on the aforementioned building nodes, a node thermal balance relationship is constructed; Based on the nodal thermal balance relationship and combined with building characteristic data, the thermal resistance and heat capacity parameters are determined by the least squares method. Based on the thermal resistance and thermal capacity parameters, the building's thermal inertia is transformed into "virtual battery" characteristics, and the building's energy storage status and vertical adjustment capacity are quantitatively output to complete the model construction.

[0073] In this embodiment, the process of constructing the Model Predictive Control Algorithm (MPC) in the MPC construction unit 002 is as follows: The control period, prediction time domain, and data sampling interval are set to obtain the time parameters; Based on the time parameters, an objective function is constructed that includes temperature deviation and equipment energy consumption; Based on the objective function and the upper and lower regulation capacities output by the building thermodynamic RC virtual energy storage model, constraints such as temperature comfort range and equipment power limit are set, the optimization feasible region is delineated, and the construction is completed.

[0074] In this embodiment, in the alternating direction multiplier method distributed optimization model construction unit 003, the user-side sub-problem optimization model aims to minimize the energy cost of a single building; the expression of the objective function is: ; In the formula, N is the total number of heating areas; n is the number of heating areas; T is the number of daily control cycles; and t is the number of daily control operations. For time intervals; The electricity price for users; The power purchased from the microgrid by a single heating zone; The constraints of the user-side sub-problem optimization model include: thermal balance constraints, indoor temperature comfort range constraints, HVAC system constraints, real-time electricity price constraints, and average electricity price constraints.

[0075] In this embodiment, in the alternating direction multiplier method distributed optimization model construction unit 003, the aggregation quotient sub-problem optimization model aims to maximize the microgrid operation benefits of the building cluster; the objective function expression is: ; In the formula, A represents the total number of building clusters; a represents the building cluster number. The electricity price for microgrids to supply power to users; The power output of the microgrid sold to a single heating area; The price at which the microgrid system purchases electricity from the distribution network; This refers to the power purchased by the microgrid from the distribution network. The economic cost imposed on building aggregators when a blockage occurs in the distribution network; The constraints of the optimization model for the aggregation business side problem include: microgrid power balance constraints, electricity purchase and sale price constraints, and distribution network node voltage constraints.

[0076] It should be noted that the information interaction and execution process between the modules of the above system are based on the same concept as the method embodiment in Embodiment 1 of this application, and the resulting technical effects are the same as those in the method embodiment of this application. For details, please refer to the description in the method embodiment shown above in this application, and it will not be repeated here.

[0077] Example 3

[0078] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium storing program code of a building cluster cloud-edge collaborative optimization control method. The program code includes instructions for executing the building cluster cloud-edge collaborative optimization control method of Embodiment 1 or any possible implementation thereof.

[0079] Computer-readable storage media can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

[0080] Example 4

[0081] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor; The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor can execute a building cluster cloud-edge collaborative optimization control method according to Embodiment 1 or any possible implementation thereof by calling the program instructions.

[0082] Specifically, a processor can be implemented in hardware or software. When implemented in hardware, the processor can be a logic circuit, an integrated circuit, etc. When implemented in software, the processor can be a general-purpose processor that reads software code stored in memory. This memory can be integrated into the processor or located outside the processor and exist independently.

[0083] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable system. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0084] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing systems. They can be centralized on a single computing system or distributed across a network of multiple computing systems. Optionally, they can be implemented using program code executable by a computing system, thereby storing them in a storage system for execution by the computing system. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0085] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A cloud-edge collaborative optimization control method for building clusters, characterized in that, include: Based on the characteristics of building envelope and the physical laws of heat transfer, a building thermodynamic RC virtual energy storage model is constructed. Based on the building thermal dynamic characteristics output by the building thermodynamic RC virtual energy storage model, a model predictive control algorithm (MPC) is constructed. Based on the adjustable potential parameters output by the building thermodynamic RC virtual energy storage model and the control constraint boundary of the model predictive control algorithm MPC, an alternating direction multiplier method distributed optimization model is constructed; the alternating direction multiplier method distributed optimization model includes a user-side sub-problem optimization model and an aggregation quotient-side sub-problem optimization model; The dummy variables, dual variables, and iteration count of the alternating direction multiplier method distributed optimization model are initialized. The user-side edge control module collects real-time building operation data through sensors; based on the real-time building operation data, it calculates the current adjustable potential of the building through the building thermodynamic RC virtual energy storage model. Based on the current adjustable potential and the current incentive price, the optimization model for the user-side sub-problem is solved to obtain user response data, and the user response data is uploaded to the cloud. Based on the user response volume data and microgrid operation constraints, the cloud processing module solves the optimization model of the aggregator-side sub-problem to obtain the differentiated incentive price, and then sends the differentiated incentive price to the user side. The dummy variables and dual variables of the alternating direction multiplier method distributed optimization model are updated based on the user response volume data and the differentiated incentive price to obtain the updated variables; Based on the updated variables, the residuals and dual residuals are calculated; based on the residuals and dual residuals, the convergence of the alternating direction multiplier method distributed optimization model is determined; if convergence is achieved, the optimal user response quantity and the optimal differentiated incentive price are output. If convergence is not achieved, the user-side subproblem optimization model and the aggregation quotient-side subproblem optimization model are iteratively optimized until the alternating direction multiplier method distributed optimization model converges. Based on the optimal user response volume, the optimal differentiated incentive price, and the disturbance prediction data for a future set duration, the optimal control strategy is generated through the Model Predictive Control (MPC) algorithm. By shifting the control time domain forward by one cycle, the real-time operating data of the building is re-acquired for the next round of optimization, thus achieving rolling optimization control.

2. The cloud-edge collaborative optimization control method for building clusters according to claim 1, characterized in that, The process of constructing the building thermodynamic RC virtual energy storage model is as follows: Building nodes are divided using a thermal resistance-thermal capacity network strategy; the building nodes include wall nodes and indoor air nodes. Based on the aforementioned building nodes, a node thermal balance relationship is constructed; Based on the nodal thermal balance relationship and combined with building characteristic data, the thermal resistance and heat capacity parameters are determined by the least squares method. Based on the thermal resistance and thermal capacity parameters, the building's thermal inertia is transformed into "virtual battery" characteristics, and the building's energy storage status and vertical adjustment capacity are quantitatively output to complete the model construction.

3. The cloud-edge collaborative optimization control method for building clusters according to claim 2, characterized in that, The process of constructing the Model Predictive Control (MPC) algorithm is as follows: The control period, prediction time domain, and data sampling interval are set to obtain the time parameters; Based on the time parameters, an objective function is constructed that includes temperature deviation and equipment energy consumption; Based on the objective function and the upper and lower regulation capacities output by the building thermodynamic RC virtual energy storage model, constraints such as temperature comfort range and equipment power limit are set, the optimization feasible region is delineated, and the construction is completed.

4. The cloud-edge collaborative optimization control method for building clusters according to claim 3, characterized in that, The user-side sub-problem optimization model aims to minimize the energy cost of a single building; the objective function is expressed as follows: ； In the formula, N is the total number of heating zones; n is the number of heating zones; T represents the number of intraday adjustment cycles; t represents the number of intraday adjustments. For time intervals; The electricity price for users; The power purchased from the microgrid by a single heating zone; The constraints of the user-side sub-problem optimization model include: thermal balance constraints, indoor temperature comfort range constraints, HVAC system constraints, real-time electricity price constraints, and average electricity price constraints.

5. The cloud-edge collaborative optimization control method for building clusters according to claim 4, characterized in that, The optimization model for the aggregator side subproblem aims to maximize the microgrid operation benefits of the building cluster; the objective function expression is: ； In the formula, A represents the total number of building clusters; a represents the building cluster number. The electricity price for microgrids to supply power to users; The power output of the microgrid sold to a single heating area; This refers to the price at which the microgrid system purchases electricity from the distribution network. This refers to the power purchased by the microgrid from the distribution network. The economic cost imposed on building aggregators when a blockage occurs in the distribution network; The constraints of the optimization model for the aggregation business side problem include: microgrid power balance constraints, electricity purchase and sale price constraints, and distribution network node voltage constraints.

6. A building cluster cloud-edge collaborative optimization control device, employing the building cluster cloud-edge collaborative optimization control method according to any one of claims 1-5, characterized in that, include: The building thermodynamic RC virtual energy storage model construction unit is used to construct a building thermodynamic RC virtual energy storage model based on the characteristics of the building envelope and the physical laws of heat transfer. The Model Predictive Control (MPC) algorithm construction unit is used to construct the Model Predictive Control (MPC) algorithm based on the building thermal dynamic characteristics output by the building thermodynamic RC virtual energy storage model. The alternating direction multiplier method distributed optimization model construction unit is used to construct an alternating direction multiplier method distributed optimization model based on the adjustable potential parameters output by the building thermodynamic RC virtual energy storage model and the control constraint boundary of the model predictive control algorithm MPC; the alternating direction multiplier method distributed optimization model includes a user-side sub-problem optimization model and an aggregation quotient-side sub-problem optimization model; An initialization unit for the distributed optimization model of the alternating direction multiplier method is used to initialize the dummy variables, dual variables, and iteration number of the distributed optimization model of the alternating direction multiplier method. The current adjustable potential calculation unit is used by the user-side edge control module to collect real-time building operation data through sensors; based on the real-time building operation data, the current adjustable potential of the building is calculated through the building thermodynamic RC virtual energy storage model. The user response data acquisition and uploading unit is used to solve the user-side sub-problem optimization model based on the current adjustable potential and the current incentive price, obtain user response data, and upload the user response data to the cloud. The differentiated incentive price acquisition and distribution unit is used by the cloud processing module to solve the optimization model of the aggregator-side sub-problem based on the user response volume data and microgrid operation constraints, obtain the differentiated incentive price, and distribute the differentiated incentive price to the user side. The variable update unit is used to update the dummy variables and dual variables of the alternating direction multiplier method distributed optimization model based on the user response volume data and the differentiated incentive price, so as to obtain the updated variables; The model convergence judgment and processing unit is used to calculate the residual and dual residual based on the updated variables; to perform convergence judgment on the alternating direction multiplier method distributed optimization model based on the residual and the dual residual; if convergence is achieved, the optimal user response quantity and the optimal differentiated incentive price are output. If convergence is not achieved, the user-side subproblem optimization model and the aggregation quotient-side subproblem optimization model are iteratively optimized until the alternating direction multiplier method distributed optimization model converges. The optimal control strategy acquisition unit is used to generate the optimal control strategy based on the optimal user response volume, the optimal differentiated incentive price, and disturbance prediction data for a future set duration, through the Model Predictive Control (MPC) algorithm. The rolling optimization control unit is used to move the control time domain forward by one cycle, re-collect the real-time operating data of the building, and perform the next round of optimization to achieve rolling optimization control.

7. The building cluster cloud-edge collaborative optimization control device according to claim 6, characterized in that, In the building thermodynamic RC virtual energy storage model construction unit, the process of constructing the building thermodynamic RC virtual energy storage model is as follows: Building nodes are divided using a thermal resistance-thermal capacity network strategy; the building nodes include wall nodes and indoor air nodes. Based on the aforementioned building nodes, a node thermal balance relationship is constructed; Based on the nodal thermal balance relationship and combined with building characteristic data, the thermal resistance and heat capacity parameters are determined by the least squares method. Based on the thermal resistance and thermal capacity parameters, the building's thermal inertia is transformed into "virtual battery" characteristics, and the building's energy storage status and vertical adjustment capacity are quantitatively output to complete the model construction.

8. The building cluster cloud-edge collaborative optimization control device according to claim 7, characterized in that, In the Model Predictive Control (MPC) algorithm construction unit, the process of constructing the MPC algorithm is as follows: The control period, prediction time domain, and data sampling interval are set to obtain the time parameters; Based on the time parameters, an objective function is constructed that includes temperature deviation and equipment energy consumption; Based on the objective function and the upper and lower regulation capacities output by the building thermodynamic RC virtual energy storage model, constraints such as temperature comfort range and equipment power limit are set, the optimization feasible region is delineated, and the construction is completed.

9. A building cluster cloud-edge collaborative optimization control device according to claim 8, characterized in that, In the alternating direction multiplier method distributed optimization model construction unit, the user-side sub-problem optimization model aims to minimize the energy cost of a single building; the expression of the objective function is: ； In the formula, N is the total number of heating zones; n is the number of heating zones; T represents the number of intraday adjustment cycles; t represents the number of intraday adjustments. For time intervals; The electricity price for users; The power purchased from the microgrid by a single heating zone; The constraints of the user-side sub-problem optimization model include: thermal balance constraints, indoor temperature comfort range constraints, HVAC system constraints, real-time electricity price constraints, and average electricity price constraints. In the alternating direction multiplier method distributed optimization model construction unit, the aggregation quotient sub-problem optimization model aims to maximize the microgrid operation benefits of the building cluster; the objective function expression is: ； In the formula, A represents the total number of building clusters; a represents the building cluster number. The electricity price for microgrids to supply power to users; The power output of the microgrid sold to a single heating area; This refers to the price at which the microgrid system purchases electricity from the distribution network. This refers to the power purchased by the microgrid from the distribution network. The economic cost imposed on building aggregators when a blockage occurs in the distribution network; The constraints of the optimization model for the aggregation business side problem include: microgrid power balance constraints, electricity purchase and sale price constraints, and distribution network node voltage constraints.

10. An application architecture for a cloud-edge collaborative optimization control method for building clusters, characterized in that, The architecture adopts a layered design consisting of a perception layer, a control execution layer, a communication layer, and an algorithm layer. The hardware integration of the perception layer is as follows: based on edge computing devices, it is equipped with infrared sensing devices, light intensity detection devices, and temperature and humidity acquisition devices for monitoring the building environment, so as to realize the comprehensive acquisition of real-time data on indoor and outdoor temperature, light intensity, and environmental humidity of the building. The control execution layer is designed as follows: a two-way communication and control link is established through an infrared signal transmitting component and an infrared signal receiving component to realize remote control of the building's air conditioning system; the infrared receiving component converts the infrared remote control signal with the set protocol into a digital pulse signal for edge computing devices to parse and store; the infrared transmitting component outputs a pulse signal with a specific frequency and duty cycle according to the parsed control command, which is modulated and sent to the air conditioning equipment to dynamically adjust the air conditioning operating parameters. The architecture and security mechanism of the communication layer are as follows: a low-power long-distance radio frequency communication module is used to build a communication link, combined with a data encryption algorithm; the communication module is adapted to the communication needs of multi-node, distributed building environments; the communication module is connected to the edge computing core device and the cloud processing device through a serial port interface to realize bidirectional data interaction between the edge side and the cloud. The software module integration and collaboration process of the algorithm layer is as follows: The edge-side embedded software has the functions of periodically collecting data from the perception layer, preprocessing the collected data, performing model prediction based on the data, and generating control commands; by calling the optimization solver to execute the model prediction control algorithm, combined with the building thermal dynamics model and real-time environmental data, it generates the optimal air conditioning operation command, and uses an infrared protocol parsing tool to encode and send infrared control signals; The cloud software platform integrates a distributed optimization algorithm module, a data storage system, and a network service interface, performs distributed optimization operations based on the optimization solver, receives power demand data uploaded from the edge side, performs global coordination in combination with the microgrid operation model, generates differentiated electricity price signals and sends them to the edge side, and saves historical operation data through the data storage system; The hardware platform setup process is as follows: taking the edge computing core device as the center, the selection and assembly of various acquisition devices in the perception layer, the deployment and functional debugging of the communication module in the communication layer, and the connection and operation verification of the infrared control component in the control execution layer are completed in sequence. The communication protocol and edge-cloud data interaction mechanism are as follows: the system communication protocol is formulated based on a custom instruction set, and an radio frequency communication link is built based on a long-distance radio frequency communication module to meet the data interaction needs in edge-cloud collaborative scenarios.

Images