A building energy consumption dynamic simulation and optimization system based on digital twinning

Abstract

The application discloses a kind of based on digital twinning building energy consumption dynamic simulation and optimization system, it is related to building energy-saving control technical field.The system uses the distributed architecture of edge and cloud cooperation, including deployment in each subzone of building edge optimizer and cloud center coordinator.Edge optimizer executes local fast control, its local optimization target is integrated by the associated influence matrix issued by center to consider global influence, and its constraint is limited by the global load boundary issued by center.Center coordinator maintains building level digital twinning model, and issues by parallel simulation dynamic calculation associated influence matrix and global load boundary, drives distributed collaborative optimization.The application solves the problems of large response delay, optimization lag and poor adaptability of centralized system, realizes the unity of local fast response and global dynamic optimization, and improves building energy efficiency and operation resilience.

Description

Technical Field

[0001] This invention relates to the field of building energy conservation control technology, specifically to a building energy consumption dynamic simulation and optimization control system that combines digital twin and distributed collaborative optimization technologies. Background Technology

[0002] Digital twin technology provides core support for energy consumption simulation, analysis, and optimization by constructing a virtual mapping of a physical building. Currently, the mainstream application model is a centralized digital twin system: a sensor network is widely deployed in the building to upload collected data such as temperature, humidity, and energy consumption to a central server in real time; the server integrates building information modeling and physical simulation engine to run a unified, high-fidelity digital twin model for energy consumption simulation, fault diagnosis, and optimization calculations; finally, the calculated optimized control strategies (such as equipment setpoints) are distributed to the building automation system for execution.

[0003] However, this centralized architecture faces a series of challenges in practical engineering applications. First, the system's real-time response is insufficient. The entire link from data sensing, remote transmission, central processing to command feedback is long, introducing significant latency (often several minutes). This makes it difficult to meet the demand for rapid response within seconds or minutes to sudden local disturbances such as a surge in conference room occupants, equipment start-up and shutdown, and short-term changes in solar radiation, leading to a decline in local environmental control quality and energy waste. Second, global optimization calculations are lagging and lack flexibility. High-precision simulation and global optimization calculations for the entire building are computationally intensive and time-consuming, making high-frequency execution impossible (usually measured in hours). The optimization strategy update cycle is much slower than the dynamic changes in the building's internal and external states, resulting in control strategies often based on outdated information and ineffective optimization. Third, the system is highly dependent on models and lacks adaptability. The optimization effect heavily relies on the accuracy of the digital twin model. As building equipment ages, the performance of the building envelope changes, and the functions and usage patterns of indoor spaces change, model mismatches inevitably occur. Existing systems typically lack effective online calibration mechanisms, leading to increased simulation prediction bias and a continuous decline in optimization efficiency over long-term operation. Fourth, the system has a weak ability to coordinate and respond to emergencies. When faced with sudden demands such as power grid demand response commands or extreme operating conditions inside buildings, the centralized system struggles to quickly coordinate the load adjustment or operation mode switching of various terminal devices, resulting in an insufficiently agile and precise response.

[0004] Therefore, existing technologies need a technical solution that can effectively address issues such as response delay, optimization lag, model mismatch, and insufficient coordination, in order to achieve more real-time, accurate, and adaptive building energy consumption management. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a building energy consumption dynamic simulation and optimization system based on digital twins to address the problems mentioned in the background art. Specifically, it overcomes the shortcomings of existing centralized digital twin systems in terms of real-time response, dynamic optimization, and adaptability to uncertain operating conditions.

[0006] To address the aforementioned technical problems, this invention provides a building energy consumption dynamic simulation and optimization system based on digital twins, including edge optimizers deployed in multiple physical zones of a building and a central coordinator deployed in the cloud.

[0007] The edge optimizer is used to perform the following operations in the local control loop: Collect environmental status data for this partition; based on the data, the internal prediction model, and the coordination parameters received from the central coordinator, solve a local optimization problem to obtain the optimal setpoint sequence for the controlled devices in this partition within a finite future time period and execute the current instruction; The collaborative parameters include an association influence matrix and a global load boundary. The optimization objective of the local optimization problem includes an association cost term calculated from the association influence matrix, and its constraints include load constraints determined by the global load boundary. Upload information containing the status and control summary of this partition to the central coordinator.

[0008] The central coordinator is used to perform the following operations in the global coordination loop: Maintain an architectural-level digital twin model consisting of multiple coupled sub-models, with each sub-model corresponding to a physical partition; Receive information uploaded by all edge optimizers; based on the digital twin model and the received information, perform parallel simulation analysis to calculate the correlation influence matrix describing the influence intensity of any partition control behavior on the state of other partitions, and the global load boundary for coordinating the operating state of each partition to enable the overall building energy supply system to operate in the high-efficiency zone; The calculated correlation influence matrix and the global load boundary are sent as collaborative parameters to the corresponding edge optimizer.

[0009] Furthermore, the associated cost term is obtained by performing a quadratic operation on the candidate setpoint sequence vector and the associated influence matrix.

[0010] Furthermore, the parallel simulation analysis includes: performing simulation based on the planned control input reported by each edge optimizer to obtain the system state baseline trajectory; for each partition, performing simulation after applying a preset test disturbance to the baseline input to obtain the system state response trajectory under the disturbance; and calculating the correlation influence matrix by analyzing the difference between the state response trajectory and the state baseline trajectory.

[0011] Furthermore, calculating the global load boundary includes: determining the efficient operating load range of the building's central heating and cooling source system based on the building-level digital twin model; optimizing the solution using the planned load of each zone in the future collaborative cycle as the decision variable, so that the total building load falls into the efficient operating load range, and determining the specific load boundary of each zone based on the solution results.

[0012] Furthermore, the central coordinator also executes a model parameter update process, including: Based on the historical operation data sequence of building equipment, the coupling relationship parameters between the sub-models in the digital twin model are updated online using a parameter identification algorithm; Based on the deviation between the actual running data and the corresponding simulation prediction data fed back by the edge optimizer, when the deviation continues to exceed the threshold, a calibration operation on the internal parameters of the relevant sub-model is triggered and executed.

[0013] Furthermore, the parameter identification algorithm is a recursive least squares method with a forgetting factor; the calibration operation employs an iterative optimization algorithm aimed at minimizing prediction bias.

[0014] Furthermore, the system also includes an emergency response mechanism. When the central coordinator determines an internal emergency based on information uploaded by the edge optimizers, or receives an external scheduling instruction, it interrupts the current global coordination loop and initiates an emergency coordination process. This emergency coordination process uses a shortened simulation prediction time domain and an optimization objective tailored to the event target to recalculate emergency coordination parameters and prioritize their distribution to the relevant edge optimizers.

[0015] Furthermore, the criteria for determining internal emergencies include: any edge optimizer reporting that the rate of change of the environmental state of its partition exceeds a preset threshold.

[0016] Furthermore, when responding to external dispatch instructions, the global load boundary calculated by the emergency coordination process is used to ensure that the total building load meets the adjustment target required by the instruction.

[0017] Furthermore, the edge optimizer executes the local control loop for a first period, and the central coordinator executes the global coordination loop for a second period, the second period being an integer multiple of the first period.

[0018] Compared with the prior art, the technical solution provided by the present invention has the following beneficial effects: 1. Achieved a synergistic improvement in control response speed and global optimization depth: By delegating local control decisions with high real-time requirements to the edge optimizer and entrusting complex global coupling analysis and strategy formulation to the central coordinator for asynchronous execution, the system can quickly respond to local disturbances on a second / minute time scale, while continuously performing global energy efficiency optimization on a minute / hour time scale, thus solving the contradiction that centralized systems cannot balance real-time performance and global performance.

[0019] 2. Enhanced dynamic adaptability and foresight of optimization strategies: The central coordinator, through periodic parallel simulations, can quickly quantify and evaluate the dynamic coupling relationships between different building zones, and proactively solve for the globally optimal load allocation based on the updated digital twin model. This allows the collaborative parameters issued to the edge optimizers to reflect the system state in real time, guiding local control behavior to dynamically approach the global optimum, avoiding the strategy rigidity problem caused by slow model updates and low optimization frequency in traditional methods.

[0020] 3. Improved accuracy and reliability of long-term system operation: By integrating online parameter identification based on historical data and model calibration mechanisms based on prediction error feedback, the digital twin model can continuously evolve and maintain consistency with the physical characteristics of real buildings. This self-learning capability reduces reliance on accurate initial models, effectively mitigating model mismatch caused by factors such as equipment aging and changes in usage patterns, and ensuring the long-term stability of optimization results.

[0021] 4. Enhanced system's ability to cope with internal and external uncertainties: The system's embedded emergency response mechanism provides a layered and collaborative flexible control paradigm. In the face of sudden high-load events, emergency resource coordination can be achieved by rapidly adjusting local constraints and coupling penalties; in the face of external commands such as grid demand response, proactive load coordination scheduling can be performed. This enhances the robustness of the building energy system and its ability to participate in grid interaction as a flexible resource.

[0022] 5. Improved system scalability and engineering practicality: The distributed architecture reduces the instantaneous peak pressure on the central node's computing and communication, facilitating modular deployment and expansion in large building complexes. The clear functional boundaries and standardized data interfaces between the edge optimizer and the central coordinator also reduce the complexity of system integration, debugging, and maintenance. Attached Figure Description

[0023] Figure 1 This is a flowchart of the calculation process for the central coordinator's collaborative parameters according to the present invention. Figure 2The branch structure diagram for updating the parameters of the digital twin model of this invention; Figure 3 This is a branch structure diagram of the system emergency collaborative response mechanism of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] As attached Figures 1 to 3 The system illustrates a digital twin-based dynamic simulation and optimization system for building energy consumption. This system includes edge optimizers deployed in various physical zones within the building, and a central coordinator deployed on a remote server. The edge optimizers perform rapid local environmental control, while the central coordinator performs global simulation and collaborative optimization. Through data interaction, they achieve dynamic management of the building's overall energy consumption.

[0026] Example 1: System operation under normal operating conditions This embodiment details the standard workflow and data interaction of the system when there are no external unexpected interferences.

[0027] 1. System Initialization Edge optimizers are deployed in independent control areas of a building, such as individual rooms or open-plan office spaces. Each edge optimizer establishes a data connection with the temperature sensors and air conditioning terminal actuators in its respective zone.

[0028] The edge optimizer internally sets the local control cycle. , The value of is between to Between. The edge optimizer's built-in predictive model is used to predict future environmental states based on control commands. This model's parameters have initial values ​​and can be updated during operation.

[0029] The central coordinator loads building information model (BIM) and equipment system data to construct a building-level digital twin simulation model. This model can simulate heat transfer in the building envelope, indoor airflow, and the hydraulic and thermal processes of the HVAC system.

[0030] The model is divided into multiple coupled sub-models based on the physical topology of the air conditioning supply loop and water supply network. Each sub-model corresponds to a physical partition controlled by the edge optimizer.

[0031] The coupling relationships between sub-models are defined by parameters such as thermal conductivity, air exchange rate, and hydraulic correlation coefficient. The central coordinator sets the global coordination cycle. , yes Multiples of , whose range can be . of Up to 10 times.

[0032] 2. Periodic control of the edge optimizer In each length of During the control cycle, each edge optimizer independently performs the following operations: (1) Data acquisition: Read the indoor temperature of this zone Real-time power of air conditioning terminal equipment .

[0033] (2) Local optimization problem construction and solution: based on the future One control step ( greater than In a preferred embodiment, the integer is... Corresponding to the future Using the air conditioning setpoint sequence as the decision variable, construct the objective function and constraints, and solve for the optimal sequence.

[0034] The objective function is to minimize a compound cost. The calculation formula is as follows: ; in, and The future number is calculated based on the internal prediction model of the edge optimizer and the candidate setpoint sequence. Energy consumption and room temperature prediction for each step. Set the desired indoor temperature value, for example The set value It is usually set within the comfort range, for example, the midpoint of the comfort range can be taken. , , The preset non-negative weighting coefficients satisfy... A set of examples takes the value . This is a related cost item.

[0035] The related cost item The calculation formula is: ; in, for The candidate setpoint sequence vector. It is The real symmetric matrix, called the local association influence matrix, is calculated and distributed by the central coordinator. The elements of this matrix... Quantified in the first The control action of the system in the first control step affects the system at the... The overall influence intensity of the control step size on the overall state.

[0036] The constraints of the optimization problem include: physical limits of the device actuator; and predicted room temperature. It needs to be maintained within the preset comfort range, for example... Predicted average power The global load boundary issued by the central coordinator must be met. .

[0037] The edge optimizer calls the built-in optimization solver to solve the above problem and obtain the optimal setpoint sequence. The optimization solver can be a quadratic programming solver, whose input is the composite cost function. The aforementioned related costs The calculation relationships and all constraints are used to output the optimal sequence. .

[0038] (3) Control instruction execution: Execute the optimal sequence The first element in the setpoint, which should be executed immediately in the current cycle, is sent to the air conditioning terminal device.

[0039] (4) Upload of operational information: Generate a data package containing current period measurement data and a summary of future forecasts, and upload it to the central coordinator. The forecast summary includes data based on... The future of computing Average forecast load within the period .

[0040] 3. Global collaborative computation of the central coordinator The central coordinator operates on a cyclical basis. Perform the following steps: (1) State synchronization: Receive and integrate all information uploaded by edge optimizers, and update the current state and plan control input of each sub-model in the digital twin model.

[0041] (2) Parallel simulation and coupling analysis: Run a set of parallel digital twin simulations to quantify the dynamic effects between partitions.

[0042] First, using the planned control inputs reported by each edge optimizer as a baseline, a simulation is performed to obtain the baseline trajectory of each partition's state (such as temperature and humidity). ,in This indicates the future simulation time. Subsequently, for each partition... Apply a preset test perturbation to its reference input. (In a preferred embodiment, To increase its planned average load The simulation was run again to obtain the state trajectory after the disturbance. .

[0043] By calculating the state deviation Extract partition The disturbance affects all other partitions Influence intensity coefficient of critical conditions (such as temperature) .

[0044] In a preferred embodiment, the influence intensity coefficient Calculated using the following formula: ,in yes Middle partition Temperature deviation component, This represents the simulation time domain length.

[0045] (3) Calculation of co-parameters: Based on the results of coupling analysis, two types of parameters are calculated.

[0046] The first type is the correlation influence matrix. All partitions ( The influence strength coefficient between each pair of partitions (total number of partitions) To standardize and form a Global correlation influence matrix Normalization can be used to treat all influence intensity coefficients. Divide by the largest absolute value of these coefficients. From Extract the corresponding partition Submatrix, as the partition .

[0047] The second category is the global load boundary. This involves using a digital twin model to simulate a central cooling / heating source system (such as a chiller unit) under different total loads. Operating efficiency curve Determine how to improve system efficiency. Higher than the set threshold (In a preferred embodiment, ,in Total load operating range (for peak efficiency) .

[0048] In the future The period is the optimization time period, which is the average planned load of each zone within that time period. ( Given the decision variables, construct a system-level optimization problem with the objective of maximizing the total building load. fall into Within the range, the basic needs of each sub-region are also taken into account.

[0049] Solve this problem to obtain the optimal load recommendations for each zone. .

[0050] Based on this value, set an allowable fluctuation range. (For example Desirable ), forming the global load boundary for each partition. ,in , .

[0051] (4) Parameter distribution: The calculated parameters for each partition are distributed. and It is then sent to the corresponding edge optimizer.

[0052] 4. Continuous learning and updating of model parameters To improve the long-term accuracy of the digital twin model, the system executes a background learning process: (1) Online identification of coupling relationship: The central coordinator continuously collects historical operating data of the building equipment monitoring system, such as the water flow of each air conditioning terminal. Air volume and the corresponding return air temperature ,in For historical timestamps.

[0053] For any two partitions and To identify the partition For partitions To determine the thermal effects, the following first-order linear model is established for online identification: ,in For discrete-time indexing, It is noise.

[0054] Using a forgetting factor ( The recursive least squares method (with a preferred value of 0.98) is used for online identification. The input of this algorithm is the historical time. and The sequence outputs the coupling parameters. The real-time estimate is used to update the corresponding coupling parameters in the digital twin model.

[0055] (2) Prediction error-driven calibration: After each global collaborative calculation, the central coordinator records the predicted load sequences of each partition for simulation. .

[0056] After a period of delay (In one implementation, Obtain the actual average power measurement value of each zone within the corresponding time period. .

[0057] Calculate the load forecasting error for each zone. .

[0058] If a specific partition mean prediction error In continuous One calibration check cycle (e.g.) All exceeded the set threshold. (For example Set as the rated power of this zone If this is the case, then automatic calibration of the sub-model parameters corresponding to that partition will be triggered.

[0059] The calibration process involves adjusting the key parameter set of the sub-model. (such as building envelope heat capacity, indoor heat generation coefficient), to minimize a recent historical window (such as the past) The sum of squares of the load forecasting errors for that zone The objective is solved through iterative optimization.

[0060] The iterative optimization employs the gradient descent method, where the required gradient... Obtained through numerical perturbation (finite difference method), that is, sequentially... A small perturbation is applied to each parameter (the step size of the perturbation can be selected based on the physical meaning and typical order of magnitude of the parameter, for example, set to a small proportion of the current value of the parameter). to (or a fixed, small absolute value), the model is rerun to calculate the change in prediction error, thereby estimating the gradient. The optimization process continues until the error meets the requirements or the maximum number of iterations is reached.

[0061] Example 2: Response Procedure for Internal Emergencies This embodiment illustrates the system's response mechanism when an unforeseen surge in load occurs inside a building.

[0062] (1) Event detection: The edge optimizer in a conference room monitors the indoor temperature in real time. When detected In two consecutive rate of increase within the cycle Exceeding the preset threshold (In a preferred embodiment, ), and in the two consecutive If the personnel presence sensor detects an increase of more than 10 people indoors within a given period, the edge optimizer determines this as a sudden high-load event. The edge optimizer immediately appends an emergency event flag to the data packets uploaded in the current period, including the current... , And personnel change data.

[0063] (2) Emergency Coordination Trigger: Upon receiving a data packet containing an emergency flag, the central coordinator suspends any currently running regular collaborative computing threads and immediately starts a high-priority emergency collaborative computing thread. The emergency thread uses the latest reported status of all partitions as its initial condition and shortens the simulation prediction time domain to [a specific timeframe]. (Preferred) to ).

[0064] The optimization objective has been temporarily adjusted to: ensure that the event occurs in the partition (denoted as partition). The temperature in the future Within not exceeding the acceptable emergency limit (like Under the premise of minimizing the peak total energy consumption of the entire building's air conditioning system in the short term in the future.

[0065] (3) Rapid calculation and instruction issuance: Based on the adjusted time domain and target, the emergency thread quickly executes a simplified global collaborative calculation. By using shorter simulation steps in emergency simulations (e.g., compared to regular simulations) The step size is shortened to The coupling effect matrix is ​​recalculated. A set of emergency coordination parameters is generated, which includes at least: For event partitioning Issue a temporarily relaxed comfort constraint range, for example... And issue the correlation and influence matrix of the enhanced penalty. .

[0066] For partitions In the simulation, one or more adjacent partitions are shown as strongly coupled: a temporary reduction in the load ceiling is issued. For example, adjust to the original of To prioritize the protection of partitions The supply of cooling capacity.

[0067] (4) Execution and Recovery: Upon receiving the emergency coordination parameters in the next communication cycle, the relevant edge optimizers immediately use them to overwrite the original regular parameters for local optimization and control. The central coordinator continuously monitors the partitions. Temperature feedback.

[0068] When partition temperature Persistently below achieve (like ), and its temperature change rate Below When the event is over, the central coordinator determines that the event has ended and broadcasts a notification to all edge optimizers to clear the emergency parameters and resume using the latest issued regular coordination parameters.

[0069] Example 3: Flowchart for responding to external scheduling instructions This example illustrates how the system responds to demand response commands from the power grid or a higher-level energy management system.

[0070] (1) Command Reception and Parsing: The central coordinator receives external commands through a standard communication interface (such as a RESTful API). The command format includes key parameters, such as: start time. Duration The target load reduction value to be achieved (Unit is) ).

[0071] The central coordinator parses and verifies the instructions. Check if the time difference with the current time is sufficient to complete the scheduling preparation. If the time is insufficient, return a non-responsive message to the superior system; if the time is sufficient, continue executing the subsequent scheduling process.

[0072] (2) Proactive collaborative scheduling: In Previously, the central coordinator initiated a response-oriented period. This involves a special global collaborative calculation. The optimization objective function for this calculation incorporates a strong constraint term or a high-weight penalty term regarding the total load to ensure that the total predicted load of the building is within the response period. satisfy ,in This represents the predicted base load for this period. Meanwhile, optimization still needs to minimize the overall deviation in comfort levels across different zones.

[0073] (3) Task decomposition and pre-execution: By solving the above optimization problem, the optimal load setpoint for each partition during the response period is obtained. The central coordinator then calculates the load adjustment amount that each zone needs to handle. ,in This represents the predicted load under normal circumstances.

[0074] Subsequently, the updated load boundary will be... (generally The reduced (and potentially fine-tuned) correlation matrix is ​​distributed in advance to all edge optimizers. During local optimization, the edge optimizers utilize the new boundary constraints to pre-adjust and smoothly adjust their control setpoint sequences, ensuring that... At any given moment, the total building load can smoothly transition to the target level and throughout the entire process. The system remains stable during the response period. After the response period ends, the system automatically resumes its normal collaborative mechanism.

[0075] The above embodiments illustrate how the system works in different scenarios. It should be understood that the specific parameters, values, algorithm selections, and scenario descriptions in the above embodiments are merely examples to clearly illustrate the implementation process of the present invention and are not intended to limit the invention. Those skilled in the art can substitute or adjust the above parameters, values, and algorithms according to actual building characteristics, equipment performance, and control requirements, without departing from the core concept of the present invention. All such adjustments should be considered within the scope of protection claimed by the present invention.

[0076] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A building energy consumption dynamic simulation and optimization system based on digital twins, characterized in that, include: Multiple edge optimizers, deployed in different physical zones of the building, are used to perform local control. as well as The central coordinator, deployed in the cloud, is used to perform global coordination. The edge optimizer is configured to perform the following steps in a loop: S1. Collect environmental status data for this partition; S2. Based on the collected data, the internal prediction model, and the coordination parameters received from the central coordinator, solve a local optimization problem to obtain the optimal setpoint sequence of the controlled equipment in this partition within a finite future time period; wherein, the coordination parameters include an association influence matrix and a global load boundary, the optimization objective of the local optimization problem includes an association cost term determined by the association influence matrix, and the constraints of the local optimization problem include load constraints determined by the global load boundary; S3. Execute the current control command in the optimal setpoint sequence; S4. Upload information containing the status of this partition and a summary of the optimal setting value sequence to the central coordinator; The central coordinator is configured to perform the following steps in a loop: P1. Maintain an architectural-level digital twin model, which consists of multiple interdependent sub-models, each corresponding to a physical partition; P2. Receives all information uploaded by the edge optimizer; P3. Based on the building-level digital twin model and the received information, perform parallel simulation analysis to calculate the correlation influence matrix that describes the intensity of the influence of any zone control behavior on the state of other zones, and the global load boundary that coordinates the operating state of each zone. P4. The calculated correlation influence matrix and the global load boundary are sent to the corresponding edge optimizer as the collaborative parameters.

2. The building energy consumption dynamic simulation and optimization system based on digital twins according to claim 1, characterized in that, In step S2, the associated cost term is obtained by performing a quadratic operation on the candidate set value sequence vector and the associated influence matrix.

3. The building energy consumption dynamic simulation and optimization system based on digital twins according to claim 1, characterized in that, In step P3, the parallel simulation analysis includes: P31. Simulation is performed using the planned control input uploaded by each edge optimizer as a reference to obtain the system state reference trajectory; P32. For each partition, after applying a preset test perturbation to the reference input, perform simulation to obtain the state response trajectory of the system under the perturbation; P33. The correlation influence matrix is ​​calculated by analyzing the difference between the state response trajectory and the state reference trajectory.

4. The building energy consumption dynamic simulation and optimization system based on digital twins according to claim 1 or 3, characterized in that, In step P3, calculating the global load boundary includes: determining the efficient operating load range of the building energy supply system based on the building-level digital twin model; optimizing the planned load of each zone within the target time period as a variable so that the total building load falls into the efficient operating load range, and determining the load boundary of each zone based on the optimization results.

5. The building energy consumption dynamic simulation and optimization system based on digital twins according to claim 1, characterized in that, The central coordinator is also configured to execute a model update process, which includes: Based on the building's historical operational data, update the coupling relationship parameters between the sub-models in the building-level digital twin model; Based on the deviation between the actual running data fed back by the edge optimizer and the corresponding simulation prediction data, the calibration of the internal parameters of the relevant sub-model is triggered and executed.

6. The building energy consumption dynamic simulation and optimization system based on digital twins according to claim 5, characterized in that, The updated coupling parameters are obtained using a recursive least squares method with a forgetting factor; the internal parameters of the calibration correlation sub-model are obtained using an iterative optimization algorithm aimed at minimizing prediction bias.

7. The building energy consumption dynamic simulation and optimization system based on digital twins according to claim 1, characterized in that, The system also includes an emergency response mechanism; when the central coordinator determines that an event requiring urgent response has occurred, it interrupts the current global coordination loop and executes the emergency coordination process. The emergency coordination process includes: recalculating emergency coordination parameters by using a shortened simulation time domain and an optimization objective adapted to the event target, and then sending them to the relevant edge optimizers.

8. The building energy consumption dynamic simulation and optimization system based on digital twins according to claim 7, characterized in that, The events requiring urgent response include internal emergencies, which are determined based on the following criteria: any edge optimizer reports a rate of change in its partition environment state exceeding a preset threshold.

9. The building energy consumption dynamic simulation and optimization system based on digital twins according to claim 7, characterized in that, The events requiring emergency response include external dispatch instructions and the global load boundary calculated by the emergency coordination process, which is used to ensure that the total building load meets the requirements of the external dispatch instructions.

10. The building energy consumption dynamic simulation and optimization system based on digital twins according to claim 1, characterized in that, The local control cycle of the edge optimizer is a first cycle, and the global coordination cycle of the central coordinator is a second cycle, which is an integer multiple of the first cycle.

Images