Building energy consumption simulation and prediction method and system based on digital twinning

Abstract

The present application relates to the technical field of digital twinning, in particular to a building energy consumption simulation and prediction method and system based on digital twinning, comprising the following steps: identifying a step excitation starting time generated by a device state jump, calculating a thermal response lag time of a physical and virtual environment, correcting a thermal inertia parameter according to a time lag deviation to establish a dynamic calibration model, and generating an energy consumption trend prediction sequence in combination with a weather forecast. In the present application, the step time is determined by monitoring the operating state jump, the thermal response lag is calculated in combination with the measured temperature data, the deviation is obtained by comparing the lag time of the virtual environment, the thermal inertia parameter of the digital twinning model is iteratively updated using the deviation, a dynamic calibration mechanism is constructed to eliminate the influence of parameter misalignment, the thermal mechanical behavior of the virtual model is ensured to be highly consistent with the entity building, the predictive simulation is carried out according to the dynamic calibration model, and the scheduling plan and the weather forecast are imported to efficiently generate the building energy consumption trend prediction sequence.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and in particular to a method and system for simulating and predicting building energy consumption based on digital twins. Background Technology

[0002] The field of digital twin technology involves integrating multidisciplinary, multi-physical, multi-scale, and multi-probabilistic simulation processes using physical models, sensor updates, and operational data. This process maps these simulations in a virtual space, reflecting the entire lifecycle of the corresponding physical equipment. By constructing virtual copies of physical entities, this technology enables real-time monitoring, diagnosis, prediction, and optimized control, and is widely applied in industrial manufacturing, smart cities, and building management scenarios. Traditional building energy consumption simulation and prediction methods refer to the technical aspects of calculating and estimating the energy consumption of buildings during operation. These methods employ physical modeling, based on the building's geometric topology, thermal properties of the building envelope, and HVAC equipment parameters, combined with typical meteorological year data, using heat balance equations or transfer functions to calculate the building's heating and cooling loads and energy consumption. Alternatively, a data-driven approach is used, collecting historical meter readings and environmental temperature and humidity data, and constructing linear regression or time series models to fit the functional relationship between energy consumption and influencing factors to extrapolate future energy consumption values.

[0003] Traditional building energy consumption simulation relies on geometric topology and design thermal properties to construct physical models, or simply fits historical electricity meter readings and environmental data. This makes it difficult to reflect in real time the decay of the building envelope's thermal performance due to long-term operation. Calculations are based solely on typical meteorological year data and static heat balance equations, ignoring the hysteresis characteristics of the actual thermal response in the physical environment. This results in deviations between the thermal inertia parameters of the virtual model and the actual physical state, failing to accurately characterize the actual thermodynamic behavior of the building under dynamic operating conditions. Consequently, the accuracy of energy consumption prediction results is reduced, making it difficult to meet the needs of refined energy management. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method and system for building energy consumption simulation and prediction based on digital twins.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for simulating and predicting building energy consumption based on digital twins, comprising the following steps:

[0006] S1: Monitor the level status of the operation control circuit of the air conditioning terminal equipment in the physical building, identify the signal characteristics of the transition from non-operation to operation, identify the transition time, and generate the start time of the step excitation.

[0007] S2: Call the start time of the step excitation, collect indoor temperature data, calculate the temperature change rate of adjacent sampling points, filter the time when the temperature change rate exceeds the preset temperature response judgment standard, and generate the physical environment thermal response hysteresis time.

[0008] S3: Input the start time of the step excitation into the digital twin model to drive the virtual air conditioner to run, calculate the virtual temperature distribution and rate of change, filter the time when the rate of change exceeds the temperature response judgment standard, and obtain the virtual environment thermal response lag time.

[0009] S4: Calculate the deviation between the thermal response lag time of the physical environment and the thermal response lag time of the virtual environment. Based on the positive and negative polarity and magnitude of the deviation, iteratively correct the thermal inertia parameters in the digital twin model to establish a dynamically calibrated digital twin model.

[0010] S5: Using the dynamically calibrated digital twin model, import meteorological forecast data and personnel scheduling plans for future time periods to perform predictive simulation, calculate the dynamic heating and cooling load curves required to maintain the set environmental comfort level, and generate a building energy consumption trend prediction sequence.

[0011] As a further embodiment of the present invention, the step excitation start time includes a physical device start-up timestamp, the physical environment thermal response lag time includes the measured temperature fluctuation delay period, the virtual environment thermal response lag time includes the simulated temperature change delay period, the dynamically calibrated digital twin model includes the corrected building envelope thermal resistance and updated indoor air heat capacity parameters, and the building energy consumption trend prediction sequence includes future heating and cooling load prediction values ​​and expected power consumption time series data.

[0012] As a further aspect of the present invention, the step of obtaining the start time of the step excitation specifically includes:

[0013] S111: Monitor the operation control line level status of the air conditioning terminal equipment in the physical building. Use the analog-to-digital conversion interface to collect the line voltage at high frequency. Extract the voltage amplitude data at several moments according to the preset sampling frequency. Map and bind each voltage amplitude to the corresponding clock timestamp one by one. Construct a two-dimensional matrix including time dimension index and voltage amplitude data to generate the line level time sequence sampling set.

[0014] S112: Based on the line level timing sampling set, calculate the first-order difference value of the voltage amplitude corresponding to adjacent time steps in the two-dimensional matrix, construct the voltage change rate sequence, compare the difference value in the voltage change rate sequence with the preset operating state activation voltage threshold, filter the data segments whose difference value exceeds the operating state activation voltage threshold, extract the sequence position index information corresponding to the data segment, and generate a positive jump signal index.

[0015] S113: Call the positive transition signal index, retrieve the voltage amplitude rising edge start point pointed to by the positive transition signal index in the time axis data, extract the absolute time value corresponding to the voltage amplitude rising edge start point, use the absolute time value as the time breakpoint for switching from non-operating state to operating state, filter the steady-state fluctuation interference data before and after the time breakpoint, identify the single edge trigger time, and establish the step excitation start time.

[0016] As a further aspect of the present invention, the step of obtaining the physical environment thermal response hysteresis time specifically includes:

[0017] S211: Call the step excitation start time, start the indoor temperature sensing unit to perform discrete continuous sampling, record the thermal signal amplitude of each sampling period, map and bind the collected temperature data with the corresponding timestamp, and generate an indoor temperature time series.

[0018] S212: Based on the indoor temperature time series, extract the temperature values ​​at adjacent sampling times, calculate the dynamic response intensity value of the thermal environment, quantify the unsteady fluctuation characteristics of the spatial thermal field, and generate an instantaneous temperature change rate sequence.

[0019] S213: Call the instantaneous temperature change rate sequence, compare the elements of the instantaneous temperature change rate sequence with the preset temperature response judgment benchmark value, filter the time index corresponding to the instantaneous temperature change rate data that exceeds the temperature response judgment benchmark value, subtract the start time of the step excitation from the time index, calculate the time difference, and obtain the physical environment thermal response lag time.

[0020] As a further aspect of the present invention, the dynamic response intensity value of the thermal environment is expressed by the formula:

[0021] ;

[0022] in, Represents the dynamic response intensity value of the thermal environment. This represents the temperature value at the current sampling point. This represents the temperature value at the previous sampling point. Represents the sampling time interval. Represents the air convection coupling coefficient. Represents the sensor's thermal inertia factor. Represents the variance of environmental thermal noise. This is the time resolution constant.

[0023] As a further aspect of the present invention, the step of obtaining the virtual environment thermal response hysteresis time specifically includes:

[0024] S311: Call the step excitation start time and set it as the trigger zero point of the virtual air conditioning cooling logic. Perform unsteady heat transfer calculation on the pre-constructed three-dimensional mesh space, solve the thermodynamic state parameters of each mesh node under continuous time step, and perform multi-dimensional correlation mapping between the temperature values ​​of the differentiated coordinate points and the corresponding timestamps to generate the virtual node temperature field evolution matrix.

[0025] S312: Based on the virtual node temperature field evolution matrix, identify the temperature data vector of the specified monitoring point in the virtual space, perform differential operation on the temperature values ​​corresponding to adjacent time indices in the temperature data vector, calculate the temperature amplitude change within a unit time step, construct a numerical sequence reflecting the temperature fluctuation of the virtual thermal environment, and arrange them in chronological order to generate a simulation temperature change rate time series set.

[0026] S313: Call the simulated temperature change rate time series set, compare the simulated temperature change rate values ​​at each time with the preset temperature response judgment benchmark value, filter the time node when the simulated temperature change rate value first exceeds the temperature response judgment benchmark value, calculate the time difference between the timestamp corresponding to the specific time node and the start time of the step excitation, quantify the response delay characteristics of the virtual air conditioner to the control command, and establish the thermal response lag time of the virtual environment.

[0027] As a further aspect of the present invention, the steps for obtaining the dynamically calibrated digital twin model are specifically as follows:

[0028] S411: Call the physical environment thermal response lag time and the virtual environment thermal response lag time, subtract the two time quantities, calculate the time response difference between the physical entity and the digital twin model under the step excitation, and extract the positive and negative polarity signs of the time response difference to generate the lag time deviation value.

[0029] S412: Based on the lag time deviation value, obtain the thermodynamic parameters under real-time operating conditions, call the preset thermal response-thermal capacity sensitivity function, map the scalar deviation in the time dimension to the parameter adjustment direction in the energy dimension, and combine the preset calibration gain coefficient to calculate the correction amount for the thermal capacity parameters of the virtual model, quantify the energy tolerance between the model's thermal inertia and the physical properties of the entity, and generate the thermal inertia parameter compensation gradient.

[0030] The thermal response-heat capacity sensitivity function is used to describe the numerical mapping relationship between time delay deviation and heat capacity correction amount;

[0031] S413: Call the thermal inertia parameter compensation gradient, retrieve the configuration file describing the building envelope and air thermal properties in the digital twin model, and apply the thermal inertia parameter compensation gradient to the original heat capacity parameter field in a weighted update manner according to the gradient descent optimization principle. Reconstruct the heat transfer differential equation system of the virtual environment using the updated parameters to establish a dynamically calibrated digital twin model.

[0032] As a further aspect of the present invention, the steps for obtaining the building energy consumption trend prediction sequence are as follows:

[0033] S511: Acquire meteorological forecast data and personnel scheduling plans including future time windows, synchronize the multi-source heterogeneous data stream with the internal simulation clock of the dynamically calibrated digital twin model, map the predicted outdoor temperature, solar radiation intensity and personnel density distribution values ​​to the virtual building envelope and internal space grid nodes, and generate a multi-source environmental excitation matrix.

[0034] S512: Call the multi-source environmental excitation matrix, load the multi-source environmental excitation matrix data into the unsteady heat transfer calculation kernel of the digital twin model, perform iterative solution of the regional heat balance equation set according to the thermodynamic energy conservation law, calculate the instantaneous cooling and heating compensation power required to offset the input thermal disturbance and maintain the indoor air state parameters within the preset comfort range, aggregate the power demand values ​​calculated for each zone to form a continuous fluctuating load curve, and generate a dynamic heat load demand spectrum;

[0035] S513: Based on the dynamic heat load demand spectrum, retrieve the partial load performance characteristic curves and variable operating condition energy efficiency ratio configuration parameters of the HVAC unit, map the heat load demand values ​​to the power consumption data domain, calculate the expected operating energy consumption at each time step, and serialize the expected operating energy consumption according to the time extrapolation order to generate a building energy consumption trend prediction sequence.

[0036] The digital twin-based building energy consumption simulation and prediction system is used to execute the aforementioned digital twin-based building energy consumption simulation and prediction method. The system includes:

[0037] The excitation moment identification module monitors the level status of the operation control circuit of the air conditioning terminal equipment in the physical building, identifies the signal characteristics of the transition from non-operation to operation, identifies the transition moment, and generates the step excitation start moment.

[0038] The physical hysteresis acquisition module calls the start time of the step excitation, collects indoor temperature data, calculates the rate of temperature change of adjacent sampling points, filters the time when the rate of temperature change exceeds the preset temperature response judgment standard, and generates the physical environment thermal response hysteresis duration.

[0039] The virtual response calculation module inputs the start time of the step excitation into the digital twin model to drive the virtual air conditioner to run, calculates the virtual temperature distribution and rate of change, filters out the time when the rate of change exceeds the temperature response judgment standard, and obtains the virtual environment thermal response lag time.

[0040] The model dynamic calibration module calculates the deviation between the thermal response lag time of the physical environment and the thermal response lag time of the virtual environment. Based on the positive or negative polarity and magnitude of the deviation, it iteratively corrects the thermal inertia parameters in the digital twin model to establish a dynamically calibrated digital twin model.

[0041] The energy consumption trend prediction module uses the dynamically calibrated digital twin model to import meteorological forecast data and personnel scheduling plans for future time periods to perform predictive simulations, calculate the dynamic cooling and heating load curves required to maintain the set environmental comfort level, and generate a building energy consumption trend prediction sequence.

[0042] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0043] In this invention, the start time of step excitation is identified by monitoring the sudden changes in the operating status of the equipment. The lag time of the physical environment's thermal response is determined by combining the measured temperature change rate. The virtual model is driven to run and obtain the lag time of the virtual environment's thermal response. The deviation between the two is calculated and the thermal inertia parameters in the digital twin model are iteratively corrected accordingly. A dynamic calibration mechanism is established to eliminate model parameter errors and ensure that the thermodynamic behavior of the virtual space is highly consistent with that of the physical building. The calibrated model is then used to import weather forecasts and scheduling plans for predictive simulation, generating dynamic heating and cooling load curves required to maintain comfort, thereby achieving accurate prediction of building energy consumption trends. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the workflow of the present invention;

[0045] Figure 2 This is a flowchart of the process for obtaining the start time of the step excitation in this invention;

[0046] Figure 3 This is a flowchart illustrating the process of obtaining the hysteresis time of the physical environment thermal response in this invention.

[0047] Figure 4 This is a flowchart illustrating the process of obtaining the thermal response hysteresis time of the virtual environment in this invention.

[0048] Figure 5 This is a flowchart illustrating the process of obtaining the dynamically calibrated digital twin model in this invention.

[0049] Figure 6 This is a flowchart illustrating the process of obtaining the building energy consumption trend prediction sequence in this invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0051] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0052] Please see Figure 1 This invention provides a technical solution: a method for simulating and predicting building energy consumption based on digital twins, comprising the following steps:

[0053] S1: Monitor the level status of the operation control circuit of the air conditioning terminal equipment in the physical building, identify the signal characteristics of the transition from non-operation state to operation state, use the time node of the operation state transition action as the test excitation benchmark, and generate the step excitation start time.

[0054] S2: Call the step excitation start time to trigger the indoor ambient temperature acquisition process, obtain indoor air temperature data in continuous time series, calculate the temperature change rate between adjacent sampling points, filter the time nodes where the temperature change rate exceeds the preset temperature response judgment standard, and generate the physical environment thermal response lag time.

[0055] S3: Based on the thermal response lag time of the physical environment, the start time of the step excitation is input into the digital twin model to drive the virtual air conditioning model to start running. Using the finite element difference algorithm, the temperature distribution of the virtual wall and air nodes is iteratively calculated, the rate of change of virtual temperature with time step is calculated, the rate of change is compared with the temperature response judgment standard, the time node that meets the judgment temperature response standard is located, and the thermal response lag time of the virtual environment is obtained.

[0056] S4: Calculate the deviation between the physical environment thermal response lag time and the virtual environment thermal response lag time, compare the difference polarity between the physical lag time and the virtual lag time, and perform incremental or decremental corrections on the specific heat capacity parameters of the inner surface of the building envelope and indoor air nodes in the digital twin model according to the difference polarity, and establish a dynamically calibrated digital twin model.

[0057] S5: Using a dynamically calibrated digital twin model, reconstruct the thermal parameter set of the digital twin model, input meteorological forecast data including outdoor temperature and solar radiation intensity and personnel schedule information, solve the building heat balance equation set, calculate the total cooling and heating load required to maintain the indoor set temperature within a specified period, and generate a building energy consumption trend prediction sequence.

[0058] The step excitation start time includes the physical device start-up timestamp; the physical environment thermal response lag time includes the measured temperature fluctuation delay period; the virtual environment thermal response lag time includes the simulated temperature change delay period; the dynamically calibrated digital twin model includes the corrected building envelope thermal resistance and updated indoor air heat capacity parameters; and the building energy consumption trend prediction sequence includes future heating and cooling load predictions and expected power consumption time series data.

[0059] Please see Figure 2 The specific steps for obtaining the start time of the step excitation are as follows:

[0060] S111: Monitor the operation control line level status of the air conditioning terminal equipment in the physical building. Use the analog-to-digital conversion interface to collect the line voltage at high frequency. Extract the voltage amplitude data at several moments according to the preset sampling frequency. Map and bind each voltage amplitude to the corresponding clock timestamp one by one. Construct a two-dimensional matrix including time dimension index and voltage amplitude data to generate the line level time sequence sampling set.

[0061] A high-precision analog-to-digital converter (ADC) data acquisition probe is connected to the 0-10V DC control circuit of the building's air conditioning terminal fan coil unit (FCU) to build a bridge for the conversion between physical and digital signals. The ADC sampling frequency is set to 1000 Hz, i.e., capturing 1000 discrete voltage data points per second, to ensure that the microsecond-level jitter characteristics of instantaneous changes in the control signal can be captured. A continuous acquisition task is started, capturing 60,000 voltage amplitude data points in real time within a 60-second sliding time window, and these data are temporarily stored in a high-speed buffer. The microsecond-level clock signal provided by the Network Time Protocol (NTP) server is read synchronously, and each voltage sampling point is bound to the current absolute timestamp (format YYYY-MM-DD HH:MM:SS.mmmmmm) one-to-one. After binding, a two-dimensional matrix containing two columns of data is constructed. The first column is the time dimension index with monotonically increasing characteristics, and the second column is the corresponding analog voltage quantization value (unit: volts). To eliminate the contamination of the signal by electromagnetic interference from the line, Kalman filtering is performed on the generated line level timing sample set. The process noise covariance is set to 0.01 and the measurement noise covariance is set to 0.1. The smoothed voltage timing data is output to provide a high signal-to-noise ratio basic data source for subsequent step signal identification.

[0062] S112: Based on the line level timing sampling set, calculate the first-order difference value of the voltage amplitude corresponding to adjacent time steps in the two-dimensional matrix, construct the voltage change rate sequence, compare the difference value in the voltage change rate sequence with the preset operating state activation voltage threshold, filter the data segments whose difference value exceeds the operating state activation voltage threshold, extract the sequence position index information corresponding to the data segment, and generate a positive jump signal index.

[0063] Read the filtered line level timing sample set and scan the two-dimensional matrix row by row in chronological order. Select the current time. voltage amplitude Compared with the previous sampling time voltage amplitude Perform a first-order backward difference operation, the calculation formula is as follows: The entire sample set is traversed to generate a voltage change rate sequence reflecting instantaneous voltage fluctuations. The activation voltage threshold for the operating state is set to 0.5 volts. This threshold is based on dead-zone testing of air conditioning control valve actuators; experiments show that the valve only produces substantial mechanical action when the control voltage jump exceeds 0.5 volts. Each difference value in the voltage change rate sequence is compared to 0.5 volts. If the difference value is consistently greater than 0.5 volts within a certain continuous time period, that time period is determined to be an active area for control command transmission. The row index of this data segment in the original matrix is ​​extracted and marked as a positive jump signal index. This index set accurately identifies the electrical signal abrupt change interval during the control's transition from "standby" to "operation" state.

[0064] S113: Call the positive transition signal index, retrieve the voltage amplitude rising edge start point pointed to by the positive transition signal index in the time axis data, extract the absolute time value corresponding to the voltage amplitude rising edge start point, use the absolute time value as the time breakpoint for switching from non-running state to running state, filter the steady-state fluctuation interference data before and after the time breakpoint, identify the single edge trigger time, and establish the step excitation start time.

[0065] In the original voltage time-series data, the rising edge region of the voltage waveform is located. A least-squares method is used to perform linear fitting on the data points in the rising edge region, and the slope and intercept of the fitted line are calculated. The intersection of the fitted line and the steady-state reference voltage line (typically 0 volts) is defined as the theoretical starting point of the voltage amplitude rising edge. The absolute time value corresponding to this starting point is extracted and accurate to the millisecond level as the start time of the step excitation. To verify the accuracy of this moment, 0.5 seconds are backtracked in the negative direction of the time axis and 0.5 seconds are extended in the positive direction, calculating the standard deviation of the voltage data within this 1-second time window. If the standard deviation of the backtracking window (the first 0.5 seconds) is less than 0.02 volts (i.e., in steady state), and the standard deviation of the extended window (the last 0.5 seconds) is greater than 0.2 volts (i.e., in non-steady state), then this time point is confirmed as a valid single edge trigger time. This moment marks the zero point at which the physical air conditioner officially receives the start command, providing a unique time reference for subsequent calculations of thermal response hysteresis.

[0066] Please see Figure 3 The specific steps for obtaining the hysteresis time of the physical environment thermal response are as follows:

[0067] S211: Call the start time of the step excitation, start the indoor temperature sensing unit to perform discrete continuous sampling, record the thermal signal amplitude of each sampling period, map and bind the collected temperature data with the corresponding timestamp, and generate an indoor temperature time series.

[0068] At a determined step excitation start time To trigger the signal, the wireless sensor network deployed indoors is immediately activated. A typical working area location, 3 meters horizontally and 1.5 meters vertically from the air conditioner vent, is selected, and discrete continuous sampling is performed using a PT1000 platinum resistance temperature sensor. The temperature sampling period is set to 1 second, taking into account the physical characteristics of air's large heat capacity and slow temperature change, avoiding data redundancy caused by oversampling. During continuous acquisition, the thermal signal amplitude (i.e., Celsius temperature value) within each sampling period is recorded in real time, and the data is transmitted to the central processing unit via the ZigBee wireless protocol. After receiving the data, the processor binds each temperature value to the timestamp of the acquisition time, removes abnormal data caused by packet loss or out-of-order delivery, and reassembles it to generate a continuous and evenly spaced indoor temperature time series. For example, in a 300-second test, the sequence contains 300 temperature-time data pairs, with a temperature resolution of 0.01 degrees Celsius.

[0069] S212: Based on the indoor temperature time series, extract the temperature values ​​of adjacent sampling times using the following formula:

[0070] ;

[0071] Calculate the dynamic response intensity value of the thermal environment, quantify the unsteady-state fluctuation characteristics of the spatial thermal field, and generate a sequence of instantaneous temperature change rate;

[0072] in, Represents the dynamic response intensity value of the thermal environment. This represents the temperature value at the current sampling point. This represents the temperature value at the previous sampling point. Represents the sampling time interval. Represents the air convection coupling coefficient. Represents the sensor's thermal inertia factor. Represents the variance of environmental thermal noise. This is a time resolution constant;

[0073] Data is extracted from the time series of indoor temperatures, assuming the current time is... The previous moment was Read the temperature value of the current sampling point. (e.g., 26.50 degrees Celsius) and the temperature value of the previous sampling point (e.g., 26.52 degrees Celsius). Obtain preset physical parameters: sampling time interval. Set to 1 second, air convection coupling coefficient The sensor thermal inertia factor was set to 3.5 watts per square meter using computational fluid dynamics pre-simulation. The sensor's factory calibration specification specifies a value of 0.8, and the environmental thermal noise variance is... The value, obtained from statistical analysis of idle period data, is 0.005, representing a time resolution constant. The value is set to 1.0. Substitute the above value into the formula for calculating the dynamic response intensity of the thermal environment;

[0074] The first step is to calculate the absolute difference in temperature changes:

[0075] ;

[0076] The second step is to calculate the numerator (weighted heat flux intensity):

[0077] ;

[0078] The third step is to calculate the denominator (damping and time factor):

[0079] ;

[0080] The fourth step is to solve for the dynamic response intensity value of the thermal environment. :

[0081] ;

[0082] Calculation results Based on the temperature response judgment benchmark value (0.05) set in step S213, a comparison was performed: 0.087 > 0.05;

[0083] Conclusion: The calculated dynamic response intensity value of the thermal environment exceeds the judgment benchmark value. Therefore, the current moment (i.e., the moment when the temperature difference changes by 0.02℃) is determined to be a valid thermal response time node, and the timestamp of this point is extracted as... The calculation result is consistent with the statement that "when the air conditioner's cold air effectively disturbs the indoor temperature field," The physical description states that the value will quickly exceed 0.05.

[0084] The physical meaning of this formula is as follows: the numerator amplifies the instantaneous temperature change through differential calculation and uses the convection coupling coefficient for weighting, reflecting the actual heat transfer rate; the denominator constructs a damping term using the thermal inertia factor and noise variance to suppress spurious fluctuations caused by sensor delay and random environmental thermal disturbances. The calculated dynamic response intensity value of the thermal environment is... This method can quantify the unsteady-state fluctuation characteristics of the spatial thermal field with high sensitivity. Performing this calculation on all data points in the sequence generates a sequence of instantaneous temperature change rates.

[0085] S213: Call the instantaneous temperature change rate sequence, compare the elements of the instantaneous temperature change rate sequence with the preset temperature response judgment benchmark value, filter the time index corresponding to the instantaneous temperature change rate data that exceeds the temperature response judgment benchmark value, subtract the start time of the step excitation from the time index, calculate the time difference, and obtain the physical environment thermal response lag time.

[0086] The temperature response benchmark was set at 0.05, determined through multiple cooling load step experiments on a typical office environment. Experimental data showed that, without air conditioning, the temperature response fluctuation was calculated as... The value remained below 0.03; however, when the air conditioner's cold air effectively disturbed the indoor temperature field, The value will quickly exceed 0.05. Traverse the instantaneous temperature change rate sequence, find the element whose value first exceeds 0.05, and extract the timestamp corresponding to that position. Perform a subtraction operation: For example, if It is 10:00:00. If the time is 10:01:30, the calculated time difference is 90 seconds. This value is the physical environment thermal response lag time, which characterizes the physical delay from when the air conditioner is turned on to when a detectable change occurs in the indoor temperature field. It includes the sum of air duct transmission delay, air mixing delay, and sensor response delay.

[0087] Please see Figure 4The specific steps for obtaining the virtual environment thermal response lag time are as follows:

[0088] S311: Call the start time of the step excitation, set it as the trigger zero point of the virtual air conditioning cooling logic, perform unsteady heat transfer calculation on the pre-constructed three-dimensional mesh space, solve the thermodynamic state parameters of each mesh node under continuous time steps, perform multi-dimensional correlation mapping between the temperature values ​​of the differentiated coordinate points and the corresponding timestamps, and generate the virtual node temperature field evolution matrix.

[0089] Using simulation kernels such as EnergyPlus or OpenFOAM, a 3D geometric model of the target building is loaded, and the determined step excitation start time in the physical world is determined. Mapped to second 0 (TimeStep 0) of the virtual simulation environment. Initialize the boundary conditions of the 3D mesh space, setting the initial temperature of walls, windows, and floors to constant values ​​(e.g., 26 degrees Celsius), and setting the virtual air conditioner's air supply to maximum cooling mode (e.g., supply air temperature 18 degrees Celsius, air volume 1000 cubic meters per hour). Start the unsteady (transient) heat transfer calculation solver, setting the time step to 1 second, and iteratively solve the governing equations (including mass conservation, momentum conservation, and energy conservation equations). Calculate the temperature, velocity, and pressure parameters of all 10,000 mesh nodes at each time step. Focus on extracting the temperature values ​​of virtual monitoring points (coordinates x=3.0m, y=4.5m, z=1.5m) that correspond to the physical sensor coordinates, and perform multi-dimensional correlation mapping of these values ​​with the simulation time axis to generate a virtual node temperature field evolution matrix containing spatial coordinates, timestamps, and temperature values.

[0090] S312: Based on the virtual node temperature field evolution matrix, identify the temperature data vector of the specified monitoring point in the virtual space, perform differential operation on the temperature values ​​corresponding to adjacent time indices in the temperature data vector, calculate the temperature amplitude change within a unit time step, construct a numerical sequence reflecting the temperature fluctuation of the virtual thermal environment, and arrange them in chronological order to generate a simulation temperature change rate time series set.

[0091] Filter out the column vector of temperature data from the specified monitoring points, and perform discrete difference operation on this vector. The calculation formula is as follows: ,in The time step is 1 second. This calculation aims to quantify the change in temperature amplitude within a unit time step in virtual space. Since numerical simulations are free of measurement noise, direct difference can reflect temperature fluctuations. The calculated changes are arranged in chronological order of the simulation time to construct a time series set of simulated temperature change rates. For example, in the first 30 seconds after the simulation begins, the temperature change rate remains at 0 because the virtual wind front has not yet reached the monitoring point; when the virtual cold wind front arrives, the temperature change rate will increase significantly.

[0092] S313: Call the simulated temperature change rate time series set, compare the simulated temperature change rate values ​​at each time with the preset temperature response judgment benchmark value, filter the time node when the simulated temperature change rate value first exceeds the temperature response judgment benchmark value, calculate the time difference between the timestamp corresponding to the specific time node and the start time of the step excitation, quantify the response delay characteristics of the virtual air conditioner to the control command, and establish the thermal response lag time of the virtual environment.

[0093] A temperature response benchmark value (0.05) identical to the physical environment is set. The simulated temperature change rate time series is traversed, and the simulated temperature change rate values ​​are compared one by one. Once a certain moment is detected... The moment the rate value first exceeds 0.05, that time point is immediately locked. Calculate that specific time point. The time difference between the virtual trigger zero point and the virtual trigger zero point yields the virtual environment thermal response hysteresis time. Typically, simplified models often assume ideal mixing or neglect some heat capacity, resulting in calculated... (For example, 60 seconds) will be less than the physical lag time. This metric quantifies the response delay characteristics of the current uncalibrated virtual air conditioner to control commands and is a key basis for subsequent model parameter tuning.

[0094] Please see Figure 5 The specific steps for obtaining a dynamically calibrated digital twin model are as follows:

[0095] S411: Call the physical environment thermal response lag time and the virtual environment thermal response lag time, subtract the two time quantities, calculate the time response difference between the physical entity and the digital twin model under step excitation, extract the positive and negative polarity signs of the time response difference, and generate the lag time deviation value.

[0096] Read the physical environment thermal response hysteresis time calculated in step S213. (For example, 90 seconds) and the virtual environment thermal response hysteresis time calculated in step S313 (For example, 60 seconds). Perform subtraction: The difference is 30 seconds. A positive difference indicates that the digital twin model responds faster than the physical entity, meaning that the thermal inertia in the model is underestimated. Extracting the difference of 30 and its positive sign "+" generates the lag time deviation value. A negative difference indicates that the model's thermal inertia is too large. This deviation value directly reflects the degree of distortion in the model's dynamic characteristics.

[0097] S412: Based on the hysteresis deviation value, obtain the thermodynamic parameters under real-time operating conditions using the following formula:

[0098] ;

[0099] Calculate the correction amount for the thermal capacity parameters of the virtual model, quantify the energy tolerance between the thermal inertia of the model and the physical properties of the entity, and generate the thermal inertia parameter compensation gradient.

[0100] in, The amount representing the correction to the heat capacity parameters of the virtual model; Represents the hysteresis time of the physical environment's thermal response. Represents the thermal response lag time of the virtual environment. Represents the calibration gain coefficient. Represents the real-time cooling capacity of the air conditioning unit. Represents the indoor return air temperature. Represents the outdoor ambient temperature. Represents the residual energy modulus. The nominal heat capacity constant of the representative model, Represents the damping temperature factor;

[0101] To eliminate hysteresis, the heat capacity parameters of the virtual model need to be corrected to obtain the thermodynamic parameters under real-time operating conditions: real-time cooling power of the air conditioning unit. 5000 watts, indoor return air temperature The outdoor temperature is 26 degrees Celsius. At 32 degrees Celsius, calibrate the gain coefficient. Based on experience, the value is set to 0.1, and the model's nominal heat capacity constant is... The damping temperature factor is 100,000 joules per Kelvin. Set the temperature to 2.0 degrees Celsius. Calculate the residual energy modulus. Its value is the product of cooling power and hysteresis deviation, i.e. Joules. Substitute the above parameters into the formula for calculating the thermal inertia parameter compensation gradient. In this calculation logic: numerator part This represents the total energy that needs to be replenished or reduced, estimated by integrating the cooling power over the deviation time to measure the thermo-mass difference; the denominator is... This represents the driving temperature difference potential energy under the current environment, and introduces... This is to prevent singularities in the denominator when the indoor and outdoor temperature difference is close to zero, ensuring calculation stability; tail terms It is a dimensionless energy weighting factor used to dynamically adjust the correction step size based on the magnitude of the residual energy.

[0102] Substitute the parameter values ​​into the calculation (the values ​​are for illustrative purposes only): Assume the calculated result of the base correction gradient (the part in square brackets) is:

[0103] ;

[0104] The energy weighting factor (tail term) is:

[0105] ;

[0106] The correction amount is obtained:

[0107] ;

[0108] This result is the thermal inertia parameter compensation gradient, indicating that the total heat capacity in the virtual model needs to be increased by 3560 joules per Kelvin to increase its thermal inertia and extend its response lag time to match that of the physical entity.

[0109] S413: Call the thermal inertia parameter compensation gradient, retrieve the configuration file describing the building envelope and air thermal properties in the digital twin model, and apply the thermal inertia parameter compensation gradient to the original heat capacity parameter field in a weighted update manner according to the gradient descent optimization principle. Use the updated parameters to reconstruct the heat transfer differential equations of the virtual environment and establish a dynamically calibrated digital twin model.

[0110] Based on the calculated thermal inertia parameter compensation gradient (+3560 J / K), a script is written to read the configuration file of the digital twin model (such as an IDF file from EnergyPlus or an MO model file from Modelica). The script locates the fields describing the building envelope (walls, furniture) and the thermal and mass properties of the indoor air. Finally, it obtains the total mass of the indoor air in the current virtual model. (For example, 100kg), according to the specific heat capacity definition formula Calculate the increase in the specific heat capacity of air. J / (kg·K). This increment is added to the original air specific heat capacity parameter, for example, adjusting the air specific heat capacity from 1005 J / (kg·K) to 1040.6 J / (kg·K), or increasing the equivalent mass of the virtual furniture. Using the updated parameter set, the heat transfer differential equations of the virtual environment are recompiled to complete the dynamic reconstruction of the model. As shown in Table 1 below, after calibration, the error between the predicted lag time and the physical measured value of the digital twin model is significantly reduced;

[0111] Table 1: Comparison of lag time before and after model calibration:

[0112]

[0113] Please see Figure 6 The specific steps for obtaining the building energy consumption trend prediction sequence are as follows:

[0114] S511: Acquire meteorological forecast data and personnel scheduling plans including future time windows, synchronize and calibrate the multi-source heterogeneous data stream with the internal simulation clock of the dynamically calibrated digital twin model, map the predicted outdoor temperature, solar radiation intensity and personnel density distribution values ​​to the virtual building envelope and internal space grid nodes, and generate a multi-source environmental excitation matrix.

[0115] By connecting to the local meteorological station database via API, hourly weather forecast data for the next 24 hours is retrieved, including outdoor dry-bulb temperature (e.g., high of 35 degrees Celsius, low of 24 degrees Celsius), relative humidity, total horizontal radiation intensity, and wind speed and direction. Simultaneously, the personnel access control scheduling plan for the next day is exported from Building Automation (BAS) to obtain the expected indoor personnel density distribution data (e.g., 0.1 people / square meter). The above multi-source heterogeneous data streams are aligned in time granularity, uniformly interpolated to a 15-minute time step, and synchronized with the internal simulation clock of the dynamically calibrated digital twin model (starting from 00:00 the next day). Using a data mapping algorithm, outdoor temperature boundary conditions are applied to the surface of the virtual building's outer envelope. Solar radiation intensity is calculated using ray tracing to convert it into the heat gain transmitted through windows facing each direction. Personnel density values ​​are converted into corresponding sensible and latent heat dissipation power and loaded into the internal spatial grid nodes, thereby generating a multi-source environmental excitation matrix with full spatiotemporal coverage.

[0116] S512: Call the multi-source environmental excitation matrix, load the multi-source environmental excitation matrix data into the unsteady heat transfer calculation kernel of the digital twin model, perform iterative solution of the regional heat balance equations according to the thermodynamic energy conservation law, calculate the instantaneous cooling and heating compensation power required to offset the input thermal disturbance and maintain the indoor air state parameters within the preset comfort range, aggregate the power demand values ​​calculated for each zone to form a continuous fluctuating load curve, and generate a dynamic heat load demand spectrum;

[0117] The generated multi-source environmental excitation matrix is ​​used as the boundary input condition and loaded into the unsteady heat transfer calculation kernel of the digital twin model. Based on the first law of thermodynamics (the law of conservation of energy), a set of regional heat balance equations is established for each building zone: .in, This includes solar radiation, heat dissipation from personnel and equipment, and heat generated by lighting. For building heat storage (accurately calibrated in S4 stage). The heat transfer loss is calculated based on the building envelope. An iterative solution is performed to calculate the instantaneous cooling and heating compensation power required by the air conditioning system to maintain indoor air parameters (e.g., temperature 24 degrees Celsius, humidity 50%) within a preset comfort range. For example, the instantaneous cooling load at 14:00 is calculated to be 150 kW. The calculation results for each zone of the entire floor are aggregated to form a continuously fluctuating load curve, generating a dynamic heat load demand spectrum. This demand spectrum not only reflects the static load but also accurately includes the load peak lag effect due to thermal inertia.

[0118] S513: Based on the dynamic heat load demand spectrum, retrieve the partial load performance characteristic curves and variable operating condition energy efficiency ratio configuration parameters of HVAC units, map the heat load demand values ​​to the power consumption data domain, calculate the expected operating energy consumption at each time step, and serialize the expected operating energy consumption according to the time projection order to generate a building energy consumption trend prediction sequence.

[0119] Consult the equipment technical manual of the HVAC unit to digitally extract the Part Load Ratio (PLR) curve and variable operating condition coefficient of performance (COP) configuration parameters. For example, a centrifugal chiller unit has a COP of 5.5 at 100% load, which increases to 6.2 at 50% load. Read the heat load value (e.g., 150 kW) for each time step in the dynamic heat load demand spectrum and calculate its proportion of the unit's rated capacity (PLR). Based on the PLR ​​value, look up a table or interpolate it on the characteristic curve to obtain the real-time COP value. Use the formula... Calculate the expected energy consumption for that time step. For example, dividing a 150 kW cooling load by a COP of 6.0 yields a 25 kW electrical power input. Sequentially arrange the energy consumption values ​​for all time steps according to the time-based projection sequence (00:00 to 23:45) to generate a building energy consumption trend prediction sequence.

[0120] Table 2: Validation Data of Building Energy Consumption Prediction Accuracy

[0121] As shown in Table 2, the experimental results confirm that the prediction sequence of building energy consumption trend generated based on the dynamic calibration digital twin model has a prediction error of less than 2% in different time periods. It can effectively capture the nonlinear influence of building thermal inertia on energy consumption and provide reliable data support for subsequent power demand management and flexible load scheduling.

[0122] The digital twin-based building energy consumption simulation and prediction system is used to execute the above-mentioned digital twin-based building energy consumption simulation and prediction method. The system includes:

[0123] The excitation moment identification module monitors the level status of the operation control circuit of the air conditioning terminal equipment in the physical building, identifies the signal characteristics of the transition from non-operation to operation, identifies the transition moment, and generates the step excitation start moment.

[0124] The physical hysteresis acquisition module calls the start time of the step excitation, collects indoor temperature data, calculates the rate of temperature change of adjacent sampling points, filters the time when the rate of temperature change exceeds the preset temperature response judgment standard, and generates the physical environment thermal response hysteresis duration.

[0125] The virtual response calculation module inputs the start time of the step excitation into the digital twin model to drive the virtual air conditioner, calculates the virtual temperature distribution and rate of change, filters out the time when the rate of change exceeds the temperature response judgment standard, and obtains the lag time of the virtual environment thermal response.

[0126] The model dynamic calibration module calculates the deviation between the thermal response lag time of the physical environment and the thermal response lag time of the virtual environment. Based on the positive or negative polarity and magnitude of the deviation, it iteratively corrects the thermal inertia parameters in the digital twin model to establish a dynamically calibrated digital twin model.

[0127] The energy consumption trend prediction module uses a dynamically calibrated digital twin model to import weather forecast data and staff scheduling plans for future time periods for predictive simulation, calculates the dynamic cooling and heating load curves required to maintain the set environmental comfort level, and generates a building energy consumption trend prediction sequence.

[0128] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for simulating and predicting building energy consumption based on digital twins, characterized in that, Includes the following steps: S1: Monitor the level status of the operation control circuit of the air conditioning terminal equipment in the physical building, identify the signal characteristics of the transition from non-operation to operation, identify the transition time, and generate the start time of the step excitation. S2: Call the start time of the step excitation, collect indoor temperature data, calculate the temperature change rate of adjacent sampling points, filter the time when the temperature change rate exceeds the preset temperature response judgment standard, and generate the physical environment thermal response hysteresis time. S3: Input the start time of the step excitation into the digital twin model to drive the virtual air conditioner to run, calculate the virtual temperature distribution and rate of change, filter the time when the rate of change exceeds the temperature response judgment standard, and obtain the virtual environment thermal response lag time. S4: Calculate the deviation between the thermal response lag time of the physical environment and the thermal response lag time of the virtual environment. Based on the polarity and magnitude of the deviation, iteratively correct the thermal inertia parameters in the digital twin model to establish a dynamically calibrated digital twin model. The specific steps are as follows: S411: Call the physical environment thermal response lag time and the virtual environment thermal response lag time, subtract the two time quantities, calculate the time response difference between the physical entity and the digital twin model under the step excitation, and extract the positive and negative polarity signs of the time response difference to generate the lag time deviation value. S412: Based on the lag time deviation value, obtain the thermodynamic parameters under real-time operating conditions, call the preset thermal response-thermal capacity sensitivity function, map the scalar deviation in the time dimension to the parameter adjustment direction in the energy dimension, and combine the preset calibration gain coefficient to calculate the correction amount for the thermal capacity parameters of the virtual model, quantify the energy tolerance between the model's thermal inertia and the physical properties of the entity, and generate the thermal inertia parameter compensation gradient. The thermal response-heat capacity sensitivity function is used to describe the numerical mapping relationship between time delay deviation and heat capacity correction amount; S413: Call the thermal inertia parameter compensation gradient, retrieve the configuration file describing the enclosure structure and air thermal properties in the digital twin model, and apply the thermal inertia parameter compensation gradient to the original heat capacity parameter field in a weighted update manner according to the gradient descent optimization principle. Reconstruct the heat transfer differential equation system of the virtual environment using the updated parameters to establish a dynamically calibrated digital twin model. S5: Using the dynamically calibrated digital twin model, import meteorological forecast data and personnel scheduling plans for future time periods to perform predictive simulation, calculate the dynamic heating and cooling load curves required to maintain the set environmental comfort level, and generate a building energy consumption trend prediction sequence.

2. The building energy consumption simulation and prediction method based on digital twins according to claim 1, characterized in that, The step excitation start time includes the physical device turn-on timestamp, the physical environment thermal response lag time includes the measured temperature fluctuation delay period, the virtual environment thermal response lag time includes the simulated temperature change delay period, the dynamically calibrated digital twin model includes the corrected building envelope thermal resistance and updated indoor air heat capacity parameters, and the building energy consumption trend prediction sequence includes future heating and cooling load predictions and expected power consumption time series data.

3. The building energy consumption simulation and prediction method based on digital twins according to claim 1, characterized in that, The specific steps for obtaining the start time of the step excitation are as follows: S111: Monitor the operation control line level status of the air conditioning terminal equipment in the physical building. Use the analog-to-digital conversion interface to collect the line voltage at high frequency. Extract the voltage amplitude data at several moments according to the preset sampling frequency. Map and bind each voltage amplitude to the corresponding clock timestamp one by one. Construct a two-dimensional matrix including time dimension index and voltage amplitude data to generate the line level time sequence sampling set. S112: Based on the line level timing sampling set, calculate the first-order difference value of the voltage amplitude corresponding to adjacent time steps in the two-dimensional matrix, construct the voltage change rate sequence, compare the difference value in the voltage change rate sequence with the preset operating state activation voltage threshold, filter the data segments whose difference value exceeds the operating state activation voltage threshold, extract the sequence position index information corresponding to the data segment, and generate a positive jump signal index. S113: Call the positive transition signal index, retrieve the voltage amplitude rising edge start point pointed to by the positive transition signal index in the time axis data, extract the absolute time value corresponding to the voltage amplitude rising edge start point, use the absolute time value as the time breakpoint for switching from non-operating state to operating state, filter the steady-state fluctuation interference data before and after the time breakpoint, identify the single edge trigger time, and establish the step excitation start time.

4. The building energy consumption simulation and prediction method based on digital twins according to claim 3, characterized in that, The specific steps for obtaining the hysteresis time of the physical environment thermal response are as follows: S211: Call the step excitation start time, start the indoor temperature sensing unit to perform discrete continuous sampling, record the thermal signal amplitude of each sampling period, map and bind the collected temperature data with the corresponding timestamp, and generate an indoor temperature time series. S212: Based on the indoor temperature time series, extract the temperature values ​​at adjacent sampling times, calculate the dynamic response intensity value of the thermal environment, quantify the unsteady fluctuation characteristics of the spatial thermal field, and generate an instantaneous temperature change rate sequence. S213: Call the instantaneous temperature change rate sequence, compare the elements of the instantaneous temperature change rate sequence with the preset temperature response judgment benchmark value, filter the time index corresponding to the instantaneous temperature change rate data that exceeds the temperature response judgment benchmark value, subtract the start time of the step excitation from the time index, calculate the time difference, and obtain the physical environment thermal response lag time.

5. The building energy consumption simulation and prediction method based on digital twins according to claim 4, characterized in that, The dynamic response intensity value of the thermal environment is expressed by the formula: ； in, Represents the dynamic response intensity value of the thermal environment. This represents the temperature value at the current sampling point. This represents the temperature value at the previous sampling point. Represents the sampling time interval. Represents the air convection coupling coefficient. Represents the sensor's thermal inertia factor. Represents the variance of environmental thermal noise. This is the time resolution constant.

6. The building energy consumption simulation and prediction method based on digital twins according to claim 4, characterized in that, The specific steps for obtaining the virtual environment thermal response hysteresis duration are as follows: S311: Call the step excitation start time and set it as the trigger zero point of the virtual air conditioning cooling logic. Perform unsteady heat transfer calculation on the pre-constructed three-dimensional mesh space, solve the thermodynamic state parameters of each mesh node under continuous time step, and perform multi-dimensional correlation mapping between the temperature values ​​of the differentiated coordinate points and the corresponding timestamps to generate the virtual node temperature field evolution matrix. S312: Based on the virtual node temperature field evolution matrix, identify the temperature data vector of the specified monitoring point in the virtual space, perform differential operation on the temperature values ​​corresponding to adjacent time indices in the temperature data vector, calculate the temperature amplitude change within a unit time step, construct a numerical sequence reflecting the temperature fluctuation of the virtual thermal environment, and arrange them in chronological order to generate a simulation temperature change rate time series set. S313: Call the simulated temperature change rate time series set, compare the simulated temperature change rate values ​​at each time with the preset temperature response judgment benchmark value, filter the time node when the simulated temperature change rate value first exceeds the temperature response judgment benchmark value, calculate the time difference between the timestamp corresponding to the specific time node and the start time of the step excitation, quantify the response delay characteristics of the virtual air conditioner to the control command, and establish the thermal response lag time of the virtual environment.

7. The building energy consumption simulation and prediction method based on digital twins according to claim 1, characterized in that, The specific steps for obtaining the building energy consumption trend prediction sequence are as follows: S511: Acquire meteorological forecast data and personnel scheduling plans including future time windows, synchronize the multi-source heterogeneous data stream with the internal simulation clock of the dynamically calibrated digital twin model, map the predicted outdoor temperature, solar radiation intensity and personnel density distribution values ​​to the virtual building envelope and internal space grid nodes, and generate a multi-source environmental excitation matrix. S512: Call the multi-source environmental excitation matrix, load the multi-source environmental excitation matrix data into the unsteady heat transfer calculation kernel of the digital twin model, perform iterative solution of the regional heat balance equation set according to the thermodynamic energy conservation law, calculate the instantaneous cooling and heating compensation power required to offset the input thermal disturbance and maintain the indoor air state parameters within the preset comfort range, aggregate the power demand values ​​calculated for each zone to form a continuous fluctuating load curve, and generate a dynamic heat load demand spectrum; S513: Based on the dynamic heat load demand spectrum, retrieve the partial load performance characteristic curves and variable operating condition energy efficiency ratio configuration parameters of the HVAC unit, map the heat load demand values ​​to the power consumption data domain, calculate the expected operating energy consumption at each time step, and serialize the expected operating energy consumption according to the time extrapolation order to generate a building energy consumption trend prediction sequence.

8. A building energy consumption simulation and prediction system based on digital twins, characterized in that, The system is used to implement the building energy consumption simulation and prediction method based on digital twins as described in any one of claims 1-7, and the system comprises: The excitation moment identification module monitors the level status of the operation control circuit of the air conditioning terminal equipment in the physical building, identifies the signal characteristics of the transition from non-operation to operation, identifies the transition moment, and generates the step excitation start moment. The physical hysteresis acquisition module calls the start time of the step excitation, collects indoor temperature data, calculates the rate of temperature change of adjacent sampling points, filters the time when the rate of temperature change exceeds the preset temperature response judgment standard, and generates the physical environment thermal response hysteresis duration. The virtual response calculation module inputs the start time of the step excitation into the digital twin model to drive the virtual air conditioner to run, calculates the virtual temperature distribution and rate of change, filters out the time when the rate of change exceeds the temperature response judgment standard, and obtains the virtual environment thermal response lag time. The model dynamic calibration module calculates the deviation between the thermal response lag time of the physical environment and the thermal response lag time of the virtual environment. Based on the polarity and magnitude of the deviation, iteratively corrects the thermal inertia parameters in the digital twin model to establish a dynamically calibrated digital twin model. Specifically, this includes: The physical environment thermal response lag time and the virtual environment thermal response lag time are called, and the two time quantities are subtracted to calculate the time response difference between the physical entity and the digital twin model under the step excitation. The positive and negative polarity signs of the time response difference are extracted to generate the lag time deviation value. Based on the lag time deviation value, the thermodynamic parameters under real-time operating conditions are obtained, the preset thermal response-heat capacity sensitivity function is called, the scalar deviation in the time dimension is mapped to the parameter adjustment direction in the energy dimension, and combined with the preset calibration gain coefficient, the correction amount for the heat capacity parameters of the virtual model is calculated, the energy tolerance between the thermal inertia of the model and the physical properties of the entity is quantified, and the thermal inertia parameter compensation gradient is generated. The thermal inertia parameter compensation gradient is invoked, and the configuration file describing the building envelope and air thermal properties in the digital twin model is retrieved. Based on the gradient descent optimization principle, the thermal inertia parameter compensation gradient is applied to the original heat capacity parameter field in a weighted update manner. The heat transfer differential equations of the virtual environment are reconstructed using the updated parameters to establish a dynamically calibrated digital twin model. The energy consumption trend prediction module uses the dynamically calibrated digital twin model to import meteorological forecast data and personnel scheduling plans for future time periods to perform predictive simulations, calculate the dynamic cooling and heating load curves required to maintain the set environmental comfort level, and generate a building energy consumption trend prediction sequence.

Images