A real-time digital twin modeling method and system for building facilities

Abstract

The present invention provides a real-time digital twin modeling method and system for building facilities, relating to the field of building parameter modeling and monitoring technology. The purpose is to accurately refine and design more precise modeling methods in building twin digital modeling. The method includes establishing multiple detection devices for building settings, each of which is used to collect different actual real-time data and associated data; collecting actual real-time data and associated data; establishing a digital twin model based on the actual real-time data and associated data over a collection time period; synchronously and periodically collecting the twin real-time data of the digital twin model and the actual real-time data of the building; updating the digital twin model if the difference between the twin real-time data and the actual real-time data exceeds a preset threshold; and updating the digital twin model if the time since the last digital twin model update reaches a certain length. The present invention has the advantages of digital twin modeling, which accurately and reliably simulates the status of building facilities.

Description

Technical Field

[0001] The present invention relates to the technical field of building parameter modeling and monitoring, and in particular to a real-time digital twin modeling method and system for building facilities. Background Art

[0002] Digital twins make full use of physical models, sensor updates, operation history and other data, integrate multi-disciplinary, multi-physical quantity, multi-scale, and multi-probability simulation processes, complete mapping in virtual space, and thus reflect the continuous cycle process of the corresponding physical facilities.

[0003] When it comes to building monitoring, conventional sensors have a limited real-time field of view, making it difficult to achieve comprehensive and predictive monitoring. Digital twin models, however, can reflect multiple aspects of a building's status in real time, enabling the timely detection and prediction of potential issues, reducing downtime and repair costs. Analyzing a building's digital twin model can optimize resource utilization, such as energy and water management, and reduce operating costs. However, numerous factors influence the observed data in building digital twins, posing certain challenges and leading to issues such as redundancy or incompleteness.

[0004] Therefore, it is necessary to accurately refine and design more precise modeling methods in the twin digital modeling of buildings to achieve more accurate predictive observation of buildings by digital twin modeling. Summary of the Invention

[0005] The purpose of the present invention is to provide a real-time digital twin modeling method and system for building facilities, which can accurately refine and design more precise modeling methods in the twin digital modeling of buildings.

[0006] The embodiments of the present invention are achieved through the following technical solutions:

[0007] A real-time digital twin modeling method for building facilities includes the following steps:

[0008] Establishing multiple detection devices for building settings, each of which is used to collect different actual real-time data and related data, and the detection devices include image acquisition devices and multiple sensors;

[0009] The actual real-time data and the associated data are collected with T1 as a first unit time length;

[0010] Establishing a digital twin model based on the actual real-time data and the associated data during a collection period;

[0011] At every second time unit T2, the twin real-time data of the digital twin model and the actual real-time data of the building are synchronously and periodically collected within the third unit time unit T3;

[0012] If the difference between the twin real-time data and the actual real-time data is greater than a preset threshold, the digital twin model is updated;

[0013] If the time since the last update of the digital twin model reaches M first unit time lengths, the digital twin model is updated.

[0014] Preferably, the real-time data includes power consumption per unit time, operating power of various power facilities in the building, vibration frequency of multiple points in the building, and static image pixel data.

[0015] Preferably, the associated data includes temperature, humidity, month, time and rainfall.

[0016] Preferably, the method for establishing a digital twin model based on the actual real-time data and the associated data in a collection time period is:

[0017] Determine the fitting relationship type between the actual real-time data and the associated data through a trained model recognition network;

[0018] A fitting parameter is obtained based on the fitting relationship type, and a functional relationship between the actual real-time data and the associated data is established.

[0019] Preferably, the method for determining the type of fitting relationship between the actual real-time data and the associated data by using the trained model recognition network is:

[0020] Inputting the actual real-time data and the associated data into the model recognition network;

[0021] Perform feature extraction;

[0022] Output the fitted relationship type through the activation function;

[0023] Determine the fitting relationship.

[0024] Preferably, the method for performing feature extraction is:

[0025] For each of the actual real-time data, fitting relationship characteristic parameters are established based on the associated data:

[0026]

[0027] in, and are respectively the first fitting relationship characteristic parameter, the second fitting relationship characteristic parameter, the third fitting relationship characteristic parameter and the fourth fitting relationship characteristic parameter for the bth actual real-time data based on the ath associated data, b,i and x a,iare the i-th actual real-time data and the i-th associated data collected in the first unit time, N is the total number of the actual real-time data collected in the first unit time, and d1, d2 and d3 are coefficients respectively;

[0028] For each of the actual real-time data, a fitting relationship feature matrix REL is established b :

[0029]

[0030] Among them, REL b (a,q) represents the element in the ath row and the qth column.

[0031] Preferably, the method of outputting the fitting relationship type through the activation function is:

[0032] Based on the REL b Outputting the functional relationship type of the b-th actual real-time data to each of the associated data respectively through a Softmax function, wherein the functional relationship type includes a linear function, a quadratic function, an ln function, and an exp function;

[0033] Obtain the fitting relationship type of the bth actual real-time data to the associated data:

[0034] y b =∑ a γ ab *f ab (x a );

[0035] Among them, y b represents the fitting output of the actual real-time data of type b, namely the twin real-time data, x a is the collected value of the associated data described in type a, f ab (.) represents the functional relationship type of the actual real-time data of type b to the associated data of type a, γ ab It is the coefficient to be determined of the function of the actual real-time data of type b to the associated data of type a.

[0036] Preferably, the method for determining the fitting relationship is to solve the γ by the least square method or the gradient descent method. ab The value of y b and x a Functions between .

[0037] Preferably, a method for determining whether the difference between the twin real-time data and the actual real-time data is greater than a preset threshold is:

[0038] Get the comprehensive difference parameter ε:

[0039]

[0040] Among them, y b,t and They are respectively the tth twin real-time data and the actual real-time data of the bth type of actual real-time data within the third unit time length T3, J is the total number of the twin real-time data collected within the third unit time length T3, and B is the number of types of the actual real-time data.

[0041] The present invention further provides a real-time digital twin modeling system for building facilities, which is applied to any of the above-mentioned real-time digital twin modeling methods for building facilities, comprising:

[0042] The perception module collects different actual real-time data and related data through various detection devices. These real-time data include monitoring images of the corresponding area and the device posture data corresponding to the images. The related data includes the comprehensive working status S1 of the equipment, such as the current main electrical parameters and network status. The related data is transmitted back through the unified external network interface of the general industrial control system such as PLC or DCS;

[0043] The data acquisition module is configured to collect the actual real-time data and the associated data using T1 as the first unit duration. The actual real-time data is aggregated into a database via the building's network. Image data is processed using image recognition algorithms such as Faster-RCNN to identify and record the device's image location information P1 within T1 and status information E1 regarding whether there has been an intrusion. Both the location information and status information are statically associated data for the current device image and are stored in a database via network transmission for subsequent module access.

[0044] A model building module is used to build a digital twin model based on the actual real-time data and the associated data in a collection time period, wherein the model parameters and appearance status are mainly determined by the image position information P1, the equipment working status S1, and the personnel intrusion status E1, and the model status is rendered in real time;

[0045] A model checking module is used to synchronously and periodically collect the twin real-time data of the digital twin model and the actual real-time data of the building within a third unit time T3 at every second time unit T2;

[0046] A model update module is used to update the digital twin model if the difference between the twin real-time data and the actual real-time data is greater than a preset threshold; and if the time since the last update of the digital twin model reaches M of the first unit time lengths, the digital twin model is updated.

[0047] The technical solutions of the embodiments of the present invention have at least the following advantages and beneficial effects:

[0048] The present invention uses digital twin technology to remotely simulate the real-time status of a building in a virtual environment. It does not need to rely on remote communication to receive building status data, and can quickly and timely obtain the building status.

[0049] The present invention can create a virtual model that is highly consistent with the actual state of the building, facilitating accurate simulation of the building's performance, environmental impact, etc. through analysis of the model;

[0050] The present invention achieves real-time monitoring and prediction through remote simulation to help timely discover and deal with potential building problems, reducing the risk of building and equipment failure and operational interruption;

[0051] The present invention helps to realize the visual observation of building status, and extracts effective, accurate and comprehensive models based on environmental parameters that affect the status. The model can then be used to perform extended calculations to obtain prediction data.

[0052] The present invention is reasonably designed and helps to achieve forward-looking observation of the state of a building and realize comprehensive control of the state of the building. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 A flowchart of a real-time digital twin modeling method for building facilities provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0055] Example 1

[0056] This embodiment provides a real-time digital twin modeling method for building facilities. Figure 1 , including the following steps:

[0057] Detection device construction:

[0058] Establishing multiple detection devices for building settings, each of which is used to collect different actual real-time data and related data, and the detection devices include image acquisition devices and multiple sensors;

[0059] Collect actual real-time data and relevant data:

[0060] The actual real-time data and the associated data are collected with T1 as a first unit time length;

[0061] Building a digital twin model:

[0062] Establishing a digital twin model based on the actual real-time data and the associated data during a collection period;

[0063] Perform model update operations:

[0064] Every second time unit T2, within a third time unit T3, the real-time twin data of the digital twin model, the actual real-time data of the building, and the static image pixel data are synchronously and periodically collected;

[0065] If the difference between the twin real-time data and the actual real-time data is greater than a preset threshold, the digital twin model is updated;

[0066] If the time since the last update of the digital twin model reaches M first unit time lengths, the digital twin model is updated.

[0067] This embodiment primarily relies on the relationship between correlated data and actual real-time data to establish a digital twin model. Actual real-time data refers to the state data required for building observation. Based on the digital twin model, building status can be effectively simulated remotely, and the state can be observed based on the simulation results. To ensure model validity, certain conditions are required to calibrate and update the model. This embodiment utilizes sensors, image, and video acquisition devices to comprehensively observe the building and extract data, enabling comprehensive digital twin modeling and data prediction based on this data.

[0068] Example 2

[0069] This embodiment is based on the technical solution of Example 1 and further explains the specific implementation method of each step.

[0070] In this embodiment, the real-time data includes power consumption per unit time, operating power of various power facilities in the building, vibration frequencies at multiple points in the building, and static image pixel data.

[0071] It should be noted that the static image pixel data can be obtained through an image acquisition device, or through frame extraction through a video acquisition device. The static image pixel data finally obtained can be used to extract the standard deviation and / or average value of the pixel values, or the image can be cut into regions to extract the standard deviation and / or average value as the final parameters to be added to the data processing in subsequent steps.

[0072] Furthermore, the associated data includes temperature, humidity, month, time and rainfall.

[0073] This embodiment builds a model based on the relationship between the environmental factors and the time factors and the data to be observed.

[0074] As a preferred solution, the method for establishing a digital twin model based on the actual real-time data and the associated data in a collection time period is:

[0075] Determine the fitting relationship type between the actual real-time data and the associated data through a trained model recognition network;

[0076] A fitting parameter is obtained based on the fitting relationship type, and a functional relationship between the actual real-time data and the associated data is established.

[0077] The method of determining the type of fitting relationship between the actual real-time data and the associated data through the trained model recognition network is preferably:

[0078] Inputting the actual real-time data and the associated data into the model recognition network;

[0079] Perform feature extraction;

[0080] Output the fitted relationship type through the activation function;

[0081] Determine the fitting relationship.

[0082] Specifically, the method for performing feature extraction is preferably:

[0083] For each of the actual real-time data, fitting relationship characteristic parameters are established based on the associated data:

[0084]

[0085]

[0086] in, and are respectively the first fitting relationship characteristic parameter, the second fitting relationship characteristic parameter, the third fitting relationship characteristic parameter and the fourth fitting relationship characteristic parameter for the bth actual real-time data based on the ath associated data, b,i and x a,i are the i-th actual real-time data and the i-th associated data collected in the first unit time, N is the total number of the actual real-time data collected in the first unit time, and d1, d2 and d3 are coefficients respectively;

[0087] For each of the actual real-time data, a fitting relationship feature matrix REL is established b :

[0088]

[0089] Among them, REL b (a,q) represents the element in the ath row and the qth column.

[0090] On this basis, the method of outputting the fitting relationship type through the activation function can be:

[0091] Based on the REL b Outputting the functional relationship type of the b-th actual real-time data to each of the associated data respectively through a Softmax function, wherein the functional relationship type includes a linear function, a quadratic function, an ln function, and an exp function;

[0092] Obtain the fitting relationship type of the bth actual real-time data to the associated data:

[0093] y b =∑ a γ ab *f ab (x a );

[0094] Among them, y b represents the fitting output of the actual real-time data of type b, namely the twin real-time data, x a is the collected value of the associated data described in type a, f ab (.) represents the functional relationship type of the actual real-time data of type b to the associated data of type a, γ ab It is the coefficient to be determined of the function of the actual real-time data of type b to the associated data of type a.

[0095] As a preferred solution of this embodiment, the method for determining the fitting relationship is to solve the γ by the least square method or the gradient descent method. ab The value of y b and x a Functions between .

[0096] This embodiment adjusts the coefficients preceding different types of functions to assist the neural network in capturing the impact of different types of function coefficients on the fitting relationship between different types of associated data and actual real-time data. After determining the closest fitting type between associated data and actual real-time data on a one-to-one basis, the relationship between all independent variables and the dependent variable is integrated to establish a functional relationship for each dependent variable to the multi-independent variable combination. The most appropriate function coefficients are then fitted using methods such as least squares or gradient descent, thereby establishing a real-time digital twin model for various building status data. This embodiment trains the neural network using the widest possible range of function type labels to more accurately capture the functional relationship type under the influence of different independent variables for each dependent variable. The resulting function coefficients and other parameters help further reduce fitting errors and ensure the reliability of the fit.

[0097] Finally, the method for determining whether the difference between the twin real-time data and the actual real-time data is greater than a preset threshold is as follows:

[0098] Get the comprehensive difference parameter ε:

[0099]

[0100] Among them, y b,t and They are respectively the tth twin real-time data and the actual real-time data of the bth type of actual real-time data within the third unit time length T3, J is the total number of the twin real-time data collected within the third unit time length T3, and B is the number of types of the actual real-time data.

[0101] In this embodiment, in order to ensure the reliability of the digital twin model, the model built in this embodiment is updated periodically and verified.

[0102] Example 3

[0103] This embodiment provides a real-time digital twin modeling system for building facilities, which is applied to any of the above-mentioned real-time digital twin modeling methods for building facilities, including:

[0104] The perception module collects different actual real-time data and related data through various detection devices;

[0105] A data collection module, configured to collect the actual real-time data and the associated data using T1 as a first unit time length;

[0106] A model building module, configured to build a digital twin model based on the actual real-time data and the associated data during a collection period;

[0107] A model checking module is used to synchronously and periodically collect the twin real-time data of the digital twin model and the actual real-time data of the building within a third unit time T3 at every second time unit T2;

[0108] A model update module is used to update the digital twin model if the difference between the twin real-time data and the actual real-time data is greater than a preset threshold; and if the time since the last update of the digital twin model reaches M of the first unit time lengths, the digital twin model is updated.

[0109] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A real-time digital twin modeling method for building facilities, characterized by: The following steps are involved: Establishing multiple detection devices for building settings, each of which is used to collect different actual real-time data and related data, and the detection devices include image acquisition devices and multiple sensors; The actual real-time data and the associated data are collected with T1 as a first unit time length; Establishing a digital twin model based on the actual real-time data and the associated data during a collection period; Every second time unit T2, the twin real-time data of the digital twin model and the actual real-time data of the building are synchronously and periodically collected within the third unit time unit T3; If the difference between the twin real-time data and the actual real-time data is greater than a preset threshold, the digital twin model is updated; If the time since the last update of the digital twin model reaches M first unit time lengths, the digital twin model is updated; The method for establishing a digital twin model based on the actual real-time data and the associated data in a collection time period is: Determine the fitting relationship type between the actual real-time data and the associated data through a trained model recognition network; Obtaining fitting parameters based on the fitting relationship type, and establishing a functional relationship between the actual real-time data and the associated data; The method for determining the fitting relationship type between the actual real-time data and the associated data through the trained model recognition network is: Inputting the actual real-time data and the associated data into the model recognition network; Perform feature extraction; Output the fitted relationship type through the activation function; Determine the fitting relationship; The method for performing feature extraction is: For each of the actual real-time data, fitting relationship characteristic parameters are established based on the associated data: in, and are respectively the first fitting relationship characteristic parameter, the second fitting relationship characteristic parameter, the third fitting relationship characteristic parameter and the fourth fitting relationship characteristic parameter for the bth actual real-time data based on the ath associated data, b,i and x a,i are the i-th actual real-time data and the i-th associated data collected in the first unit time, N is the total number of the actual real-time data collected in the first unit time, and d1, d2 and d3 are coefficients respectively; For each of the actual real-time data, a fitting relationship feature matrix REL is established b : Among them, REL b (a,q) represents the element in the ath row and the qth column.

2. A real-time digital twin modeling method for building facilities according to claim 1, characterized in that: The real-time data includes power consumption per unit time, operating power of various power facilities in the building, vibration frequency of multiple points in the building, and static image pixel data.

3. A real-time digital twin modeling method for building facilities according to claim 2, characterized in that: The associated data includes temperature, humidity, month, time and rainfall.

4. A real-time digital twin modeling method for building facilities according to claim 1, characterized in that: The method of outputting the fitting relationship type through the activation function is: Based on the REL b Outputting the functional relationship type of the b-th actual real-time data to each of the associated data respectively through a Softmax function, wherein the functional relationship type includes a linear function, a quadratic function, an ln function, and an exp function; Obtain the fitting relationship type of the bth actual real-time data to the associated data: y b =∑ a c ab *f ab (x a ); Among them, y b represents the fitting output of the actual real-time data of type b, namely the twin real-time data, x a is the collected value of the associated data described in type a, f ab (.) represents the functional relationship type of the actual real-time data of type b to the associated data of type a, γ ab It is the coefficient to be determined of the function of the actual real-time data of type b to the associated data of type a.

5. A real-time digital twin modeling method for building facilities according to claim 4, characterized in that: The method for determining the fitting relationship is to solve the γ by the least square method or the gradient descent method. ab The value of y b and x a Functions between .

6. A real-time digital twin modeling method for building facilities according to claim 1, characterized in that: The method for determining whether the difference between the twin real-time data and the actual real-time data is greater than a preset threshold is as follows: Get the comprehensive difference parameter ε: Among them, y b,t and They are respectively the tth twin real-time data and the actual real-time data of the bth type of actual real-time data within the third unit time length T3, J is the total number of the twin real-time data collected within the third unit time length T3, and B is the number of types of the actual real-time data.

7. A real-time digital twin modeling system for building facilities, applied to a real-time digital twin modeling method for building facilities according to any one of claims 1 to 6, characterized in that: include: The perception module collects different actual real-time data and related data through various detection devices; A data collection module, configured to collect the actual real-time data and the associated data using T1 as a first unit time length; A model building module, configured to build a digital twin model based on the actual real-time data and the associated data during a collection period; A model checking module is used to synchronously and periodically collect the twin real-time data of the digital twin model and the actual real-time data of the building within a third unit time T3 at every second time unit T2; A model update module is used to update the digital twin model if the difference between the twin real-time data and the actual real-time data is greater than a preset threshold; and if the time since the last update of the digital twin model reaches M of the first unit time lengths, the digital twin model is updated.

Images