Building energy-saving and carbon-reducing method, device and equipment and storage medium

Abstract

The invention belongs to the technical field of building energy conservation and carbon reduction, discloses a building energy conservation and carbon reduction method, device and equipment and a storage medium, and constructs a high-precision building BIM model. A digital twinborn model of a Xian mansion is constructed by applying a BIM technology, and the model not only depicts the physical structure of the building in detail, but also simulates key operation details such as personnel flow and equipment operation. By strategically simplifying the model, it is ensured that the model is efficient and accurate. The comparative analysis of the corrected model and actually measured data shows that the prediction error of the model is only 0.10%, the extremely high consistency of the prediction result and the actual situation is highlighted, the high accuracy of model correction is verified, and the high standard requirement of in-depth research is met; according to the invention, balance optimization of carbon emission minimization and comfort maximization is realized. By intelligently adjusting the operation strategy of the air conditioning equipment, the method obviously reduces the carbon emission of the building while ensuring the living comfort.

Description

Technical Field

[0001] The present invention belongs to the technical field of building energy conservation and carbon reduction, and in particular relates to a building energy conservation and carbon reduction method, device, equipment and storage medium. Background Art

[0002] At present, the construction industry is extensively managed and has a low level of intelligence, resulting in serious waste of resources and a high proportion of carbon emissions. Research on carbon emissions in the construction industry is crucial to mitigating climate change and achieving sustainable development. The operation and maintenance stage occupies most of the entire life cycle of a building, with a large total carbon emission and great potential for carbon reduction. At present, in the field of carbon emission control in the operation and maintenance of high-rise buildings, existing technologies mainly use traditional energy management systems to adjust air conditioning, lighting and other equipment based on historical data and simple prediction models; some buildings use building automation systems, deploy various sensors and automation equipment, and monitor and control various facilities and equipment in the building in real time.

[0003] Through the above analysis, the problems and defects of the prior art are as follows:

[0004] (1) Insufficient data integration: Existing technical solutions generally perform poorly in integrating multi-source data. They are unable to effectively unify and analyze multiple data sources such as energy consumption, temperature sensors, and air-conditioning equipment, and lack a comprehensive understanding of the operating status of the entire building.

[0005] (2) Limited level of intelligence: Existing systems often rely on pre-set rules and logic, lack adaptability and learning capabilities, and are unable to update judgments and strategies based on actual conditions.

[0006] (3) Insufficient consideration of environmental factors: Traditional systems rarely consider the impact of external environmental factors (such as weather changes, humidity, etc.) on energy use and comfort inside buildings, resulting in unreasonable optimization plans.

[0007] (4) Lack of consideration for building structure: Traditional systems do not adequately consider the particularities of the internal structure of high-rise buildings (such as atrium structure) and lack targeted optimization strategies. Summary of the invention

[0008] In view of the problems existing in the prior art, the present invention provides a method, device, equipment and storage medium for building energy conservation and carbon reduction.

[0009] The present invention is implemented as follows: a method for energy saving and carbon reduction in buildings, comprising:

[0010] Step 1: Establish a high-precision building BIM model;

[0011] Step 2, model simulation to construct a data set;

[0012] Step 3: Multi-scenario building operation and maintenance prediction based on random forest;

[0013] Step 4: Strengthen deep learning Double-DQN to achieve building carbon reduction optimization.

[0014] Furthermore, the establishment of a high-precision building BIM model:

[0015] (1) Construct a building BIM model in Revit software based on building parameters;

[0016] (2) Export the building geometry data and performance information in Revit as a gbXML file and import it into DesignBuilder (DB) for simulation analysis;

[0017] (3) Correct the building model in DB.

[0018] Furthermore, the building BIM model is constructed in Revit software according to the building parameters: according to the needs and design requirements of the construction, the necessary building parameters of the building plan drawings and structural material information are obtained; these parameters include but are not limited to the overall layout of the building, the size of each room, the location of doors and windows, the distribution of walls and columns, the type of specific structural materials used and detailed information on their properties; after obtaining the above detailed building parameters, the building model is constructed in Revit software.

[0019] Furthermore, the model simulation constructs a data set:

[0020] The cooling set temperature of each floor of the building was selected as the key input variable, while carbon emissions and indoor environmental comfort were set as output indicators for evaluating system performance;

[0021] The cooling temperature of the 11th floor was selected as the input variable, and the carbon emissions and comfort index PPD were selected as the output indicators. An orthogonal experiment was designed under three temperature scenarios, and a data set was established through DB simulation.

[0022] Furthermore, the random forest-based building multi-scenario operation and maintenance prediction:

[0023] On the basis of establishing a complete data set, the random forest algorithm was used to build a prediction model; a stratified strategy was adopted to build a separate model for each floor;

[0024] The coefficient of determination (R 2 ) and Mean Absolute Error (MAE), as defined in formulas (1) and (2);

[0025]

[0026]

[0027] In the formula, y i Represent the predicted value and the actual value respectively. represents the average of the actual values, and n represents the number of samples.

[0028] Furthermore, the enhanced deep learning Double-DQN achieves building carbon reduction optimization:

[0029] The Double-DQN (DDQN) method is introduced to conduct multi-objective optimization analysis on the two indicators of building carbon emissions and comfort, and the optimal cooling solution under each scenario is obtained;

[0030] The state should be the cooling temperature of each floor, the action should be the change of the cooling temperature of each floor, and the reward should be the weighted sum of the standardized carbon emissions and the PPD value;

[0031] Multi-objective optimization tasks involve weighing multiple competing objective functions in a single solution process; they can be defined as formula (3)(4);

[0032] miny=f(X)=[f1(x),f2(x),…,f m (x)] T (3)

[0033] st.g i (X)≤0(j=1,2,…,p)

[0034] h k (X)=0(k=1,2,…,q)

[0035]

[0036] X=[x1,x2,…,x n ] T ∈Θ

[0037] Y=[y1,y2,…,y n ] T ∈ψ

[0038] Where m is the number of optimization objective functions, Θ is the n-dimensional search space, and and lower limit Determine; ψ is the m-dimensional vector space of the objective function, determined by Θ and the objective function f(x). Equation g i (X)≤0(j=1,2,…,p) and h k(X) = 0 (k = 1, 2, ..., q) represents p inequality constraints and q equality constraints. In particular, if p = q = 0, the problem is simplified to an unconstrained multi-objective optimization problem.

[0039] The algorithm principle of DDQN; Compared with the DQN algorithm, only the method of calculating the target value is changed from formula (5) to formula (6);

[0040]

[0041]

[0042] Where r is the immediate reward, γ is the discount factor, and s t+1 is the next state, θ is the parameter of the Q network, are the parameters of the target network, Indicates using the target network in the next state s t+1 All actions on a t+1 The maximum Q value in represents the use of Q network to select the best action, Indicates that the Q value of the selected action is calculated using the target network.

[0043] Another object of the present invention is to provide a building energy-saving and carbon-reducing system comprising:

[0044] BIM model building module, used to build high-precision building BIM models;

[0045] The simulation module is used to construct data sets for model simulation;

[0046] Prediction module, used for multi-scenario operation and maintenance prediction of buildings based on random forest;

[0047] Optimization module, used to enhance deep learning Double-DQN to achieve building carbon reduction optimization.

[0048] Another object of the present invention is to provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the building energy conservation and carbon reduction method.

[0049] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute the steps of the building energy conservation and carbon reduction method.

[0050] Another object of the present invention is to provide an information data processing terminal, which is used to implement the building energy saving and carbon reduction system.

[0051] In combination with the above technical solutions and the technical problems solved, please analyze the advantages and positive effects of the technical solutions to be protected by the present invention from the following aspects:

[0052] First, in view of the technical problems existing in the above-mentioned prior art and the difficulty of solving the problems, the technical solutions to be protected by the present invention and the results and data during the research and development process are closely combined to analyze in detail and deeply how the technical solutions of the present invention solve the technical problems, and some creative technical effects brought about after solving the problems. The specific description is as follows:

[0053] The purpose of this invention is to propose an optimization method based on artificial intelligence technologies such as three-dimensional digital modeling, machine learning and deep reinforcement learning, taking into account the influence of environmental factors and the chimney effect of high-rise buildings, to improve the overall comfort during the operation and maintenance of large-scale buildings, while achieving significant carbon reduction. It mainly solves the following problems:

[0054] (1) Use three-dimensional digital modeling technology to create high-precision BIM models to solve the problem that traditional modeling methods are difficult to accurately reflect complex operation and maintenance scenarios.

[0055] (2) A proxy model based on the random forest algorithm is used. This model can accurately predict the carbon emissions and occupant comfort of buildings under various temperature scenarios.

[0056] (3) The external weather conditions are classified and considered, and deep reinforcement learning technology is used to design dynamic air-conditioning operation strategies for different weather conditions, thereby solving the problem that traditional air-conditioning control strategies are subjective and do not take weather changes into consideration.

[0057] (4) In order to address the problem of significant differences in temperature and occupant comfort on different floors of high-rise buildings due to the chimney effect, a floor-based air conditioning operation control strategy is proposed. The set temperature of the air-conditioning terminal of the cooling system is adjusted according to the specific needs of each floor and external weather conditions.

[0058] This paper proposes a multi-scenario building operation and maintenance optimization method that integrates building information modeling (BIM), random forest algorithm and dual deep Q network (DDQN). The key technical points are:

[0059] 1. High-precision BIM model and simulation integration technology: The present invention uses Revit software to construct a three-dimensional BIM model of a building, and seamlessly connects this high-precision model to the DesignBuilder simulation platform to achieve a detailed simulation of the physical properties of the building.

[0060] 2. Construction of random forest prediction agent model: For the two core indicators of carbon emissions and indoor comfort, the present invention designs and implements a prediction model based on the random forest algorithm. The model shows excellent prediction performance, proves the robustness of the model among complex variables, and provides a scientific basis for formulating accurate optimization strategies.

[0061] 3. Multi-objective optimization strategy driven by dual deep Q network (DDQN): Facing the multi-objective challenges of carbon emission reduction and comfort improvement in building operation and maintenance, this paper creatively introduces the DDQN algorithm and applies it to the automatic optimization of building operation and maintenance strategies. It makes full use of the efficient learning mechanism of reinforcement learning in a dynamic environment, and achieves the balanced optimization of minimizing carbon emissions and maximizing comfort through a carefully designed reward mechanism. By intelligently adjusting the operation strategy of air-conditioning equipment, this method significantly reduces the carbon emissions of buildings while ensuring living comfort.

[0062] Second, the technical solution of the present invention fills the technical gap in the industry at home and abroad:

[0063] Through the research and investigation of existing literature and patents, we found that there are already some technical solutions for building energy management on the market, but these solutions are usually not intelligent enough, lack adaptability and learning ability, and cannot update judgments and strategies according to actual conditions; or focus on the optimization of a single system, lacking comprehensive solutions that fully consider the impact of environmental factors and building characteristics. There is no shortage of research on the application of advanced artificial intelligence technology to promote building energy conservation and emission reduction at the forefront of current disciplines, but most of these studies use a single machine learning method to process, and are still relatively insufficient in providing flexible and targeted carbon reduction strategies for buildings.

[0064] In this context, the present invention studies the problems of high carbon emission intensity and poor comfort of personnel caused by the chimney effect of high-rise buildings, and uses advanced computer technologies such as three-dimensional digital modeling and artificial intelligence to propose a method for building energy conservation and carbon reduction in multiple scenarios that combines BIM, machine learning, and deep reinforcement learning. This innovative method not only makes up for the shortcomings of existing technologies in data integration, limited intelligence, environmental factors, and building structure considerations, but also brings advanced and comprehensive technical perspectives to the industry at home and abroad. Through the technical solution of the present invention, we have successfully solved the shortcomings of traditional optimization methods in terms of intelligence, comprehensiveness, and pertinence, and provided theoretical support for carbon reduction in building operation and maintenance.

[0065] The technical solution of the present invention solves a technical problem that people have been eager to solve but have never succeeded in solving: For a long time, a major challenge faced by the construction industry in the operation and maintenance stage is to reduce carbon emissions while ensuring the comfort of the indoor environment. Despite many attempts, few technologies can achieve significant energy-saving effects without sacrificing comfort. The current mainstream building energy management and control technology is still based on historical data and simple prediction models. Although some building automatic control systems can monitor and control various facilities and equipment in the building in real time, they usually perform poorly in integrating multi-source data, rely on pre-set rules and logic, lack adaptability and learning ability, and cannot update judgments and strategies according to actual conditions. In addition, traditional systems rarely consider the impact of external environmental factors (such as weather changes, humidity, etc.) on internal energy use and comfort in buildings.

[0066] The technical solution of the present invention successfully solves the above technical problems by combining advanced technologies such as three-dimensional digital modeling, machine learning and deep reinforcement learning. Specifically, the present invention not only uses three-dimensional digital modeling technology to create a high-precision BIM model to solve the problem that traditional modeling methods are difficult to accurately reflect complex operation and maintenance scenarios; but also considers the classification of external weather, combines random forests and deep reinforcement learning technology, and designs dynamic air-conditioning operation strategies for different weather conditions, solving the subjectivity of traditional air-conditioning control strategies and insufficient consideration of weather changes; in addition, the present invention fully considers the obvious differences in temperature and personnel comfort on different floors in high-rise buildings due to the chimney effect, and proposes a floor-by-floor control strategy for air-conditioning operation, adjusting the temperature set at the air-conditioning terminal of the cooling system according to the specific needs of each floor and external weather conditions.

[0067] In summary, the technical solution of the present invention has made a significant breakthrough in solving the technical problems that people have been eager to solve but have not been able to solve. Its advancement and practicality can not only meet the research and development needs of the current intelligent manufacturing era, but also inject new vitality into the field of building operation and maintenance carbon reduction.

[0068] Third, the present invention addresses the deficiencies in existing building energy conservation and carbon reduction technologies and proposes a more comprehensive and accurate building energy management method. Traditional building energy conservation methods usually rely on static control strategies and limited data sets, and are unable to fully address complex energy consumption and carbon emission issues in different building scenarios, resulting in low energy efficiency and unsatisfactory carbon emission control effects. In addition, existing technologies lack systematic management of the entire life cycle of buildings, and are unable to dynamically respond to changes during the use of buildings, especially in terms of multi-scenario and multi-objective optimization, and have significant limitations.

[0069] The present invention significantly improves the prediction and optimization capabilities of building energy consumption and carbon emissions by introducing high-precision building information modeling (BIM) technology and combining it with simulation. By constructing a BIM model and performing simulation analysis, the system can obtain more accurate and comprehensive building energy consumption data, providing strong data support for subsequent optimization. Compared with traditional building management methods, the present invention can fully simulate and predict a variety of building operation scenarios, significantly improving the understanding of complex building energy consumption behavior and ensuring the accuracy of energy consumption control and carbon emission management.

[0070] In addition, the present invention introduces a random forest algorithm for predicting multi-scenario operation and maintenance of buildings, which solves the shortcomings of traditional methods in processing multi-dimensional data. The random forest algorithm can effectively deal with the energy consumption differences of different floors and scenes through hierarchical modeling and multi-objective optimization, and provide accurate energy consumption prediction and management strategies. Compared with the prior art, the present invention not only improves the accuracy of prediction, but also can flexibly adjust the operation strategy of the building according to different scenarios and needs, optimize building energy consumption, reduce carbon emissions, and realize true intelligent energy management.

[0071] Finally, the present invention further optimizes the carbon reduction effect of the building by strengthening the deep learning algorithm Double-DQN. Through multi-objective optimization analysis, the system can dynamically adjust the cooling temperature of each floor on the premise of balancing comfort and energy consumption to achieve the optimal building operation plan. This innovative control method significantly improves the efficiency of building energy conservation and carbon reduction, solves the problem of difficulty in balancing multi-objective requirements in traditional methods, promotes the development of building energy-saving technology, and has significant social and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 It is a flow chart of the building energy saving and carbon reduction method provided by an embodiment of the present invention.

[0073] Figure 2 It is a structural block diagram of a building energy-saving and carbon-reduction system provided by an embodiment of the present invention.

[0074] Figure 3 The BIM model is established by using the Revit software provided in an embodiment of the present invention: (a) 3D view; (b) vertical view.

[0075] Figure 4 It is a BIM model diagram in the DesignBuilder software provided in an embodiment of the present invention.

[0076] Figure 5 It is a comparison chart of the DesignBuilder simulation and actual carbon emissions provided by an embodiment of the present invention.

[0077] Figure 6 This is a cooling season outdoor temperature division diagram provided by an embodiment of the present invention.

[0078] Figure 7 It is a random forest flow chart provided by an embodiment of the present invention.

[0079] Figure 8 It is a schematic diagram of a DDQN provided by an embodiment of the present invention.

[0080] Fig. 9 It is a BIM model accuracy map provided by an embodiment of the present invention.

[0081] Fig.10 Graph 1 shows the R2 distribution in various scenarios provided by the embodiments of the present invention: (a) 22°C-30°C; (b) 30°C-35°C; (c) >35°C.

[0082] Fig.11 This is the MAE distribution under each scenario provided by the embodiments of the present invention: (a)(b)(c) are the carbon emission prediction errors under three scenarios; (d)(e)(f) are the indoor occupant comfort prediction error diagrams under three scenarios.

[0083] Fig.12 It is a comparison chart of three optimization strategies provided by the embodiment of the present invention and the original strategy.

[0084] Fig.13 It is a technical roadmap of the present invention provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0085] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0086] The following are two specific industrial application examples, which demonstrate the application of the building energy saving and carbon reduction method of the present invention in different fields:

[0087] Example 1: Optimization of energy management system in a large shopping mall

[0088] In large shopping malls, air conditioning systems and lighting systems are usually the main sources of energy consumption. The method of the present invention is applied to optimize the overall energy management of shopping malls. First, the architectural structure and energy consumption of the shopping mall are modeled in detail using Revit software and DesignBuilder simulation tools, and the model is corrected to ensure the accuracy of the data. After building a high-precision BIM model, a data set covering different temperature settings and operating scenarios is created through simulation. Next, the random forest algorithm is used to analyze these data to predict the energy consumption and carbon emissions of different floors and areas under various operating conditions. Finally, the Double-DQN algorithm is used to optimize the air conditioning system of the shopping mall in real time. By dynamically adjusting the cooling temperature of each area, it is ensured that while maintaining customer comfort, the optimal use of energy is achieved and the overall carbon emissions are reduced. This method effectively reduces the operating costs of the shopping mall and improves the level of intelligence in energy management.

[0089] Example 2: Optimization of air conditioning system in smart office buildings

[0090] In smart office buildings, the operating efficiency of the air-conditioning system directly affects the energy consumption and carbon emissions of the building. Through the method of the present invention, a detailed BIM model of the office building is first constructed in Revit, the geometric and performance data of the building are obtained, and simulation correction is performed through DesignBuilder. Subsequently, a multi-scenario operation and maintenance data set covering different floors and temperature settings of the office building is constructed through simulation and random forest algorithm, and the energy consumption and carbon emissions of each floor are predicted. On this basis, the Double-DQN algorithm is used to perform multi-objective optimization on the air-conditioning system of the office building, and the cooling temperature of each floor is dynamically adjusted to balance the comfort and energy consumption of employees. This method not only significantly reduces the energy consumption and carbon emissions of office buildings, but also enhances the intelligence of building management, achieves the goal of green and energy-saving buildings, and helps enterprises fulfill their social responsibilities and reduce operating costs.

[0091] The building energy saving and carbon reduction method of the present invention realizes the optimization of energy consumption and carbon emissions of buildings by constructing a high-precision building information model (BIM), combining simulation and advanced machine learning algorithms. First, by constructing a high-precision building BIM model in Revit software, a three-dimensional model of the building is established according to the actual parameters of the building, including detailed information such as floor plans, structural materials, and room dimensions. Subsequently, the model is exported to gbXML format and imported into DesignBuilder (DB) software for simulation analysis, the geometry and performance information of the model are corrected, its accuracy is ensured, and a reliable data basis is provided for subsequent simulation.

[0092] After completing the high-precision BIM model, simulations were performed to construct a data set for energy conservation and carbon reduction. Specifically, the cooling set temperature of each floor of the building was selected as the key input variable, and carbon emissions and indoor environmental comfort (such as PPD value) were set as output indicators. By designing orthogonal experiments in DesignBuilder to simulate the building energy consumption and carbon emissions under different temperature scenarios, a data set covering a variety of building operation scenarios was finally established, laying the foundation for the subsequent prediction and optimization process.

[0093] Next, the multi-scenario operation and maintenance of the building is predicted based on the random forest algorithm. The random forest algorithm performs layered modeling on the data set and establishes a prediction model for each floor to improve the accuracy of the prediction. 2 ) and mean absolute error (MAE) are used to evaluate the performance of the model on each floor. The stratification strategy of random forest enables the model to more accurately capture the operating characteristics of different floors and provide accurate predictions of building energy consumption and carbon emissions.

[0094] Finally, the building carbon reduction optimization is achieved by strengthening the deep learning algorithm Double-DQN (DDQN). In this process, the DDQN algorithm is introduced for multi-objective optimization, and a balance analysis is performed on the two competing goals of carbon emissions and comfort. The cooling temperature of each floor is used as the state, the change of cooling temperature is used as the action, and the reward is defined as the weighted sum of carbon emissions and comfort indicators. The DDQN algorithm dynamically adjusts the cooling temperature to find the optimal cooling solution in multiple scenarios, thereby achieving the goals of building energy saving and carbon reduction. Compared with the traditional DQN algorithm, DDQN reduces the estimation error and improves the optimization effect through more accurate target value calculation, providing a more intelligent and efficient energy management solution for buildings.

[0095] like Figure 1 As shown, a method for building energy saving and carbon reduction provided by an embodiment of the present invention comprises the following steps:

[0096] S101, establish a high-precision building BIM model.

[0097] (1) Construct the building BIM model in Revit software according to the building parameters. According to the needs and design requirements of the construction, obtain the necessary building parameters such as building floor plans and structural material information. These parameters include but are not limited to the overall layout of the building, the size of each room, the location of doors and windows, the distribution of walls and columns, the type of specific structural materials used (such as concrete, steel bars, glass, etc.) and their properties, and other detailed information. After obtaining the above detailed building parameters, construct the building model in Revit software Figure 3 .

[0098] (2) Export the building geometry data and performance information in Revit as a gbXML file and import it into DesignBuilder (DB) for simulation analysis. DesignBuilder is an open building simulation platform based on EnergyPlus that can handle complex building system models and conduct detailed building energy consumption analysis, including energy consumption of heating, ventilation, and air conditioning systems, as well as lighting and equipment loads, covering multiple aspects such as sunlight analysis, daylighting analysis, thermal comfort assessment, and carbon emission calculation. Figure 4 .

[0099] (3) Correct the building model in DB. In view of the building operation characteristics in the cooling season, the model shows a high degree of simulation realism. Through fine-tuning and reasonable simplification (such as excluding the second and third underground floors and non-crowded areas that contribute less to the total energy consumption), the model effectively focuses on the key spaces that affect energy consumption, ensuring that the accuracy and efficiency of the simulation analysis are given equal importance. In particular, considering the computational complexity in actual operations, we merged physically connected and functionally similar rooms, which greatly reduced the number of computing units and significantly improved the simulation efficiency, while having little impact on the overall simulation accuracy. By adjusting the COP parameters of the air-conditioning equipment and comparing them with the actual data, when the error between the simulated data and the actual data is less than 10%, the model is considered to have high accuracy and fully meets the needs of subsequent research. In this study, the model was compared with the actual data after correction. Figure 5 shown.

[0100] S102, model simulation to construct a data set.

[0101] Given that cooling demand varies significantly with seasonal temperature differences, this study particularly emphasizes the need to consider the adaptability of strategies under different external ambient temperature conditions. Therefore, it is necessary to construct data sets under different temperature ranges to simulate the energy efficiency performance of refrigeration equipment and the corresponding carbon emission levels in various outdoor environments. Based on this, the building performance analysis is divided into three scenarios, corresponding to different outdoor temperature ranges: Scenario 1 (22℃-30℃), Scenario 2 (30℃-35℃), and Scenario 3 (above 35℃). Figure 6 shown.

[0102] This study focuses on specific scenarios in the building operation and maintenance phase, so structural parameters that are determined and immutable during construction are excluded as output variables, and it is noted that the density of people in office buildings remains stable most of the time and does not meet the conditions for dynamic output. Taking comprehensive considerations, this study selects the cooling set temperature of each floor of the building as the key input variable, and sets carbon emissions and indoor environmental comfort as output indicators for evaluating system performance.

[0103] In summary, the cooling temperature of the 11th floor was selected as the input variable, the carbon emissions and the comfort index PPD (Predicted Percentage of Dissatisfied) were selected as the output indicators, an orthogonal experiment was designed under three temperature scenarios, and a data set was established through DB simulation.

[0104] S103, Multi-scenario operation and maintenance prediction of buildings based on random forest.

[0105] Based on the establishment of a complete data set, the random forest algorithm is used to build a prediction model. This study adopts a hierarchical strategy and builds a separate model for each floor to more accurately reflect the heterogeneity and energy efficiency characteristics of each floor and avoid the errors caused by overall modeling.

[0106] Random Forest is a typical machine learning method that has high prediction accuracy, can handle large-scale data sets and high-dimensional feature spaces, and is easy to use and adjust. The random forest algorithm builds multiple decision trees, randomly selects features, and each decision tree is independently generated and trained on a randomly selected training subset until the maximum depth of the tree or other stopping conditions (such as the minimum number of samples, node purity, etc.) are met. Figure 7 This is a flowchart of the random forest algorithm.

[0107] In order to make the model more efficient and more accurate, the use of RandomizedSearchCV (random grid search) requires the optimization of algorithm parameters. RandomizedSearchCV is a very practical tool provided by the scikit-learn library for automatically tuning the parameters of machine learning models (hyperparameter tuning). It cross-validates by randomly selecting parameter combinations, which can efficiently find the optimal parameter settings for the model.

[0108] The prediction performance of the random forest model can be measured by the commonly used indicators for evaluating the performance of regression models. This study uses the coefficient of determination (R 2 ) and mean absolute error (MAE), as defined in formulas (1) and (2).

[0109]

[0110]

[0111] In the formula, y i Represent the predicted value and the actual value respectively. represents the average of the actual values, and n represents the number of samples.

[0112] Based on the data set established in step 2 (149 sets of data in each of the three scenarios), prediction proxy models were established for the carbon emissions and PPD of each floor, and the accuracy of the prediction model was improved through hyperparameter optimization.

[0113] S104, Strengthening deep learning Double-DQN to achieve building carbon reduction optimization.

[0114] This study introduces the Double-DQN (DDQN) method to perform multi-objective optimization analysis on the two indicators of building carbon emissions and comfort, and obtain the optimal cooling solution under each scenario. For the problem in this study, the state should be the cooling temperature of each floor, the action is the change of the cooling temperature of each floor, and the reward is the weighted sum of the standardized carbon emissions and the PPD value.

[0115] Multi-objective optimization tasks involve weighing multiple competing objective functions in a single solution process. It can be defined as formula (3)(4). In this study, optimizing building carbon emissions and comfort is a typical multi-objective optimization task. The weighted summation method is adopted (normalizing different optimization objectives and multiplying them by weight factors to obtain a single objective function, converting the multi-objective problem into a single objective problem for solution), and the weighted sum of carbon emissions and PPD is used as the reward in the DDQN problem for solution.

[0116] miny=f(X)=[f1(x),f2(x),…,f m (x)] T (3)

[0117] st.g i (X)≤0(j=1,2,…,p)

[0118] h k (X)=0(k=1,2,…,q)

[0119]

[0120] X=[x1,x2,…,x n ] T ∈Θ

[0121] Y=[y1,y2,…,y n ] T ∈ψ

[0122] Where m is the number of optimization objective functions, Θ is the n-dimensional search space, and and lower limit Determine; ψ is the m-dimensional vector space of the objective function, determined by Θ and the objective function f(x). Equation g i(X)≤0(j=1,2,…,p) and h k (X) = 0 (k = 1, 2, ..., q) represents p inequality constraints and q equality constraints. In particular, if p = q = 0, the problem is simplified to an unconstrained multi-objective optimization problem.

[0123] DDQN is a deep reinforcement learning framework, an improved version of Q-learning, which is used to solve the problem of decision makers making decisions in uncertain environments. DDQN combines two important technologies: deep neural networks and dual Q-learning to improve the stability and performance of the Q-learning algorithm. The main points of DDQN are as follows:

[0124] (1) Q-learning basics: In reinforcement learning, Q-learning is an algorithm for learning how to choose the best action in a given state. The Q-value (Q-function) represents the expected value of the cumulative reward for taking a specific action in a specific state. Traditional Q-learning algorithms use Q-values ​​to guide decision makers in choosing the best strategy.

[0125] (2) Deep Neural Networks: DDQN introduces deep neural networks as function approximators for estimating Q-value functions. This allows DDQN to handle problems with high-dimensional state spaces and to continuously approximate the Q-value function. Typically, the input to a neural network is the state, and the output is the Q-value of each action.

[0126] (3) Dual Q-learning: The core innovation of DDQN lies in the dual Q-learning method. Traditional Q-learning algorithms estimate the Q value of the best action in the current state, and then use these estimated Q values ​​to guide decision making. However, this can lead to overestimation of Q values. DDQN introduces two neural networks, one for selecting the best action (target network) and the other for estimating Q values ​​(estimation network). These two networks are used alternately to select and estimate Q values ​​to reduce overestimation of Q values.

[0127] (4) Experience Replay: To improve sample utilization and stability, DDQN usually uses experience replay technology. This means that it saves previous experience and then randomly samples from it for training, rather than directly using the current experience for updating. This helps prevent the neural network from falling into a local optimal solution and improves the stability of learning.

[0128] (5) Target Network: DDQN uses two neural networks, one for action selection and the other for estimating Q values. One of the networks is called the target network, which is used to estimate the target Q value to reduce the volatility of the Q value and improve the stability of the algorithm. The parameters of the target network are updated regularly to keep pace with the estimation network.

[0129] Figure 8 This is the algorithm principle of DDQN. Compared with the DQN algorithm, only the method of calculating the target value is changed from formula (5) to formula (6). The algorithm pseudo code is shown in Table 1.

[0130]

[0131]

[0132] Table 1 Double DQN pseudo code

[0133]

[0134] like Figure 2 As shown, a building energy-saving and carbon-reduction system provided by an embodiment of the present invention includes:

[0135] BIM model building module, used to build high-precision building BIM models;

[0136] The simulation module is used to construct data sets for model simulation;

[0137] Prediction module, used for multi-scenario operation and maintenance prediction of buildings based on random forest;

[0138] Optimization module, used to enhance deep learning Double-DQN to achieve building carbon reduction optimization.

[0139] Another object of the present invention is to provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the building energy conservation and carbon reduction method.

[0140] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute the steps of the building energy conservation and carbon reduction method.

[0141] Another object of the present invention is to provide an information data processing terminal, which is used to implement the building energy saving and carbon reduction system.

[0142] The high-rise building operation and maintenance carbon emission optimization control method adopted by the present invention has the following advantages:

[0143] (1) A high-precision building BIM model was constructed. The digital twin model of the Xi'an Building was constructed using BIM technology. The model not only depicts the physical structure of the building in detail, but also simulates key operational details such as personnel flow and equipment operation. By strategically simplifying the model, the model was ensured to be both efficient and accurate. The comparative analysis of the calibrated model and the measured data showed that the model prediction error was only 0.10%, highlighting the extremely high consistency between its prediction results and the actual situation, verifying the high accuracy of the model calibration, which met the high standards required for in-depth research.

[0144] Fig. 9 BIM model accuracy

[0145] (2) The prediction proxy model based on random forest has high accuracy. The prediction model is constructed using the random forest algorithm. Through hyperparameter optimization (see Table 2), the prediction accuracy of the model in the training set and test set is higher than 0.99 and 0.9, respectively. Fig.10 The mean absolute error (MAE) of the model remains stable, as shown in Fig.11 As shown, the model has strong robustness and generalization ability.

[0146] Table 2 Random forest hyperparameter settings under each scenario

[0147]

[0148] (3) The Double-DQN algorithm is used to find the optimal operation and maintenance strategy for air conditioning under different temperature scenarios. While ensuring the comfort of personnel, the building carbon emissions are significantly optimized, with optimization rates reaching 42.5%, 10.13% and 12.32% respectively. The Double-DQN algorithm implements efficient multi-objective optimization for the two core factors in building operation and maintenance: carbon emissions and indoor comfort. The Euclidean distance-based evaluation method is used to calculate the Euclidean distance between the optimization strategy and the origin (0, 0), and then compared with the initial default strategy to obtain the overall optimization rate.

[0149] The specific optimization results are shown in Table 3. Through the optimization strategy, not only a good thermal comfort environment inside the building is successfully maintained, but also the efficiency of energy use is greatly improved, the environmental burden is reduced, and the effectiveness and practical value of the optimization strategy are fully verified.

[0150] Table 3 DDQN optimization results

[0151]

[0152] Fig.12 Comparison of three optimization strategies with the original strategy.

[0153] Fig.13It is a technical roadmap of the present invention provided by an embodiment of the present invention.

[0154] The technical solution of the present invention is mainly used in the field of energy management and operation and maintenance optimization of high-rise office buildings. Specific application areas or related products include:

[0155] New high-rise office buildings: The technical solution of the present invention can be introduced in the architectural design stage, and the future energy consumption and carbon emissions can be predicted and optimized through the establishment of a high-precision BIM model.

[0156] Renovation of existing high-rise office buildings: For existing high-rise office buildings, the technical solution of the present invention can help identify and improve deficiencies in energy management and achieve energy conservation and emission reduction through intelligent operation and maintenance strategies.

[0157] Intelligent building management system: The present invention can be integrated into an intelligent building management system to provide building managers with a complete set of energy consumption monitoring and optimization tools to help them achieve efficient and green operation and maintenance management.

[0158] Building energy consulting service: The present invention can also be used as one of the service contents of professional consulting companies to provide customers with building energy management consulting services based on data analysis and help customers formulate reasonable energy-saving and emission reduction strategies.

[0159] Taking Xi'an Building as an example, the application of the technical solution of the present invention in actual projects is demonstrated:

[0160] High-precision BIM model establishment: Based on the building parameters of the building, a high-precision BIM model was constructed in Revit software. By collecting the building's construction drawings, structural material information and other building parameters, a three-dimensional model that is highly consistent with the actual building was established.

[0161] Model simulation to build data sets: Based on actual operation and maintenance data, simulations were conducted in multiple scenarios through the BIM model to generate data sets that reflect the different operating states of the building in multiple scenarios.

[0162] Multi-scenario building operation and maintenance prediction based on random forest: A carbon emission and comfort prediction model under different temperature scenarios was built using the random forest algorithm, providing basic data support for subsequent optimization strategies.

[0163] Strengthen deep learning to achieve building carbon reduction optimization: Deep reinforcement learning technology, namely the Double-DQN algorithm, is used to find the best operation strategy under different environmental scenarios, achieving overall comfort improvement and significant carbon reduction during building operation and maintenance.

[0164] In order to prove the effectiveness and advantages of the technical solution of the present invention, the following are some specific technical effect evidences:

[0165] High-precision BIM model establishment: Figure 5 As shown in the figure, by comparing the BIM model constructed in the DesignBuilder software with the actual data, the simulation error is only 0.1%, which proves the accuracy and reliability of the model building of the present invention. Figure 4 .

[0166] Accuracy of the prediction model: Figure 6 As shown in Figure 2, the ambient temperature is divided into three scenarios. The accuracy of the random forest prediction model in the training set and the test set in each scenario is higher than 0.99 and 0.9 respectively. Fig.10 , 11 , which shows that the prediction model of the present invention can accurately predict carbon emissions and comfort under different environmental conditions.

[0167] Carbon reduction optimization effect: Use deep reinforcement learning technology to find the best operation strategy under different environmental scenarios, such as Fig.12 As shown, in the three temperature scenarios of 22-30°C, 30-35°C and above 35°C, the overall optimization rates of carbon emissions and comfort based on Euclidean distance reached 42.5%, 10.13% and 12.32% respectively. These data show that the technical solution of the present invention has a significant effect in achieving carbon reduction optimization.

[0168] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0169] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for energy saving and carbon reduction in buildings, characterized in that: The following steps are involved: Step 1: Establish a high-precision building BIM model; Step 2, model simulation to construct a data set; Step 3: Multi-scenario building operation and maintenance prediction based on random forest; Step 4: Strengthen deep learning Double-DQN to achieve building carbon reduction optimization.

2. The building energy saving and carbon reduction method according to claim 1, characterized in that: The establishment of a high-precision building BIM model: (1) Construct a building BIM model in Revit software based on building parameters; (2) Export the building geometry data and performance information in Revit as a gbXML file and import it into DesignBuilder (DB) for simulation analysis; (3) Correct the building model in DB.

3. The building energy saving and carbon reduction method according to claim 2, characterized in that: The construction of the building BIM model in Revit software according to the building parameters: according to the needs and design requirements of the construction, the necessary building parameters of the building plan drawings and structural material information are obtained; these parameters include but are not limited to the overall layout of the building, the size of each room, the location of doors and windows, the distribution of walls and columns, the type of specific structural materials used and detailed information on their properties; after obtaining the above detailed building parameters, the building model is constructed in Revit software.

4. The building energy saving and carbon reduction method according to claim 1, characterized in that: The model simulation constructs a data set: The cooling set temperature of each floor of the building was selected as the key input variable, while carbon emissions and indoor environmental comfort were set as output indicators for evaluating system performance; The cooling temperature of the 11th floor was selected as the input variable, and the carbon emissions and comfort index PPD were selected as the output indicators. An orthogonal experiment was designed under three temperature scenarios, and a data set was established through DB simulation.

5. The building energy saving and carbon reduction method according to claim 1, characterized in that: The random forest-based building multi-scenario operation and maintenance prediction: On the basis of establishing a complete data set, the random forest algorithm was used to build a prediction model; a stratified strategy was adopted to build a separate model for each floor; The coefficient of determination (R 2 ) and Mean Absolute Error (MAE), as defined in formulas (1) and (2); 6. The building energy saving and carbon reduction method according to claim 1, characterized in that: The enhanced deep learning Double-DQN achieves building carbon reduction optimization: The Double-DQN (DDQN) method is introduced to conduct multi-objective optimization analysis on the two indicators of building carbon emissions and comfort, and the optimal cooling solution under each scenario is obtained; The state should be the cooling temperature of each floor, the action should be the change of the cooling temperature of each floor, and the reward should be the weighted sum of the standardized carbon emissions and the PPD value; Multi-objective optimization tasks involve weighing multiple competing objective functions in a single solution process; they can be defined as formula (3)(4); min y=f(X)=[f1(x),f2(x),…,f m (x)] T (3) st.g i (X)≤0(j=1,2,…,p) h k (X)=0(k=1,2,…,q) X=[x1,x2,…,x n ] T ∈Θ Y=[y1,y2,…,y n ] T ∈ψ The algorithm principle of DDQN; Compared with the DQN algorithm, only the method of calculating the target value is changed from formula (5) to formula (6); 7. A building energy-saving and carbon-reduction system for implementing the building energy-saving and carbon-reduction method according to any one of claims 1 to 6, characterized in that: The building energy saving and carbon reduction system comprises: BIM model building module, used to build high-precision building BIM models; The simulation module is used to construct data sets for model simulation; Prediction module, used for multi-scenario operation and maintenance prediction of buildings based on random forest; Optimization module, used to enhance deep learning Double-DQN to achieve building carbon reduction optimization.

8. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the building energy conservation and carbon reduction method according to any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the building energy conservation and carbon reduction method according to any one of claims 1 to 6.

10. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the building energy-saving and carbon-reduction system as described in claim 7.

Images