Building energy consumption prediction and dynamic energy saving optimization method based on BIM

Abstract

The invention discloses a BIM (Building Information Modeling)-based building energy consumption prediction and dynamic energy-saving optimization method, which comprises the following steps of: extracting key characteristic parameters of geometry, materials and construction of a building from a BIM model, and obtaining multi-dimensional data of building types, structural characteristics, material attributes and environmental factors; cleaning the acquired multi-dimensional data, removing repeated, wrong or abnormal data points, and performing standardization processing through a minimum-maximum scaling method or performing normalization processing through a z-score standardization method; the adjusted AI model is used for building energy consumption prediction, cost models of material, labor force, maintenance and energy cost factors are combined, and a dynamic energy-saving strategy is generated through a genetic algorithm; the generated dynamic energy-saving strategy is converted into a specific system adjusting instruction to be applied to an actual building, and energy-saving control is achieved by adjusting the air volume of a building ventilation system, the brightness of an illumination system and the temperature set value of a heating ventilation air-conditioning system.

Description

Technical Field

[0001] The present invention relates to the technical field of building energy consumption optimization, and in particular to a building energy consumption prediction and dynamic energy-saving optimization method based on BIM. Background Art

[0002] In the process of using BIM models and artificial intelligence technology to predict and optimize building energy consumption, there is a key technical problem: how to effectively deal with dynamic changes in the use of buildings while ensuring prediction accuracy. Specifically, the actual use of buildings often deviates from the design stage, such as personnel flow, equipment usage frequency, indoor layout adjustment, etc. These dynamic changes will significantly affect the energy consumption pattern. Traditional static models are difficult to accurately capture these changes, and overly complex dynamic models may cause excessive computing loads and affect real-time optimization effects. Therefore, it is necessary to design an adaptive model structure that can dynamically absorb and integrate multi-source heterogeneous data from building management systems, IoT sensors, etc. while maintaining computing efficiency, and automatically adjust model parameters and structures according to data characteristics. At the same time, it is also necessary to consider how to quickly build an initial model for new or renovated buildings under limited historical data conditions through technologies such as transfer learning or small sample learning, and continuously optimize during operation. The solution to this technical problem will directly affect the practicality and promotion and application of building energy consumption prediction and optimization systems based on artificial intelligence. Summary of the invention

[0003] The present invention provides a BIM-based building energy consumption prediction and dynamic energy-saving optimization method to solve technical problems such as the difficulty of traditional static models in accurately capturing changes in the actual use of buildings, and overly complex dynamic models may lead to excessive computing loads and affect real-time optimization effects.

[0004] To achieve the above object, the present invention adopts the following technical solution: The present invention provides a method for predicting building energy consumption and dynamic energy saving optimization based on BIM, comprising the following steps: S1, data extraction: extract key characteristic parameters of building geometry, materials, and construction from the BIM model to obtain multi-dimensional data on building type, structural characteristics, material properties, and environmental factors; S2, data processing: cleaning the acquired multidimensional data, removing duplicate, erroneous or abnormal data points, and normalizing them by the minimum-maximum scaling method or the z-score normalization method; S3, determine the model: according to the characteristics of the preprocessed data and the complexity of the problem, select the neural network, support vector machine, random forest AI model, determine the input layer, hidden layer and output layer structure for the neural network model, select the kernel function for the support vector machine, and determine the number of decision trees for the random forest; S4, algorithm selection: the BIM model parameter data and historical energy consumption data after cleaning and preprocessing are used as training samples, the neural network model is trained using the back propagation algorithm, the support vector machine is trained using the sequence minimum optimization algorithm, and the random forest is trained using the bootstrap sampling method. At the same time, cross-validation and regularization techniques are used to prevent overfitting; S5, model validation: Use the reserved validation set to validate the trained model, evaluate the prediction accuracy and robustness of the model, and adjust the weights and biases of the neural network, the penalty parameters of the support vector machine, and the tree depth of the random forest based on the validation results; S6, Generate dynamic energy-saving strategies: Use the adjusted AI model for building energy consumption prediction, combine the cost model of material, labor, maintenance, and energy cost factors, and use genetic algorithms to generate dynamic energy-saving strategies; S7, instruction conversion: convert the generated dynamic energy-saving strategy into specific system adjustment instructions, apply them to actual buildings, and achieve energy-saving control by adjusting the air volume of the building ventilation system, the brightness of the lighting system, and the temperature setting value of the HVAC system; S8, monitoring optimization: collect real-time monitoring data, including energy consumption data and environmental parameters, integrate these data with the original training data set, use online learning algorithms to update AI model parameters, continuously optimize prediction accuracy and energy-saving effects, and form a continuous improvement cycle system.

[0005] Preferably, S2 comprises the following steps: Obtain multi-dimensional data on building geometry, materials and construction in the BIM model, extract key parameters through Revit API or Dynamo tools, and obtain the original data set; pre-process the original data set, use data cleaning technology to remove duplicate and erroneous or abnormal data points, and obtain the cleaned data set; According to the distribution characteristics of the cleaned data set, the minimum-maximum scaling method is used to scale the data to the [0, 1] interval, or the z-score standardization method is used to convert the data into a distribution with a mean of 0 and a variance of 1 to obtain a standardized data set; The standardized data set is divided into training subset, validation subset and test subset according to the preset ratio; according to the characteristics of the building energy consumption prediction problem, a hybrid deep learning model including convolutional neural network, recurrent neural network and long short-term memory network is constructed. The hybrid deep learning model can capture the long-term dependency in time series data; The training subset is input into the hybrid deep learning model, the predicted value is calculated through forward propagation, and the loss function value is calculated with the actual value; the gradient of the loss function to the model parameters is calculated using the back propagation algorithm, and the model parameters are updated using the optimization algorithm to minimize the loss function, thus obtaining the trained building energy consumption prediction model; Input the verification subset into the building energy consumption prediction model, calculate the accuracy index of the energy consumption prediction, and when the accuracy index reaches a preset threshold, determine the building energy consumption prediction model as the final model; The test subset is input into the final model to obtain the building energy consumption prediction results, and the prediction results are output to the building energy consumption management system.

[0006] Preferably, the training subset, validation subset and test subset are divided according to the ratio of 70%, 15% and 15%; when the model performance on the validation subset meets the preset standard, the model is finally tested using the test subset data to obtain the prediction accuracy of the model on real unknown data; the test results are used to determine whether the model can be deployed and applied. If the model performance is poor, it is necessary to return to the data preprocessing or model design stage for optimization.

[0007] Preferably, S3 comprises the following steps: According to the characteristics of the pre-processed building energy consumption data and the complexity of the problem, the decision tree algorithm is used to analyze the data to obtain the key characteristics and distribution of the data; If the data has nonlinear characteristics and high dimensions, the neural network model is selected for energy consumption prediction. When the neural network model is selected, the number of input layer nodes and output layer nodes of the neural network is determined by analyzing the dimensions and characteristics of the data, and the number of hidden layers and the number of nodes in each layer are set based on experience; If the data dimension is low and has obvious classification boundaries, the support vector machine model is selected. For the support vector machine model, different kernel functions are evaluated by cross-validation method to select the kernel function type and corresponding hyperparameters with the best prediction performance; If there are a large number of discrete variables in the data, the random forest model is used. When using the random forest model, the number of decision trees, the maximum depth, and the minimum number of samples for splitting nodes are optimized through grid search and other methods to obtain the optimal model structure. According to the selected model type, construct the training data set and validation data set of the energy consumption prediction model, and set the appropriate training batch size and number of iterations; during the model training process, monitor the changes in the loss function values ​​of the training set and validation set to determine whether the model is overfitting or underfitting, and adjust the model structure or hyperparameters according to the situation; after the model training is completed, use the test data set to evaluate the prediction performance of the model, obtain indicators such as the accuracy and mean square error of the energy consumption prediction, and verify the generalization ability of the model; If the model prediction performance meets the requirements, the trained model will be deployed to the building energy consumption management system. Through the real-time collected building operation data, the future energy consumption trend can be dynamically predicted to provide support for energy-saving optimization decisions.

[0008] Preferably, S6 comprises the following steps: Using artificial neural network algorithms, deep learning training is performed on the building energy consumption database to obtain an AI model that can accurately predict building energy consumption; The model comprehensively considers multiple influencing factors such as window-to-wall ratio, building orientation and thermal performance of the external envelope structure. It obtains energy consumption changes by real-time monitoring of building operation data and feeds them back to the AI ​​energy consumption prediction model. The AI ​​energy consumption prediction model is adjusted and optimized using the feedback data to improve the prediction accuracy and adaptability of the model. The optimized AI model is applied to building energy consumption prediction to obtain energy consumption prediction results. According to the energy consumption prediction results, combined with the cost model considering factors such as materials, labor, maintenance and energy, an energy-saving optimization objective function is constructed; a genetic algorithm is used to solve the energy-saving optimization objective function, and the optimal combination of energy-saving strategies is searched through operations such as selection, crossover and mutation; the optimal solution is selected from the solution set of the genetic algorithm and used as the dynamic energy-saving optimization strategy for the building, which includes adjusting the operating parameters of systems such as ventilation, lighting and air conditioning; the dynamic energy-saving optimization strategy is applied to the building energy management system, and real-time adjustment of energy-saving measures is achieved through automated control, and the energy-saving effect is continuously monitored to form a closed-loop optimization.

[0009] Preferably, S7 comprises the following steps: Extract relevant parameters of the building ventilation system, lighting system and HVAC system from the BIM model, including equipment type, specifications and layout location, and build an equipment parameter database; Combine the output of the AI ​​energy consumption prediction model with the dynamic energy-saving optimization algorithm according to the predicted energy consumption; Determine the energy-saving control targets for each system, such as the target air volume of the ventilation system, the target brightness of the lighting system, and the target temperature setting value of the HVAC system; Based on the information in the equipment parameter database and in combination with the energy-saving control target, an optimization algorithm is used to calculate the optimal control parameters of each system, such as the fan speed of the ventilation system, the dimming ratio of the lamps of the lighting system, and the compressor frequency of the HVAC system; the calculated optimal control parameters are converted into specific equipment control instructions, such as the frequency setting value of the fan inverter, the dimming signal of the lamp dimming controller, and the temperature setting value of the air conditioning controller; Continuously monitor the operating status and energy consumption data of each system, compare the actual energy consumption with the predicted energy consumption, and calculate the energy-saving effect; if there is a large deviation between the actual energy-saving effect and the expected target, the deviation data will be fed back to the dynamic energy-saving optimization algorithm, and the control strategy will be continuously optimized through machine learning to improve the energy-saving effect.

[0010] Preferably, performance diagnosis and fault detection are performed on the equipment of each system on a regular basis, potential faults and performance degradation of the equipment are identified through big data analysis technology, and maintenance plans are formulated in advance to ensure efficient operation of the system; the optimized control strategy and energy-saving effect data are uploaded to the cloud platform, and decision support is provided for building energy management through data mining and visualization technology, so as to continuously improve and optimize the entire BIM-AI energy consumption prediction and energy-saving optimization framework.

[0011] Preferably, S8 comprises the following steps: The IoT sensors collect building energy consumption data and environmental parameters in real time, including indoor temperature and humidity, lighting intensity, and equipment operation status, to form a real-time monitoring data stream. The real-time monitoring data is integrated with the building feature data extracted from the BIM model to construct a comprehensive data set that includes energy consumption data, environmental parameters, and building features. Preprocess the comprehensive data set, including data cleaning and normalization, to ensure data quality and consistency, in preparation for subsequent model updates; based on the preprocessed comprehensive data set, use online learning algorithms to incrementally train and update parameters of the original AI energy consumption prediction model, so that the model can adapt to changes in building operation status; During the model update process, the prediction accuracy and generalization ability of the model are evaluated through cross-validation and other technologies, and the model structure and hyperparameters are dynamically adjusted according to the evaluation results to obtain the best prediction performance; the updated AI energy consumption prediction model is applied to the actual building energy consumption prediction, and the prediction results are compared with the actual energy consumption data to calculate the prediction error; if the prediction error exceeds the preset threshold, the model automatic optimization mechanism is triggered, and the model parameters are fine-tuned through optimization algorithms such as gradient descent to further improve the prediction accuracy; Based on the optimized AI energy consumption prediction results, combined with the cost model and optimization algorithm, energy-saving strategies are dynamically generated and implemented in the building's lighting, air conditioning, ventilation and other subsystems through intelligent control systems to achieve real-time energy-saving optimization; Continuously monitor changes in building energy consumption and the effectiveness of energy-saving measures, and reincorporate feedback data into the comprehensive data set to form a closed-loop system of data-prediction-optimization-feedback to achieve continuous improvement in building energy consumption management.

[0012] The beneficial effects of the present invention are embodied in: The present invention provides a method for predicting building energy consumption and dynamic energy-saving optimization based on BIM. The method extracts multi-dimensional data from the BIM model, and after data cleaning and standardization preprocessing, selects a suitable AI model for training. The historical energy consumption data is used as a training sample, and cross-validation and other technologies are used to prevent overfitting, and the model parameters are optimized according to the verification results; combined with the cost model, the present invention uses a genetic algorithm to generate a dynamic energy-saving strategy, and converts it into a specific system adjustment instruction for application in actual buildings; by collecting real-time monitoring data and integrating the original training data, the present invention uses an online learning algorithm to continuously update the model, continuously optimize the prediction accuracy and energy-saving effect, and form a self-improving cycle system; the present invention realizes the intelligent prediction and dynamic optimization control of building energy consumption, and effectively improves the energy utilization efficiency of the building.

[0013] Other features and advantages of the present invention will be set forth in the following description, and in part will become apparent from the description, or may be understood by practicing the present invention; the main purpose and other advantages of the present invention may be realized and obtained through the solutions particularly pointed out in the description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0015] The technical solution of the present invention is described in detail below through examples. The following examples are merely exemplary and can only be used to explain and illustrate the technical solution of the present invention, but cannot be interpreted as limiting the technical solution of the present invention.

[0016] like Figure 1 The present embodiment provides a BIM-based building energy consumption prediction and dynamic energy-saving optimization method, which may specifically include: S1, data extraction: extract key characteristic parameters of building geometry, materials, and construction from the BIM model to obtain multi-dimensional data on building type, structural characteristics, material properties, and environmental factors; S2, data processing: cleaning the acquired multi-dimensional data, removing duplicate, erroneous or abnormal data points, and normalizing them by the minimum-maximum scaling method or the z-score normalization method, including: Obtain multi-dimensional data of building geometry, materials and construction in the BIM model, extract key parameters through Revit API or Dynamo tools, and obtain the original data set; pre-process the original data set, use data cleaning technology to remove duplicate and erroneous or abnormal data points, and obtain the cleaned data set; according to the distribution characteristics of the cleaned data set, use the minimum-maximum scaling method to scale the data to the [0, 1] interval, or use the z-score standardization method to convert the data into a distribution with a mean of 0 and a variance of 1, and obtain a standardized data set; divide the standardized data set into a training subset, a validation subset and a test subset according to a preset ratio; divide the training subset, validation subset and test subset according to the ratio of 70%, 15% and 15%; when the model performance on the validation subset meets the preset standard, use the test subset data to conduct a final test on the model to obtain the prediction accuracy of the model on real unknown data; The test results determine whether the model can be deployed and applied. If the model performance is poor, it is necessary to return to the data preprocessing or model design stage for optimization; according to the characteristics of the building energy consumption prediction problem, a hybrid deep learning model including convolutional neural network, recurrent neural network and long short-term memory network is constructed. The hybrid deep learning model can capture the long-term dependencies in time series data; the training subset is input into the hybrid deep learning model, the predicted value is calculated through forward propagation, and the loss function value is calculated with the actual value; the back propagation algorithm is used to calculate the gradient of the loss function to the model parameters, and the optimization algorithm is used to update the model parameters and minimize the loss function to obtain the trained building energy consumption prediction model; the verification subset is input into the building energy consumption prediction model to calculate the accuracy index of the energy consumption prediction. When the accuracy index reaches the preset threshold, the building energy consumption prediction model is determined as the final model; the test subset is input into the final model to obtain the building energy consumption prediction result, and the prediction result is output to the building energy consumption management system.

[0017] For example, obtain multi-dimensional data such as building geometry, materials, and structures in the BIM model, and extract key parameters through tools such as Revit API or Dynamo. Preprocess the extracted multi-dimensional data, and use data cleaning technology to remove duplicate, erroneous or abnormal data points to ensure the accuracy and consistency of the data. According to the distribution characteristics of the data, select an appropriate standardization method to process the data, such as the minimum-maximum scaling method to scale the data to the [0,1] interval, or use the z-score standardization method to convert the data into a distribution with a mean of 0 and a variance of 1. Divide the preprocessed data into a training set, a validation set, and a test set according to a certain ratio, usually using a 70%, 15%, and 15% division method to obtain a data subset for model training and evaluation. According to the characteristics of the building energy consumption prediction problem, select an appropriate AI model, such as a convolutional neural network, a recurrent neural network, a long short-term memory network, etc., to build a deep learning model that can capture long-term dependencies in time series data. Design the network structure of the model, determine the number of neurons and the connection mode of the input layer, hidden layer and output layer, select appropriate activation functions such as ReLU, Sigmoid, etc., and loss functions such as mean square error, mean absolute error, etc. Input the training data into the model, calculate the predicted value through forward propagation, and calculate the loss function value with the actual value. Use the backpropagation algorithm to calculate the gradient of the loss function to the model parameters, and use optimization algorithms such as gradient descent, Adam, etc. to update the model parameters and minimize the loss function. Iterate the process of forward propagation, loss calculation, backpropagation and parameter update until the performance indicators of the model on the validation set such as mean square error or mean absolute error converge or reach the preset number of training rounds. Use the test set data to evaluate the trained model and calculate the accuracy indicators of energy consumption prediction such as mean absolute percentage error, determination coefficient, etc. If the model performance meets the requirements, it will be deployed in the building energy consumption management system for practical application; otherwise, adjust the model structure or hyperparameters, and retrain and optimize.

[0018] The Revit API is used to extract multi-dimensional data such as building geometry, materials, and structures in the BIM model, such as the length, width, height, volume, exterior wall area, window-to-wall ratio, and orientation of the building, and the material properties of components such as walls, roofs, doors, and windows, such as thermal conductivity, specific heat capacity, and density. The extracted data is cleaned using Python's Pandas library to remove duplicate records and outliers such as data with negative areas to ensure data integrity and consistency. The cleaned data is subjected to minimum-maximum normalization, and each dimension of the data is scaled to the [0,1] interval to eliminate the dimensional effect. The data set is randomly divided into 70%, 15%, and 15% ratios to obtain a training set of 5,000 samples, a validation set of 1,000 samples, and a test set of 1,000 samples. A convolutional neural network model consisting of 3 convolutional layers, 2 pooling layers, and 2 fully connected layers is constructed, with the input being the multi-dimensional building data and the output being the energy consumption prediction value. The convolution kernel size of the convolution layer is 3x3, the activation function is ReLU, the pooling layer uses maximum pooling, and the number of neurons in the fully connected layer is 128 and 1. The loss function of the model selects the mean absolute error MAE, the optimization algorithm uses Adam, and the learning rate is set to 0.01. The training set data is fed into the model, and iterates for 500 rounds. After each round, the model performance is evaluated on the validation set. When the validation set MAE does not decrease for 10 consecutive rounds, the training is stopped. The energy consumption prediction performance of the model is evaluated on the test set. The mean absolute percentage error MAPE is 2%, and the determination coefficient R² is 93, indicating that the model can accurately predict the building energy consumption. Finally, the trained model is integrated into the energy consumption management system, and the building operation data is collected in real time. The model is input for energy consumption prediction and optimization control to improve the building energy efficiency.

[0019] S3, determine the model: according to the characteristics of the preprocessed data and the complexity of the problem, select the neural network, support vector machine, random forest AI model, determine the input layer, hidden layer and output layer structure for the neural network model, select the kernel function for the support vector machine, and determine the number of decision trees for the random forest, including: According to the characteristics of the pre-processed building energy consumption data and the complexity of the problem, the decision tree algorithm is used to analyze the data to obtain the key characteristics and distribution of the data; if the data has nonlinear characteristics and high dimensions, the neural network model is selected for energy consumption prediction. When the neural network model is selected, the number of input layer nodes and output layer nodes of the neural network is determined by analyzing the dimensions and characteristics of the data, and the number of hidden layers and the number of nodes per layer are set based on experience; if the data dimension is low and has obvious classification boundaries, the support vector machine model is selected. For the support vector machine model, different kernel functions are evaluated by cross-validation method, and the kernel function type and corresponding hyperparameters with the best prediction performance are selected; if there are a large number of discrete variables in the data, the random forest model is used. When using the random forest model, the decision is optimized through grid search and other methods The optimal model structure is obtained by adjusting hyperparameters such as the number of trees, maximum depth, and minimum number of samples for splitting nodes. According to the selected model type, the training data set and validation data set of the energy consumption prediction model are constructed, and the appropriate training batch size and number of iterations are set. During the model training process, the loss function value changes of the training set and validation set are monitored to determine whether the model is overfitting or underfitting, and the model structure or hyperparameters are adjusted according to the situation. When the model training is completed, the prediction performance of the model is evaluated using the test data set to obtain indicators such as the accuracy and mean square error of the energy consumption prediction to verify the generalization ability of the model. If the model prediction performance meets the requirements, the trained model is deployed to the building energy consumption management system, and the future energy consumption change trend is dynamically predicted through the real-time collected building operation data to provide support for energy-saving optimization decisions.

[0020] For example, the geometric parameters, material properties and other multi-dimensional data of the building are extracted through the Revit API, and the decision tree algorithm is used to analyze the data. It is found that the data presents obvious nonlinear characteristics and has a high characteristic dimension. Therefore, the BP neural network model is selected for energy consumption prediction. According to the data analysis results, the number of input layer nodes of the neural network is determined to be 20, the number of output layer nodes is 1, and the hidden layer is set to 2 layers, each layer contains 50 and 30 nodes respectively. The data is divided into a training set and a test set in a ratio of 8:2, and the data is normalized. The training batch size is set to 32, the number of iterations is 500, the learning rate is 01, and the Adam optimization algorithm and the mean square error loss function are used. During the training process, the loss function value changes of the training set and the validation set are monitored by the TensorBoard visualization tool. It is found that after 300 iterations, the loss function value of the validation set begins to show an upward trend, indicating that the model is overfitting, so the number of iterations is adjusted to 300. The performance of the trained model is evaluated using the test set. The results show that the energy consumption prediction accuracy of the model reaches 95%, and the mean square error is 02, which meets the requirements of practical applications. Finally, the model is deployed in the building energy consumption management system to collect building operation data in real time, and predict the energy consumption change trend in the next 24 hours every hour, providing a decision-making basis for building energy conservation optimization.

[0021] S4, algorithm selection: the BIM model parameter data and historical energy consumption data after cleaning and preprocessing are used as training samples, the neural network model is trained using the back propagation algorithm, the support vector machine is trained using the sequence minimum optimization algorithm, and the random forest is trained using the bootstrap sampling method. At the same time, cross-validation and regularization techniques are used to prevent overfitting; S5, model validation: Use the reserved validation set to validate the trained model, evaluate the prediction accuracy and robustness of the model, and adjust the weights and biases of the neural network, the penalty parameters of the support vector machine, and the tree depth of the random forest based on the validation results; S6, Generate dynamic energy-saving strategies: Use the adjusted AI model for building energy consumption prediction, combine the cost model of material, labor, maintenance, and energy cost factors, and use genetic algorithms to generate dynamic energy-saving strategies, including: The artificial neural network algorithm is used to conduct deep learning training on the building energy consumption database to obtain an AI model that can accurately predict building energy consumption. The model comprehensively considers multiple influencing factors such as window-to-wall ratio, building orientation and thermal performance of the external envelope structure. By real-time monitoring of building operation data, energy consumption changes are obtained and fed back to the AI ​​energy consumption prediction model. The AI ​​energy consumption prediction model is adjusted and optimized using the feedback data to improve the prediction accuracy and adaptability of the model. The optimized AI model is applied to building energy consumption prediction to obtain energy consumption prediction results. According to the energy consumption prediction results, combined with the cost model that considers factors such as materials, labor, maintenance and energy, an energy-saving optimization objective function is constructed. A genetic algorithm is used to solve the energy-saving optimization objective function, and the optimal combination of energy-saving strategies is searched through operations such as selection, crossover and mutation. The optimal solution is selected from the solution set of the genetic algorithm and used as the dynamic energy-saving optimization strategy for the building. The strategy includes adjusting the operating parameters of systems such as ventilation, lighting and air conditioning. The dynamic energy-saving optimization strategy is applied to the building energy management system, and real-time adjustment of energy-saving measures is achieved through automated control. The energy-saving effect is continuously monitored to form a closed-loop optimization.

[0022] For example, first, the geometric information of the building is extracted from the BIM model, such as the building area is 10,000 square meters, the window-to-wall ratio is 3, the building faces south, the exterior wall uses 200mm thick aerated concrete blocks, and the exterior insulation uses 100mm thick rock wool boards. At the same time, the material information of the building is extracted, such as the solar radiation absorption coefficient of the exterior wall paint is 6, the roof uses SBS waterproof membrane, and the thermal conductivity is 17W / (m•K). The extracted information is used to construct a multi-dimensional building energy consumption database including building type, structural characteristics, material properties, environmental factors, etc. Then, the BP neural network algorithm is used to train the building energy consumption database, setting the number of input layer nodes to 10, the number of hidden layer nodes to 8, the number of output layer nodes to 1, the learning rate to 01, and the activation function to the Sigmoid function. After 1,000 iterations of training, an AI model that can accurately predict building energy consumption is obtained. By real-time monitoring of building operation data, such as indoor temperature, humidity, and illumination, the energy consumption changes are obtained and input into the AI ​​energy consumption prediction model as feedback data. The model weights are adjusted using the gradient descent algorithm to continuously improve the prediction accuracy of the model, and finally a dynamic and adaptive building energy consumption prediction model is obtained. The model is used to predict the energy consumption of the building in the next week, and the predicted value is 12000kWh. According to the energy consumption prediction results, combined with cost factors such as materials, labor, maintenance, and energy, an energy-saving optimization objective function is constructed, such as the objective function is to minimize the cost of energy-saving transformation, and the constraint condition is that the energy saving rate is not less than 20%. The genetic algorithm is used to solve the objective function, setting the population size to 50, the crossover probability to 6, and the mutation probability to 01. After 100 iterations, the optimal energy-saving strategy combination is obtained, such as setting the air-conditioning temperature to 26 degrees Celsius and using LED lighting, the energy saving rate can reach 30%. Finally, the optimal energy-saving strategy is applied to the building energy management system, and the real-time adjustment of energy-saving measures is realized through automatic control, and the energy-saving effect is continuously monitored to form a closed-loop optimization and continuously improve the building energy efficiency level.

[0023] S7, command conversion: convert the generated dynamic energy-saving strategy into specific system adjustment commands, apply them to actual buildings, and achieve energy-saving control by adjusting the air volume of the building ventilation system, the brightness of the lighting system, and the temperature setting value of the HVAC system. Specifically, it includes: Extract relevant parameters of the building ventilation system, lighting system and HVAC system from the BIM model, including information such as equipment type, specifications and layout location, and build an equipment parameter database; combine the output results of the AI ​​energy consumption prediction model with the dynamic energy-saving optimization algorithm, and determine the energy-saving control targets of each system according to the predicted energy consumption; such as the target air volume of the ventilation system, the target brightness of the lighting system and the target temperature setting value of the HVAC system; according to the information in the equipment parameter database and the energy-saving control target, use the optimization algorithm to calculate the optimal control parameters of each system, such as the fan speed of the ventilation system, the dimming ratio of the lighting system and the compressor frequency of the HVAC system; convert the calculated optimal control parameters into specific equipment control instructions, such as the frequency setting value of the fan inverter, the dimming ratio of the lamps and lanterns, and the compressor frequency of the HVAC system. The dimming signal of the light controller and the temperature setting value of the air-conditioning controller, etc.; continuously monitor the operating status and energy consumption data of each system, compare the actual energy consumption with the predicted energy consumption, and calculate the energy-saving effect; if there is a large deviation between the actual energy-saving effect and the expected target, the deviation data will be fed back to the dynamic energy-saving optimization algorithm, and the control strategy will be continuously optimized through machine learning to improve the energy-saving effect; then, the equipment of each system will be regularly diagnosed and fault detected, and the potential failures and performance degradation of the equipment will be identified through big data analysis technology, and maintenance plans will be formulated in advance to ensure the efficient operation of the system; the optimized control strategy and energy-saving effect data will be uploaded to the cloud platform, and decision support will be provided for building energy management through data mining and visualization technology, so as to continuously improve and optimize the entire BIM-AI energy consumption prediction and energy-saving optimization framework.

[0024] For example, equipment parameters of building ventilation, lighting and HVAC systems, such as fan power, lamp power, refrigeration capacity of refrigeration units, etc., are extracted from the BIM model to build an equipment parameter database. AI energy consumption prediction models such as BP neural network are used in combination with energy-saving optimization algorithms such as dynamic programming to predict the building energy consumption in the next 24 hours, and the ventilation target air volume is determined to be 5m³ / s, the lighting target illumination is 300lx, and the air conditioning target temperature is 26℃. According to the equipment parameters, the particle swarm optimization algorithm is used to calculate the optimal control parameters of fan speed of 900rpm, lamp dimming ratio of 80%, and compressor frequency of 50Hz. The control parameters are converted into equipment control instructions and sent to the field controller for execution through the BEMS system. Energy consumption data is continuously collected. If the actual energy consumption deviates from the predicted value by more than 10%, the control model is optimized through machine learning algorithms such as random forests. Big data analysis technology is used to diagnose equipment operation data for faults and perform preventive maintenance in advance. The optimized control strategy is uploaded to the cloud platform, and the BIM-AI energy consumption prediction and energy-saving optimization framework is continuously optimized through data mining algorithms.

[0025] S8, monitoring optimization: collect real-time monitoring data, including energy consumption data and environmental parameters, integrate these data with the original training data set, use online learning algorithms to update AI model parameters, continuously optimize prediction accuracy and energy-saving effects, and form a continuous improvement cycle system, including: Through the Internet of Things sensors, real-time collection of building energy consumption data and environmental parameters, including indoor temperature and humidity, lighting intensity, and equipment operating status, forms a real-time monitoring data stream, and integrates the real-time monitoring data with the building feature data extracted from the BIM model to construct a comprehensive data set containing energy consumption data, environmental parameters, and building features; preprocessing of the comprehensive data set, including data cleaning and normalization, ensures data quality and consistency, and prepares for subsequent model updates; based on the preprocessed comprehensive data set, the original AI energy consumption prediction model is incrementally trained and parameter updated using an online learning algorithm, so that the model can adapt to changes in the building's operating status; during the model update process, the prediction accuracy and generalization ability of the model are evaluated through cross-validation and other technologies, and the model structure is dynamically adjusted according to the evaluation results. The updated AI energy consumption prediction model is applied to the actual building energy consumption prediction, and the prediction results are compared with the actual energy consumption data to calculate the prediction error. If the prediction error exceeds the preset threshold, the automatic optimization mechanism of the model is triggered, and the model parameters are fine-tuned through optimization algorithms such as gradient descent to further improve the prediction accuracy. According to the optimized AI energy consumption prediction results, combined with the cost model and optimization algorithm, energy-saving strategies are dynamically generated, and implemented to the building's lighting, air conditioning, ventilation and other subsystems through the intelligent control system to achieve real-time energy-saving optimization. The changes in building energy consumption and the implementation effects of energy-saving measures are continuously monitored, and the feedback data is re-incorporated into the comprehensive data set to form a closed-loop system of data-prediction-optimization-feedback to achieve continuous improvement in building energy consumption management.

[0026] For example, by deploying a wireless sensor network in a building, each sensor node collects environmental parameters such as temperature, humidity, illumination, and operating status data of equipment such as air conditioning and lighting, and transmits them to the data center in real time through the Zigbee protocol. The building structure, materials, equipment and other information extracted from the BIM model are associated and mapped with the real-time data collected by the sensor to construct a multi-dimensional comprehensive data set. The data is cleaned, outliers are removed, and the minimum-maximum normalization method is used to normalize the numerical data and scale the data to the [0,1] interval. Based on the comprehensive data set, the support vector machine regression (SVR) algorithm is used to perform online incremental learning on the original energy consumption prediction model, and the model parameters are dynamically updated through the sequential minimum optimization (SMO) algorithm to continuously adapt to changes in the building's operating conditions. The model performance is evaluated using 5-fold cross validation, and the hyperparameters of the SVR kernel function type, penalty coefficient C, and kernel function parameter γ are optimized through grid search to obtain the optimal model with the lowest mean absolute percentage error (MAPE). The optimized SVR model is used for building energy consumption prediction. When the deviation rate between the actual energy consumption and the predicted value exceeds 5%, the model is automatically optimized. The stochastic gradient descent (SGD) algorithm is used, and the energy consumption deviation rate is used as the loss function to fine-tune the support vector, slack variable and weight coefficient of the SVR model until the energy consumption prediction accuracy reaches more than 95%. According to the optimized energy consumption prediction results, combined with cost factors such as electricity prices, the particle swarm optimization (PSO) algorithm is used to search for the optimal equipment operation combination and control parameters, generate dynamic energy-saving strategies, and issue control instructions through the BACnet protocol to achieve automatic execution of energy-saving measures such as air-conditioning temperature setting and lighting brightness adjustment. The building energy consumption, environmental parameters and equipment status changes are continuously recorded, key performance indicators (KPIs) such as energy-saving rate are calculated, and the energy-saving effect is presented through data visualization technology to form an optimization closed loop, which provides data support and decision-making basis for subsequent energy consumption prediction and strategy optimization, and realizes the continuous optimization management of building energy consumption.

[0027] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that a technician familiar with the technical field can think of within the technical scope disclosed in the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for predicting building energy consumption and dynamic energy-saving optimization based on BIM, characterized in that: The steps include: S1, data extraction: extract key characteristic parameters of building geometry, materials, and construction from the BIM model to obtain multi-dimensional data on building type, structural characteristics, material properties, and environmental factors; S2, data processing: cleaning the acquired multidimensional data, removing duplicate, erroneous or abnormal data points, and normalizing them by the minimum-maximum scaling method or the z-score normalization method; S3, determine the model: according to the characteristics of the preprocessed data and the complexity of the problem, select the neural network, support vector machine, random forest AI model, determine the input layer, hidden layer and output layer structure for the neural network model, select the kernel function for the support vector machine, and determine the number of decision trees for the random forest; S4, algorithm selection: the BIM model parameter data and historical energy consumption data after cleaning and preprocessing are used as training samples, the neural network model is trained using the back propagation algorithm, the support vector machine is trained using the sequence minimum optimization algorithm, and the random forest is trained using the bootstrap sampling method. At the same time, cross-validation and regularization techniques are used to prevent overfitting; S5, model validation: Use the reserved validation set to validate the trained model, evaluate the prediction accuracy and robustness of the model, and adjust the weights and biases of the neural network, the penalty parameters of the support vector machine, and the tree depth of the random forest based on the validation results; S6, Generate dynamic energy-saving strategies: Use the adjusted AI model for building energy consumption prediction, combine the cost model of material, labor, maintenance, and energy cost factors, and use genetic algorithms to generate dynamic energy-saving strategies; S7, instruction conversion: convert the generated dynamic energy-saving strategy into specific system adjustment instructions, apply them to actual buildings, and achieve energy-saving control by adjusting the air volume of the building ventilation system, the brightness of the lighting system, and the temperature setting value of the HVAC system; S8, monitoring optimization: collect real-time monitoring data, including energy consumption data and environmental parameters, integrate these data with the original training data set, use online learning algorithms to update AI model parameters, continuously optimize prediction accuracy and energy-saving effects, and form a continuous improvement cycle system.

2. A method for predicting building energy consumption and dynamically optimizing energy saving based on BIM as claimed in claim 1, characterized in that: S2 includes the following steps: Obtain multi-dimensional data on building geometry, materials and construction in the BIM model, extract key parameters through Revit API or Dynamo tools, and obtain the original data set; pre-process the original data set, use data cleaning technology to remove duplicate and erroneous or abnormal data points, and obtain the cleaned data set; According to the distribution characteristics of the cleaned data set, the minimum-maximum scaling method is used to scale the data to the [0, 1] interval, or the z-score standardization method is used to convert the data into a distribution with a mean of 0 and a variance of 1 to obtain a standardized data set; Divide the standardized data set into training subset, validation subset and test subset according to the preset ratio; According to the characteristics of building energy consumption prediction problems, a hybrid deep learning model consisting of convolutional neural networks, recurrent neural networks, and long short-term memory networks is constructed. The hybrid deep learning model can capture long-term dependencies in time series data. The training subset is input into the hybrid deep learning model, the predicted value is calculated through forward propagation, and the loss function value is calculated with the actual value; the gradient of the loss function to the model parameters is calculated using the back propagation algorithm, and the model parameters are updated using the optimization algorithm to minimize the loss function, thus obtaining the trained building energy consumption prediction model; Input the verification subset into the building energy consumption prediction model, calculate the accuracy index of the energy consumption prediction, and when the accuracy index reaches a preset threshold, determine the building energy consumption prediction model as the final model; The test subset is input into the final model to obtain the building energy consumption prediction results, and the prediction results are output to the building energy consumption management system.

3. A method for predicting building energy consumption and dynamically optimizing energy saving based on BIM as claimed in claim 2, characterized in that: The training subset, validation subset, and test subset are divided into 70%, 15%, and 15% ratios; when the model performance on the validation subset meets the preset standard, the model is finally tested using the test subset data to obtain the prediction accuracy of the model on real unknown data; The test results are used to determine whether the model can be deployed for application. If the model performance is poor, it is necessary to return to the data preprocessing or model design stage for optimization.

4. The method for predicting building energy consumption and dynamically optimizing energy saving based on BIM according to claim 1, characterized in that: S3 includes the following steps: According to the characteristics of the pre-processed building energy consumption data and the complexity of the problem, the decision tree algorithm is used to analyze the data to obtain the key characteristics and distribution of the data; If the data has nonlinear characteristics and high dimensions, the neural network model is selected for energy consumption prediction. When the neural network model is selected, the number of input layer nodes and output layer nodes of the neural network is determined by analyzing the dimensions and characteristics of the data, and the number of hidden layers and the number of nodes in each layer are set based on experience; If the data dimension is low and has obvious classification boundaries, the support vector machine model is selected. For the support vector machine model, different kernel functions are evaluated by cross-validation method to select the kernel function type and corresponding hyperparameters with the best prediction performance; If there are a large number of discrete variables in the data, the random forest model is used. When using the random forest model, the number of decision trees, the maximum depth, and the minimum number of samples for splitting nodes are optimized through grid search and other methods to obtain the optimal model structure. According to the selected model type, build the training data set and validation data set of the energy consumption prediction model, and set the appropriate training batch size and number of iterations; During the model training process, the loss function value changes of the training set and the validation set are monitored to determine whether the model is overfitting or underfitting, and the model structure or hyperparameters are adjusted according to the situation. After the model training is completed, the prediction performance of the model is evaluated using the test data set to obtain indicators such as the accuracy and mean square error of the energy consumption prediction to verify the generalization ability of the model. If the model prediction performance meets the requirements, the trained model will be deployed to the building energy consumption management system. Through the real-time collected building operation data, the future energy consumption trend can be dynamically predicted to provide support for energy-saving optimization decisions.

5. The method for predicting building energy consumption and dynamically optimizing energy saving based on BIM according to claim 1, characterized in that: S6 includes the following steps: Using artificial neural network algorithms, deep learning training is performed on the building energy consumption database to obtain an AI model that can accurately predict building energy consumption; The model comprehensively considers multiple influencing factors such as window-to-wall ratio, building orientation and thermal performance of the external envelope structure. It obtains energy consumption changes by real-time monitoring of building operation data and feeds them back to the AI ​​energy consumption prediction model. The AI ​​energy consumption prediction model is adjusted and optimized using the feedback data to improve the prediction accuracy and adaptability of the model. The optimized AI model is applied to building energy consumption prediction to obtain energy consumption prediction results. According to the energy consumption prediction results, combined with the cost model considering factors such as materials, labor, maintenance and energy, an energy-saving optimization objective function is constructed; a genetic algorithm is used to solve the energy-saving optimization objective function, and the optimal combination of energy-saving strategies is searched through operations such as selection, crossover and mutation; the optimal solution is selected from the solution set of the genetic algorithm and used as the dynamic energy-saving optimization strategy for the building, which includes adjusting the operating parameters of systems such as ventilation, lighting and air conditioning; the dynamic energy-saving optimization strategy is applied to the building energy management system, and real-time adjustment of energy-saving measures is achieved through automated control, and the energy-saving effect is continuously monitored to form a closed-loop optimization.

6. The method for predicting building energy consumption and dynamically optimizing energy saving based on BIM according to claim 1, characterized in that: S7 includes the following steps: Extract relevant parameters of the building ventilation system, lighting system and HVAC system from the BIM model, including equipment type, specifications and layout location, and build an equipment parameter database; Combine the output results of the AI ​​energy consumption prediction model with the dynamic energy-saving optimization algorithm, and determine the energy-saving control targets of each system based on the predicted energy consumption, such as the target air volume of the ventilation system, the target brightness of the lighting system, and the target temperature setting value of the HVAC system; Based on the information in the equipment parameter database and in combination with the energy-saving control target, an optimization algorithm is used to calculate the optimal control parameters of each system, such as the fan speed of the ventilation system, the dimming ratio of the lamps of the lighting system, and the compressor frequency of the HVAC system; the calculated optimal control parameters are converted into specific equipment control instructions, such as the frequency setting value of the fan inverter, the dimming signal of the lamp dimming controller, and the temperature setting value of the air conditioning controller; Continuously monitor the operating status and energy consumption data of each system, compare the actual energy consumption with the predicted energy consumption, and calculate the energy-saving effect; if there is a large deviation between the actual energy-saving effect and the expected target, the deviation data will be fed back to the dynamic energy-saving optimization algorithm, and the control strategy will be continuously optimized through machine learning to improve the energy-saving effect.

7. A method for predicting building energy consumption and dynamically optimizing energy saving based on BIM as claimed in claim 6, characterized in that: Regularly perform performance diagnosis and fault detection on the equipment of each system, identify potential failures and performance degradation of equipment through big data analysis technology, formulate maintenance plans in advance, and ensure efficient operation of the system; upload the optimized control strategy and energy-saving effect data to the cloud platform, and provide decision support for building energy management through data mining and visualization technology, and continuously improve and optimize the entire BIM-AI energy consumption prediction and energy-saving optimization framework.

8. The method for predicting building energy consumption and dynamically optimizing energy saving based on BIM according to claim 1, characterized in that: S8 includes the following steps: The IoT sensors collect building energy consumption data and environmental parameters in real time, including indoor temperature and humidity, lighting intensity, and equipment operation status, to form a real-time monitoring data stream. The real-time monitoring data is integrated with the building feature data extracted from the BIM model to construct a comprehensive data set that includes energy consumption data, environmental parameters, and building features. Preprocess the comprehensive data set, including data cleaning and normalization, to ensure data quality and consistency, in preparation for subsequent model updates; based on the preprocessed comprehensive data set, use online learning algorithms to incrementally train and update parameters of the original AI energy consumption prediction model, so that the model can adapt to changes in building operation status; During the model update process, the prediction accuracy and generalization ability of the model are evaluated through cross-validation and other technologies, and the model structure and hyperparameters are dynamically adjusted according to the evaluation results to obtain the best prediction performance; the updated AI energy consumption prediction model is applied to the actual building energy consumption prediction, and the prediction results are compared with the actual energy consumption data to calculate the prediction error; if the prediction error exceeds the preset threshold, the model automatic optimization mechanism is triggered, and the model parameters are fine-tuned through optimization algorithms such as gradient descent to further improve the prediction accuracy; Based on the optimized AI energy consumption prediction results, combined with the cost model and optimization algorithm, energy-saving strategies are dynamically generated and implemented in the building's lighting, air conditioning, ventilation and other subsystems through intelligent control systems to achieve real-time energy-saving optimization; Continuously monitor changes in building energy consumption and the effectiveness of energy-saving measures, and reincorporate feedback data into the comprehensive data set to form a closed-loop system of data-prediction-optimization-feedback to achieve continuous improvement in building energy consumption management.

Images