Window-to-wall ratio strategy prediction method

Abstract

The application is suitable for the field of building technology, and discloses a building window-wall ratio strategy prediction model construction method and a window-wall ratio strategy prediction method. The prediction method comprises providing a building window-wall ratio strategy prediction model, and obtaining meteorological characteristic data at a to-be-predicted time. The obtained meteorological characteristic data is input into the building window-wall ratio strategy prediction model to obtain an optimal window-wall ratio strategy. The construction method of the building window-wall ratio strategy prediction model comprises data preprocessing on collected building window-wall ratio strategy data and meteorological characteristic data. The processed data is divided into training data and to-be-predicted data. A random forest classification model is established, and the optimal hyperparameter combination of the random forest classification model is determined based on the training data by using a grid search method and a K-fold cross-validation method. The building window-wall ratio strategy prediction model is constructed according to the optimal hyperparameter combination. The prediction method has high accuracy and can effectively help the building envelope structure to realize intelligent window-wall ratio automatic control.

Description

Technical Field

[0001] This invention relates to the field of building technology, and in particular to a method for predicting the window-to-wall ratio. Background Technology

[0002] The window-to-wall ratio is a crucial factor affecting a building's energy consumption. A well-designed window-to-wall ratio can minimize energy consumption while still meeting ventilation and lighting requirements. Currently, traditional building design employs a static design method, determining the optimal window-to-wall ratio based on the annual energy consumption demand.

[0003] However, the requirements for window-to-wall ratio vary significantly at different stages of building construction. For example, improving the indoor thermal environment in winter requires increasing the window-to-wall ratio to increase solar radiation heat gain; reducing the air conditioning load in summer requires reducing the window-to-wall ratio to decrease heat gain while ensuring sufficient lighting. With the rapid development of energy-saving technologies and the increasing demands for energy conservation, static design methods are increasingly unable to meet the needs of increasingly intelligent buildings. Dynamic, intelligent, and adaptable variable building envelopes that can meet the thermal comfort needs of different people are worthy of analysis and research.

[0004] For adjustments to variable building envelopes, accurately predicting the optimal window-to-wall ratio strategy is crucial. However, current methods for predicting the window-to-wall ratio have low accuracy, and further improvements are needed. Summary of the Invention

[0005] The first objective of this invention is to provide a building window-to-wall ratio prediction model, which aims to solve the technical problem of low prediction accuracy in existing methods for predicting window-to-wall ratio.

[0006] To achieve the above objectives, the solution provided by the present invention is as follows:

[0007] The construction method of the building window-to-wall ratio strategy prediction model includes:

[0008] Data preprocessing was performed on the collected building window-to-wall ratio strategy data and corresponding historical meteorological characteristic data;

[0009] The processed data is divided into training data and data to be predicted.

[0010] A random forest classification model is established, and the optimal combination of hyperparameters for the random forest classification model is determined based on the training data using grid search and K-fold cross-validation.

[0011] A building window-to-wall ratio prediction model is constructed based on the optimal hyperparameter combination.

[0012] Preferably, the method further includes inputting the training data to retrain the building window-to-wall ratio strategy prediction model to obtain an optimized building window-to-wall ratio strategy prediction model.

[0013] Preferably, the method further includes inputting the data to be predicted into the building window-to-wall ratio strategy prediction model, and using the accuracy index and confusion matrix to measure the building window-to-wall ratio strategy prediction model. If the accuracy is greater than 80% and the confusion matrix is ​​greater than 70%, it indicates that the building window-to-wall ratio strategy prediction model has high accuracy.

[0014] Preferably, the hyperparameter combination includes maximum depth, minimum number of leaves, and number of trees. The method for determining the optimal hyperparameter combination for the random forest classification model includes:

[0015] Given the maximum depth value, the minimum number of leaves value, and the number of trees value, a grid search method is used to cross-combine all nodes in all hyperparameters to divide them into different hyperparameter combinations;

[0016] Each hyperparameter combination was applied to a random forest classification model. The meteorological feature data of the training data was used as the input vector, and the optimal window-to-wall ratio strategy corresponding to the training data was used as the output vector to train random forest classification models with different hyperparameter combinations. The performance of the random forest classification model was verified by K-fold cross-validation.

[0017] The trained model is measured using the Accuracy metric, and the hyperparameter combination with the highest Accuracy is selected as the optimal hyperparameter combination.

[0018] Preferably, when training a random forest classification model with different hyperparameter combinations, the training data is randomly divided into K subsets, one subset is taken as the validation set, the remaining subset is taken as the training set, and the model is fitted K times.

[0019] Preferably, the maximum depth includes 5, 15, 20, 25, 35, and 45; the minimum number of leaves includes 1, 2, 4, 6, 8, and 10; and the number of trees includes 18, 27, 31, 45, and 63.

[0020] Preferably, the performance of the random forest classification model is verified using 10-fold cross-validation.

[0021] Preferably, the meteorological characteristic data includes solar irradiance, dry-bulb temperature, and relative humidity.

[0022] Preferably, the data preprocessing includes cleaning up missing meteorological feature data or category labels representing window-to-wall ratio strategies, and then removing outliers from the meteorological feature data.

[0023] A second objective of this invention is to provide a window-to-wall ratio prediction strategy method, comprising:

[0024] A building window-to-wall ratio strategy prediction model is provided, which is constructed using the method described above.

[0025] Obtain meteorological characteristic data for the time to be predicted;

[0026] The acquired meteorological characteristic data is input into the building window-to-wall ratio strategy prediction model to obtain the optimal window-to-wall ratio strategy.

[0027] The building window-to-wall ratio prediction model provided by this invention is constructed based on random forest classification and grid search methods, and the prediction results have high accuracy, specifically reflected in:

[0028] First, the building window-to-wall ratio prediction model based on random forest classification can output very high-dimensional data without dimensionality reduction or feature selection. It can also determine the importance of features and the interactions between different features, making it suitable for situations where multiple factors influence the prediction. Furthermore, it is less prone to overfitting, has a fast training speed, and is more suitable for the current multi-class classification problem of building window-to-wall ratio. In addition, the random forest classification model can balance the errors of imbalanced datasets; if an important feature is lost, accuracy can still be maintained.

[0029] Second, the building window-to-wall ratio prediction model based on grid search can automatically adjust model parameters according to different building window-to-wall ratio strategies and meteorological characteristic data, thereby improving the classification ability of the building window-to-wall ratio prediction model based on random forest.

[0030] The window-to-wall ratio prediction method provided by this invention predicts the window-to-wall ratio of a building by constructing a prediction model based on random forest classification and grid search. The prediction results have high accuracy and can effectively help the building envelope achieve intelligent automated control of the window-to-wall ratio. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0032] Figure 1 This is a flowchart of the window-to-wall ratio prediction method provided in an embodiment of the present invention;

[0033] Figure 2 This is a flowchart of the method for constructing a building window-to-wall ratio prediction model provided in an embodiment of the present invention;

[0034] Figure 3 This is an evaluation result of the accuracy of the prediction model for the optimized building window-to-wall ratio strategy using a confusion matrix, according to an embodiment of the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0036] like Figures 1 to 2 As shown, this is an embodiment of the window-to-wall ratio strategy prediction method of the present invention. It obtains the optimal window-to-wall ratio strategy through prediction, which helps to realize intelligent automatic control of the window-to-wall ratio of building envelope. A reasonable window-to-wall ratio can help reduce the energy consumption of buildings.

[0037] Please see Figures 1-2 The window-to-wall ratio prediction method of this invention includes:

[0038] Step S100: Collect building window-to-wall ratio strategy data and corresponding historical meteorological characteristic data, and construct a building window-to-wall ratio strategy prediction model based on random forest classification and grid search method.

[0039] Step S200: Meteorological characteristic data for the time to be predicted; this meteorological characteristic data can be obtained from meteorological forecast data or acquired by a meteorological information acquisition module.

[0040] Step S300: Input the acquired meteorological characteristic data into the building window-to-wall ratio strategy prediction model to obtain the optimal window-to-wall ratio strategy.

[0041] Understandably, meteorological characteristic data includes solar irradiance, dry-bulb humidity, and relative humidity. Among the numerous meteorological characteristic data, solar irradiance, dry-bulb humidity, and relative humidity have the greatest impact on the window-to-wall ratio of buildings. Therefore, the window-to-wall ratio prediction method of this invention constructs a building window-to-wall ratio prediction model by collecting solar irradiance, dry-bulb humidity, and relative humidity, and predicts the building window-to-wall ratio by collecting solar irradiance, dry-bulb humidity, and relative humidity at the time to be predicted.

[0042] Understandably, in order to improve the accuracy of the building window-to-wall ratio strategy prediction model, different building window-to-wall ratio strategy prediction models can be constructed for different buildings in different regions. Therefore, when it is necessary to construct a building window-to-wall ratio strategy prediction model for a certain building in a certain region (e.g., Jinan), the building window-to-wall ratio strategy data of that building and the meteorological characteristic data of that region can be collected.

[0043] Understandably, the window-to-wall ratio strategy includes changes in the window-to-wall ratio for south, north, east, and west directions. The strategy is represented by the window-to-wall ratio state and direction. The window-to-wall ratio state includes the most basic state of no windows (full wall), half windows, and full windows. Of course, the window-to-wall ratio state can also include other states, and can be specified to a specific window-to-wall ratio value, which helps to achieve intelligent and automated control of the window-to-wall ratio of the building envelope.

[0044] Without considering the state of north-facing windows, we consider the window-to-wall ratio changes for south, east, and west-facing windows, while keeping the window-to-wall ratios of the other two directions fixed and in a half-open state. Therefore, there are 27 possible window-to-wall ratio strategies. These strategies are represented by the window-to-wall ratio state and direction, where 00 represents a completely closed state, 40 represents a half-open state, and 80 represents a fully open state; east is represented by E, west by W, and south by S. Thus, window-to-wall ratio strategies can be represented as E_40_W_40_S_00, E_40_W_40_S_40, E_40_W_40_S_80, etc.

[0045] The window-to-wall ratio prediction method of this invention predicts the building window-to-wall ratio using a prediction model built based on random forest classification and grid search. The prediction results have high accuracy and can effectively help the building envelope achieve intelligent automated control of the window-to-wall ratio, specifically in the following ways:

[0046] First, the building window-to-wall ratio prediction model based on random forest classification can output very high-dimensional data without dimensionality reduction or feature selection. It can also determine the importance of features and the interactions between different features, making it suitable for situations where multiple factors influence the prediction. Furthermore, it is less prone to overfitting, has a fast training speed, and is more suitable for the current multi-class classification problem of building window-to-wall ratio. In addition, the random forest classification model can balance the errors of imbalanced datasets; if an important feature is lost, accuracy can still be maintained.

[0047] Second, the building window-to-wall ratio strategy prediction model, based on the grid search method, can automatically adjust according to different building window-to-wall ratio strategies and meteorological characteristic data.

[0048] Preferably, the method for constructing a building window-to-wall ratio prediction model includes:

[0049] Step S11: Preprocess the collected building window-to-wall ratio strategy data and meteorological feature data; the specific implementation method is: after cleaning the missing meteorological feature data or the category labels representing the window-to-wall ratio strategy, remove the outliers in the meteorological feature data.

[0050] Step S12: Divide the processed data into training data and data to be predicted.

[0051] Step S13: Establish a random forest classification (RFC) model and determine the optimal combination of hyperparameters for the random forest classification model based on the training data using grid search and K-fold cross-validation.

[0052] Specifically, the hyperparameters of the random forest classification model include maximum depth (max_depth), minimum number of leaves (min_samples_leaf), and number of trees (e_estimators).

[0053] Step S131: Given multiple values ​​for the maximum depth, multiple values ​​for the minimum number of leaves, and multiple values ​​for the number of trees, use a grid search method to cross-combine all nodes in all hyperparameters to divide them into different parameter combinations.

[0054] Step S132: Apply each hyperparameter combination to the random forest classification model. Use the meteorological feature data of the training data as the input vector and the optimal window-to-wall ratio strategy corresponding to the training data as the output vector to train the random forest classification model with different hyperparameter combinations. Use K-fold cross-validation to verify the performance of the random forest classification model with different hyperparameter combinations.

[0055] Step S133: Measure the trained model using the Accuracy metric, and use the parameter combination with the highest Accuracy as the hyperparameter combination for the random forest classification model.

[0056] When training a random forest classification model with different parameter combinations, the training data is randomly divided into K subsets. One subset is used as the validation set, and the remaining subsets are used as the training set. The model is then fitted K times. That is, there are K subsets of the training set, denoted as {T1, T2, ..., T...}. K The validation set also has K parts, represented as {S1, S2, ..., S}. K When training a random forest classification model with different hyperparameter combinations, the training set T1 is input first. The input vector consists of data on three meteorological characteristics—solar irradiance, dry-bulb temperature, and relative humidity—during a certain time period. The output vector is the optimal window-to-wall ratio strategy corresponding to that meteorological characteristic. By learning the influence of these three meteorological characteristics on the building window-to-wall ratio, a random forest classification model is constructed. Then, the validation set S1 corresponding to the training set T1 is input into the trained model, and the predicted value of the optimal window-to-wall ratio strategy is output. The accuracy is calculated by combining the actual values ​​of the optimal window-to-wall ratio strategy in the validation set S1. Each parameter combination is input into the training sets T1, T2, ..., T1 one by one. K And respectively through the corresponding validation sets S1, S2, ..., S KThe accuracy is obtained, and the average of the accuracy after K fittings is taken as the accuracy of the corresponding parameter combination. Finally, the parameter combination with the highest accuracy is taken as the hyperparameter combination of the random forest classification model. That is, the parameters applied by the model that contain the most accurate optimal window-to-wall ratio strategies in the prediction results are the best parameters.

[0057] Specifically, the embodiments of the present invention use 10-fold cross-validation to verify the performance of the random machine forest classification model, which can effectively avoid overlearning and underlearning states. Therefore, the final prediction results of the optimal window-to-wall ratio strategy are also quite convincing.

[0058] Step S14: To improve model performance, it is necessary to validate the performance of the current optimal parameter model using more datasets. Specifically, a new random forest classification model is constructed based on the optimal hyperparameter combination, and the model is retrained using the input training data to obtain the building window-to-wall ratio prediction model.

[0059] Step S15: Input the data to be predicted into the building window-to-wall ratio strategy prediction model, output the target prediction strategy, and use the accuracy index and confusion matrix to measure the building window-to-wall ratio strategy prediction model to evaluate its generalization ability. A larger accuracy and a larger confusion matrix indicate more accurate predictions, meaning the optimized building window-to-wall ratio strategy prediction model predicts more results, and vice versa.

[0060] Understandably, step S14 is not necessary and can be omitted, allowing the optimal hyperparameter combination to be used directly to construct a building window-to-wall ratio prediction model.

[0061] The embodiments of the present invention construct a building window-to-wall ratio strategy prediction model based on random forest classification and grid search method. The prediction results have high accuracy and can effectively help the building envelope structure achieve intelligent automated control of the window-to-wall ratio.

[0062] This invention takes a building in Jinan as an example. It collects the window-to-wall ratio strategy of the building and historical meteorological data of Jinan. It then constructs a prediction model of the building's window-to-wall ratio strategy based on random forest classification and grid search. Finally, it evaluates the optimized prediction model of the building's window-to-wall ratio strategy.

[0063] This embodiment of the invention only considers the window-to-wall ratio variation on the south side, while fixing the window-to-wall ratio on the east and west sides and assuming a half-open window state. Therefore, the target strategy is one of the following three strategy categories, with labels 1, 2, and 3 as the output targets, respectively. Label 1 is represented as E_40_W_40_S_00, label 2 as E_40_W_40_S_40, and label 3 as E_40_W_40_S_80.

[0064] The hyperparameters given when training the random forest classification model are shown in Table 1.

[0065] Table 1 provides the hyperparameters.

[0066] max_depth [5,15,20,25,35,45] min_samples_leaf [1,2,4,6,8,10] e_estimators [18,27,31,45,63]

[0067] The optimal hyperparameter combination for the random forest classification model found using the grid search method is: 'max_depth': 25, 'min_samples_leaf': 6, 'e_estimators': 27. Then, a new random forest model is constructed based on 'max_depth': 25, 'min_samples_leaf': 6, 'e_estimators': 27, and the training data is input into the new random forest model for retraining, thereby obtaining the building window-to-wall ratio policy prediction model.

[0068] When evaluating the building window-to-wall ratio strategy prediction model, the data to be predicted is input into the model, which outputs the predicted optimal window-to-wall ratio strategy. Finally, the actual historical best window-to-wall ratio strategy and the predicted optimal strategy are compared. The accuracy and confusion matrix are used to evaluate the accuracy of the optimized model. The evaluation results are shown in Table 2. The accuracy is 86.11%, meaning that the model optimized using the grid search method has a greater than 80% probability of correctly predicting the optimal window-to-wall ratio strategy. Simultaneously, the confusion matrix shows that the region in strategy category 3 represents both the actual strategy E_40_W_40_S_80 and the predicted strategy E_40_W_40_S_80. This region accounts for 74%, indicating that the actual strategy category of the test data is mostly 3, and the probability of accurately predicting strategy category 3 is 74%. This demonstrates that the building window-to-wall ratio strategy prediction model has high accuracy and can accurately predict the optimal window-to-wall ratio strategy.

[0069] Table 2 Evaluation results of the building window-to-wall ratio prediction model

[0070] Optimal hyperparameter combination 'max_depth':25,'min_samples_leaf':6,'e_estimators':27 Accuracy 86.11％ Confusion Matrix 74％

[0071] Understandably, the window-to-wall ratio prediction model of this invention can also be applied to other window-to-wall ratio prediction methods.

[0072] Embodiments of the present invention can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or contained in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier.

[0073] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention's specification and drawings under the inventive concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A window-to-wall ratio strategy prediction method, characterized by, include: A building window-to-wall ratio strategy prediction model is provided. The method for constructing the building window-to-wall ratio strategy prediction model includes: preprocessing the collected building window-to-wall ratio strategy data and corresponding historical meteorological characteristic data; dividing the processed data into training data and data to be predicted; establishing a random forest classification model; determining the optimal hyperparameter combination of the random forest classification model based on the training data using grid search and K-fold cross-validation; and constructing the building window-to-wall ratio strategy prediction model according to the optimal hyperparameter combination. Acquire meteorological characteristic data for the time to be predicted, including solar irradiance, dry-bulb temperature, and relative humidity; The acquired meteorological characteristic data is input into the building window-to-wall ratio strategy prediction model to obtain the optimal window-to-wall ratio strategy, which helps to achieve intelligent and automated control of the window-to-wall ratio of the building envelope. The hyperparameter combination includes maximum depth, minimum number of leaves, and number of trees. The method for determining the optimal hyperparameter combination for the random forest classification model includes: given the maximum depth value, the minimum number of leaves value, and the number of trees value, a grid search method is used to cross-combine all nodes in all hyperparameters to divide them into different hyperparameter combinations; each hyperparameter combination is applied to the random forest classification model, using the meteorological feature data of the training data as the input vector and the optimal window-to-wall ratio strategy corresponding to the training data as the output vector to train random forest classification models with different hyperparameter combinations, and K-fold cross-validation is used to verify the performance of the random forest classification model; the accuracy metric is used to measure the trained model, and the hyperparameter combination with the highest accuracy is taken as the optimal hyperparameter combination; the window-to-wall ratio strategy includes changes in the window-to-wall ratio in the south, north, east, and west directions, and the strategy is represented by the window-to-wall ratio state and direction.

2. The window-to-wall ratio strategy prediction method of claim 1, wherein, The method for constructing the building window-to-wall ratio strategy prediction model further includes inputting the training data to retrain the building window-to-wall ratio strategy prediction model to obtain an optimized building window-to-wall ratio strategy prediction model.

3. The window-to-wall ratio strategy prediction method of claim 1, wherein, The method for constructing the building window-to-wall ratio strategy prediction model further includes inputting the data to be predicted into the building window-to-wall ratio strategy prediction model, and using the accuracy index and confusion matrix to measure the building window-to-wall ratio strategy prediction model. If the accuracy is greater than 80% and the confusion matrix is ​​greater than 70%, it indicates that the building window-to-wall ratio strategy prediction model has high accuracy.

4. The window-to-wall ratio strategy prediction method of claim 1, wherein, When training a random forest classification model with different hyperparameter combinations, the training data is randomly divided into K subsets. One subset is taken as the validation set, and the remaining subset is taken as the training set. The model is then fitted K times.

5. The window-to-wall ratio strategy prediction method of claim 1, wherein, The maximum depth values ​​include 5, 15, 20, 25, 35, and 45; the minimum number of leaves values ​​include 1, 2, 4, 6, 8, and 10; and the number of trees values ​​include 18, 27, 31, 45, and 63.

6. The window-to-wall ratio strategy prediction method of claim 1, wherein, The performance of the random forest classification model was verified using 10-fold cross-validation.

7. The window-to-wall ratio strategy prediction method of claim 1, wherein, The data preprocessing includes removing the abnormal values in the meteorological feature data after cleaning the missing meteorological feature data or the category label representing the window-wall ratio strategy.

Images