A method and system for predicting thermal comfort of airflow environment in train passenger compartment

Abstract

The present invention belongs to the technical field of air flow environment monitoring in train passenger compartments, and discloses a method and system for predicting thermal comfort of air flow environment in train passenger compartments. The method combines numerical calculation methods with CNN-LSTM models, which can significantly reduce the time and cost required for prediction results compared with methods based on traditional numerical simulation and experiments, and has good prediction accuracy. Compared with other machine learning models, the method proposed in the present invention can significantly improve prediction accuracy and reduce training time, which has significant advantages.

Description

Technical Field

[0001] The present invention belongs to the technical field of air flow environment monitoring in train passenger compartments, and in particular relates to a method and system for predicting thermal comfort of air flow environment in train passenger compartments. Background Art

[0002] With the continuous development of society and the improvement of people's economic level, more and more people choose to travel by train. Therefore, the thermal comfort caused by the airflow environment in the train passenger compartment has always been the focus of passengers. At present, scholars at home and abroad generally use computational fluid dynamics (CFD) numerical simulation and experimental research methods to evaluate and optimize the airflow environment and thermal comfort in the train. However, the common disadvantage of CFD and experimental technology is that the research time and manpower cost are large, especially when there are many CFD working conditions to be calculated, which will consume higher computing resources. Therefore, it is difficult to quickly and accurately predict the thermal comfort in the train passenger compartment, and the current train ventilation design time and cost are high. Summary of the invention

[0003] The present invention provides a method and system for predicting thermal comfort of airflow environment in a train passenger compartment, which are used to solve the current problem that it is difficult to quickly and accurately predict thermal comfort in a train passenger compartment, and the time and cost of ventilation design of the train are high.

[0004] In order to solve the above technical problems, the technical solution proposed by the present invention is:

[0005] In a first aspect, the present application provides a method for predicting thermal comfort of airflow environment in a train passenger compartment, comprising:

[0006] S1: Orthogonal experiments are used to design sample working conditions under multiple design variables based on train ventilation parameter variables and external environment parameter variables;

[0007] S2: using a numerical method to simulate the sample working condition, and monitoring the flow field information of the surface surrounding the human body in the train passenger compartment simulation model;

[0008] S3: Based on the flow field information monitored on the package surface, a thermal comfort index of the sample working condition is calculated by a user-defined UDF function, wherein the thermal comfort index includes the values ​​of the predicted average voting PMV and the wind dissatisfaction rate DR;

[0009] S4: constructing a CNN-LSTM model based on the sample working condition, the predicted average voting PMV and the blowing sensation dissatisfaction rate DR;

[0010] S5: Based on the CNN-LSTM model, the thermal comfort of passengers in the car under different ventilation parameters and external environmental parameters is predicted.

[0011] In a second aspect, the present application also provides a thermal comfort prediction system for the air flow environment in a train passenger compartment, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in the first aspect when executing the computer program.

[0012] The present invention has the following beneficial effects:

[0013] The method for predicting thermal comfort of airflow environment in train passenger compartments provided by this application can significantly reduce the time and cost required for prediction results compared with methods based on traditional numerical simulation and experiments by combining numerical calculation methods with CNN-LSTM models, and has good prediction accuracy. Compared with other machine learning models, the method proposed by the present invention can significantly improve prediction accuracy and reduce training time, which has significant advantages. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The drawings constituting a part of this application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0015] Figure 1 A flow chart of a method for predicting thermal comfort of airflow environment in a train passenger compartment provided in this application;

[0016] Figure 2 This is a model structure framework diagram in the embodiment of the present application;

[0017] Figure 3 The train passenger compartment model in the embodiment of the present application;

[0018] Figure 4 Schematic diagram of PMV prediction errors of four models in the embodiments of the present application;

[0019] Figure 5 Schematic diagram of DR prediction error of four models in the embodiments of the present application;

[0020] Figure 6 This is a schematic diagram of the time required for a single training of the four models in the embodiments of the present application.

[0021] Reference numerals:

[0022] 1. Human body model; 2. Surface surrounding the human body. DETAILED DESCRIPTION

[0023] In order to facilitate the understanding of the present invention, the present invention will be described more comprehensively and meticulously below in conjunction with the accompanying drawings and preferred embodiments of the present invention, but the protection scope of the present invention is not limited to the following specific embodiments.

[0024] Unless otherwise defined, all professional terms used below have the same meanings as those generally understood by those skilled in the art. The professional terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the scope of protection of the present invention.

[0025] Unless otherwise specified, various raw materials, reagents, instruments and equipment used in the present invention can be purchased from the market or prepared by existing methods.

[0026] It should be understood that this application proposes an efficient method for predicting the thermal comfort of the airflow environment in the train passenger compartment by combining the numerical calculation method with the CNN-LSTM model. The feasibility and accuracy of this application in predicting the thermal comfort of the airflow environment in the train can be verified.

[0027] Embodiment 1:

[0028] See also Figure 1 In this embodiment, a method for predicting thermal comfort of airflow environment in a train passenger compartment is disclosed, comprising:

[0029] S1: Orthogonal experiments are used to design sample working conditions under multiple design variables based on train ventilation parameter variables and external environment parameter variables;

[0030] S2: using a numerical method to simulate the sample working condition, and monitoring the flow field information of the surface surrounding the human body in the train passenger compartment simulation model;

[0031] S3: Based on the flow field information monitored on the package surface, a thermal comfort index of the sample working condition is calculated by a user-defined UDF function;

[0032] S4: constructing a CNN-LSTM model based on the sample working condition and the thermal comfort index;

[0033] S5: Based on the CNN-LSTM model, the thermal comfort of passengers in the car under different ventilation parameters and external environmental parameters is predicted.

[0034] The ventilation parameters may include upper air supply ratio, air supply angle, air supply temperature, and air supply humidity, and the external environment parameters may include outdoor air temperature. These are examples only and are not intended to be limiting.

[0035] The thermal comfort index includes the predicted average voting PMV and the wind dissatisfaction rate DR. The train passenger compartment model and the thermal comfort evaluation index (PMV and DR values) are only embodiments and are not intended to limit the present invention. For those skilled in the art, the present invention may have various changes and variations, such as in buildings, aircraft cabins, air flow velocity and temperature data, etc. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

[0036] Below, the steps of the above-mentioned method for predicting thermal comfort of airflow environment in train passenger compartment are described in detail with a complete example:

[0037] 1. Select the number and range of train ventilation parameter variables and external environment parameter variables, and use orthogonal experimental design to determine the number of sample conditions under multiple design variables. Use numerical methods to simulate the sample conditions and monitor the flow field information of the envelope around the human body in the train passenger compartment simulation model, which is about 10 cm away from the human body surface.

[0038] 2. Based on the flow field information monitored on the package surface, the thermal comfort index PMV (predicted mean vote) and DR (wind dissatisfaction rate) values ​​of the sample working condition are calculated through the user-defined UDF function. The specific PMV and DR calculation formulas can be referred to GB / T 18049-2017. Several additional calculation conditions are calculated as test sets to verify the accuracy and reliability of the model.

[0039] 3. In this embodiment, if Figure 2 As shown in the figure, the CNN-LSTM model specifically includes an input layer, a convolution layer, a flat layer, an LSTM layer, a fully connected layer, and an output layer. The input layer is the thermal comfort calculation result of all grid points on the envelope surface obtained by the sample working condition calculation, and outputs it to the convolution layer. The convolution layer needs to pre-set the number of convolution layers and the range of convolution kernels, and convolve the thermal comfort results with the convolution kernel and the input features to output them to the flat layer. The flat layer converts the results of the multi-dimensional grid points on the envelope surface processed by the convolution layer into a one-dimensional vector and outputs it to the LSTM layer. The LSTM layer needs to pre-set the range of the number of internal neural units, and the internal framework includes a forget gate, an input gate, and an output gate, and outputs the results to the fully connected layer. The fully connected layer linearly combines the thermal comfort results extracted by the LSTM layer, extracts and combines complex features for final prediction, and outputs the results to the output layer. The output layer can output the thermal comfort prediction values ​​under different parameters according to the training of the model.

[0040] 4. Furthermore, in order to determine the optimal hyperparameters in the CNN-LSTM model and improve the prediction effect of the model, the grid search method is used for screening, and the root mean square error (RMSE) is selected as the evaluation indicator. The calculation formula is as follows:

[0041]

[0042] Among them, y i represents the calculated value of the i-th grid point on the envelope surface obtained by numerical simulation, It represents the predicted value of the i-th grid point on the envelope surface output by the output layer, and n represents the total number of grid points on the envelope surface. The Dropout method is used in the model to randomly discard neurons in the CNN-LSTM model to avoid the problem of overfitting of the neural network, and the Adam optimizer is used for algorithm training.

[0043] 5. Furthermore, the grid search method is used to set the value range of each key hyperparameter, and the algorithm is cycled in sequence to traverse all hyperparameter combinations; under each hyperparameter combination, cross-validation is used to divide the predicted average voting PMV and the wind dissatisfaction rate DR data set of the sample working condition into n subsets and use them as validation sets in turn, and the remaining data is used as training sets. After the cross-validation is completed, the average RMSE of the validation set of n trainings under each hyperparameter combination is calculated. Since cross-validation is evaluated on the full set of training data, the accuracy and reliability of the results are enhanced. Based on the RMSE results of the validation set, the key hyperparameters of CNN-LSTM are determined. When the average RMSE under a certain hyperparameter combination is the smallest, it is the key hyperparameter determined by CNN-LSTM.

[0044] 6. Choose MAE (mean absolute error), RMSE (root mean square error) and R 2 The coefficient of determination is used as an indicator to evaluate the prediction performance of the algorithm model. MAE and R 2 The specific calculation formula is as follows:

[0045]

[0046] where y i represents the calculated value of the i-th grid point on the envelope surface obtained by numerical simulation, Represents the predicted value of the i-th grid point on the bag surface obtained by the CNN-LSTM model, It represents the calculated average value of all grid points on the envelope surface obtained by numerical simulation, and n represents the total number of grid points on the envelope surface.

[0047] The numerical results of the test set are compared with the predicted values ​​based on the CNN-LSTM model. The PMV value requires that the average MAE and the average RMSE are not greater than 0.1, and the average R 2 Not less than 0.9; and the DR value, the results of the two must satisfy the average MAE is not greater than 1, the average RMSE is not greater than 1, and the average R 2 If the value is not less than 0.9, the prediction result is considered to meet the accuracy requirements, otherwise the number of levels of the orthogonal experiment needs to be increased to increase the number of sample conditions until the prediction value and the test set result meet the required accuracy requirements.

[0048] 7. Based on the CNN-LSTM model constructed above, the thermal comfort of passengers in the car under different air supply parameters and external environmental parameters, namely, PMV and DR values, can be quickly predicted.

[0049] In one example, a certain type of train passenger compartment model is used as an example, see Figure 3 There are two rows of seats in the train passenger compartment, each row has four seats, and a human body model 1 is placed on each seat to simulate a full load. A human body envelope 2 is set at a distance of about 10 cm from the surface of the human body model to monitor the flow field information and then calculate the thermal comfort index PMV and DR value.

[0050] There are 5 design parameters for the example working condition, of which there are 4 ventilation parameters, namely upper air supply ratio, air supply angle, air supply temperature, and air supply humidity. The external environmental parameter is the outdoor air temperature. The upper air supply ratio ranges from 50% to 100%, the air supply angle ranges from -45° to 45°, the air supply temperature ranges from 20℃ to 26℃, the air supply humidity ranges from 30% to 60%, and the outdoor air temperature ranges from 30℃ to 36℃. The orthogonal experimental design sample working condition is adopted, and each parameter adopts 7 levels, requiring a total of 49 sample working conditions, as shown in Table 1.

[0051] Table 1 Numerical calculation sample working condition parameters

[0052]

[0053]

[0054] The simulation results of 49 sample working conditions in Table 1 were obtained through numerical simulation. Figure 3 ) is used to calculate the thermal comfort index PMV and DR values ​​of the sample working condition through the user-defined UDF function. Since there are 49 sample working conditions in total, a seven-fold cross validation is used to divide the PMV and DR data sets of the sample working conditions into 7 subsets and use them as validation sets in turn, and the remaining data are used as training sets. After the cross validation is completed, the error data of the validation set is the average value of the RMSE of the seven training validation sets. The hyperparameter values ​​of CNN-LSTM are determined by the average RSME results of the validation set, as shown in Table 2.

[0055] Table 2 CNN-LSTM architecture and hyperparameter optimization

[0056]

[0057] In order to verify the reliability and accuracy of the prediction method of thermal comfort of airflow environment in train passenger compartment based on CNN-LSTM model, three additional test sets (as shown in Table 3) were selected to carry out numerical simulation calculations and compare with the predicted values. Table 4 compares the simulation results obtained from the test set with the prediction results of the method proposed in this invention. It can be found that for PMV and DR values, MAE, RMSE and R 2 All meet the specified accuracy requirements.

[0058] Table 3 Test conditions parameters

[0059]

[0060] Table 4 Comparison of numerical calculation results of test conditions and CNN-LSTM prediction results

[0061]

[0062]

[0063] Compared with the traditional numerical simulation method, the efficient method for predicting thermal comfort of airflow environment in train passenger compartment proposed in the present invention can quickly predict the thermal comfort results under different parameters with good accuracy. Specifically, for this embodiment, the CNN-LSTM model obtained by training the sample working conditions can predict the thermal comfort within each parameter range. If the interval of the upper air supply ratio is 5% (11), the interval of the air supply angle is 5° (19), the interval of the air supply temperature is 0.2°C (31), the interval of the air supply humidity is 2% (16), and the interval of the outdoor temperature is 0.2°C (31), the thermal comfort under 11×19×31×16×31=3213584 working conditions can be predicted. It takes about 4 hours to numerically calculate each working condition. If 3213584 working conditions need to be calculated, it takes about 1.29×10 7 hours. The method proposed in this invention takes about 2 hours for model training plus the time for preliminary data processing, plus 196 hours for numerical calculation of 49 sample conditions, for a total of about 200 hours. 7 The method proposed in the present invention can significantly reduce the time of thermal comfort prediction and significantly improve the design efficiency.

[0064] In addition, in order to further verify the advantages of the method proposed in the present invention, it is compared with a variety of commonly used machine learning algorithms, and the selected machine learning algorithms are BP (back propagation algorithm), SVM (support vector machine) and traditional CNN (convolutional neural network). The number of numerical samples selected by all machine learning algorithms is 49 in Table 1, and the differences between the prediction results and the simulation results on the three test sets, as well as the calculation time of each model, are compared. Figure 4 and Figure 5 The errors of PMV and DR values ​​predicted by the four models were compared respectively, and it can be found that the efficient method for predicting thermal comfort of airflow environment in train passenger compartment proposed in the present invention has higher prediction accuracy. Figure 6 By comparing the duration of a single training of the four models, it can be found that the method for predicting thermal comfort of airflow environment in a train passenger compartment based on the CNN-LSTM model proposed in the present invention requires less time and has higher efficiency.

[0065] The present application also provides a train passenger compartment airflow environment thermal comfort prediction system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program. The train passenger compartment airflow environment thermal comfort prediction system can implement the various embodiments of the train passenger compartment airflow environment thermal comfort prediction method described above, and can achieve the same beneficial effects, which will not be described in detail here.

[0066] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for predicting thermal comfort of airflow environment in a train passenger compartment, characterized in that: include: S1: Orthogonal experiments are used to design sample working conditions under multiple design variables based on train ventilation parameter variables and external environment parameter variables; S2: using a numerical method to simulate the sample working condition, and monitoring the flow field information of the surface surrounding the human body in the train passenger compartment simulation model; S3: Based on the flow field information monitored on the package surface, a thermal comfort index of the sample working condition is calculated by a user-defined UDF function; S4: constructing a CNN-LSTM model based on the sample working condition and the thermal comfort index; S5: Based on the CNN-LSTM model, the thermal comfort of passengers in the car under different ventilation parameters and external environmental parameters is predicted.

2. The method for predicting thermal comfort of airflow environment in a train passenger compartment according to claim 1, characterized in that: When constructing the CNN-LSTM model in S4, the Dropout method is used to randomly discard neurons in the CNN-LSTM model, and the Adam optimizer is used for algorithm training.

3. The method for predicting thermal comfort of airflow environment in a train passenger compartment according to claim 1, characterized in that: The thermal comfort index includes the values ​​of predicted mean voting PMV and wind dissatisfaction rate DR.

4. The method for predicting thermal comfort of airflow environment in a train passenger compartment according to claim 3, characterized in that: When constructing the CNN-LSTM model, the method further includes: Using the grid search method, we set the value range of each key hyperparameter and loop through all hyperparameter combinations in the algorithm. Under each hyperparameter combination, cross validation is used to divide the predicted average voting PMV and wind dissatisfaction rate DR data set of the sample working condition into n subsets and use them as validation sets in turn, and the rest of the data is used as training sets; After the cross validation is completed, the average value of the root mean square error RMSE of the validation set for n trainings under each hyperparameter combination is calculated; The key hyperparameters of CNN-LSTM are determined based on the average RMS value of the validation set.

5. The method for predicting thermal comfort of airflow environment in a train passenger compartment according to claim 4, characterized in that: The root mean square error RMSE satisfies the following calculation formula: Among them, y i represents the calculated value of the i-th grid point on the envelope surface obtained by numerical simulation, It represents the predicted value of the i-th grid point on the bag surface output by the output layer of the CNN-LSTM model, and n represents the total number of grid points on the bag surface.

6. The method for predicting thermal comfort of airflow environment in a train passenger compartment according to claim 1, characterized in that: After constructing the CNN-LSTM model, the method further includes: Select mean absolute error MAE, root mean square error RMSE and determination coefficient R 2 As an indicator to evaluate the predictive performance of algorithm models; The numerical results of the test set are compared with the predicted values ​​obtained based on the CNN-LSTM model. If the set conditions are met, the constructed CNN-LSTM model is considered to meet the accuracy requirements. If the set conditions are not met, the constructed CNN-LSTM model is considered to not meet the accuracy requirements.

7. The method for predicting thermal comfort of airflow environment in a train passenger compartment according to claim 6, characterized in that: The setting conditions are as follows: For the PMV value, the results of both must satisfy the following conditions: the average MAE is not greater than the first threshold, the average RMSE is not greater than the first threshold, and the average R 2 is not less than the second threshold; and the DR value requires that the results of the two satisfy that the average MAE is not greater than the third threshold, the average RMSE is not greater than the third threshold, and the average R 2 is not less than a second threshold; The first threshold is smaller than the second threshold, and the second threshold is smaller than the third threshold.

8. The method for predicting thermal comfort of airflow environment in a train passenger compartment according to claim 6, characterized in that: In the case that the constructed CNN-LSTM model does not meet the accuracy requirement, the method further includes: Increase the number of levels of the orthogonal experiment to increase the number of sample conditions until the comparison between the predicted value of the trained CNN-LSTM model and the numerical result of the test set meets the required accuracy requirements.

9. The method for predicting thermal comfort of airflow environment in a train passenger compartment according to claim 6, characterized in that: The MAE and R 2 Satisfies the following calculation formula: Among them, y i represents the calculated value of the i-th grid point on the envelope surface obtained by numerical simulation, Represents the predicted value of the i-th grid point on the bag surface obtained by the CNN-LSTM model, It represents the calculated average value of all grid points on the envelope surface obtained by numerical simulation, and n represents the total number of grid points on the envelope surface.

10. A train passenger compartment airflow environment thermal comfort prediction system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method described in any one of claims 1 to 9 are implemented.

Images