Air speed regulation and control method and system for tunnel air supply outlet based on deep learning

Abstract

The invention relates to the technical field of deep learning, in particular to an air speed regulation and control method and system for a tunnel air supply outlet based on deep learning. The method comprises the following steps: acquiring environmental parameter data at different time points in a tunnel; preprocessing the acquired environment parameter data; performing data set division based on the environmental parameter data to obtain a training set and a test set; constructing an air speed regulation and control model for the tunnel air supply outlet based on deep learning; performing feature extraction on the data set by using a wind speed regulation and control model for the tunnel air supply outlet; predicting the extracted features by using a long-short-term memory network to obtain a prediction result; and performing wind speed regulation and control according to a prediction result. The wind speed regulation and control model for the tunnel air supply outlet is constructed based on deep learning, and the complex non-linear relation between the environment parameters and the wind speed can be effectively captured.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to a wind speed control method and system for a tunnel air outlet based on deep learning. Background Art

[0002] In the modern transportation system, tunnels, as important infrastructure, play a key role in promoting regional economic development and ensuring transportation safety. However, ventilation in tunnels has always been a major challenge. Good ventilation can not only ensure that the air quality in the tunnel meets the standards and provide a safe and comfortable driving environment for drivers and passengers, but also prevent the accumulation and spread of harmful gases in emergency situations such as fires.

[0003] Traditional methods for controlling wind speed at tunnel air outlets mainly rely on fixed empirical formulas and preset thresholds, which lack adaptability to complex and changing tunnel environments. For example, adjusting wind speed based on simple traffic flow or time patterns often cannot accurately meet actual ventilation needs, which may lead to excessive ventilation that increases energy consumption, or insufficient ventilation that affects air quality.

[0004] With the continuous advancement of science and technology, deep learning technology has shown great potential in various fields. Deep learning models have strong learning and adaptive capabilities, and can automatically extract features from large amounts of data and make accurate predictions. In the field of tunnel ventilation, by using deep learning technology, various environmental parameters in the tunnel can be monitored and analyzed in real time, including temperature, humidity, traffic flow, pollutant concentration, etc. There is a complex nonlinear relationship between these parameters and the wind speed of the air outlet, which is difficult to accurately model with traditional methods, while deep learning models can effectively capture these relationships, providing a new way to achieve precise wind speed control.

[0005] In addition, with the continuous development of sensor technology, more and more high-precision sensors can be installed in tunnels to collect rich environmental data in real time. These data provide a solid foundation for the training and application of deep learning models. At the same time, high-speed data transmission and processing technology also makes real-time wind speed control possible, further improving the efficiency and reliability of tunnel ventilation systems.

[0006] In summary, a wind speed control method for tunnel air outlets based on deep learning has emerged, aiming to use advanced technical means to solve the shortcomings of traditional ventilation methods, realize the intelligence, efficiency and energy saving of tunnel ventilation, and provide strong guarantees for the safe operation and sustainable development of tunnels. Summary of the invention In order to solve the above-mentioned problems, the present invention provides a wind speed control method and system for tunnel air outlets based on deep learning.

[0007] In the first aspect, the present invention provides a method for controlling wind speed at a tunnel air outlet based on deep learning, which adopts the following technical solution: A method for controlling wind speed at a tunnel air outlet based on deep learning, comprising: Obtain environmental parameter data at different time points in the tunnel; Preprocessing the acquired environmental parameter data; Divide the data set based on environmental parameter data to obtain training set and test set; Construct a wind speed control model for tunnel air outlets based on deep learning; The wind speed control model of the tunnel air outlet is used to extract features from the data set; The extracted features are predicted using the long short-term memory network to obtain the prediction results; The wind speed is controlled according to the forecast results.

[0008] Furthermore, the obtaining of environmental parameters at different time points in the tunnel includes collecting environmental parameters at different time points in the tunnel based on sensors in the tunnel, including temperature sensors, humidity sensors, vehicle flow sensors, and pollutant concentration sensors, and transmitting the environmental parameters to a data processing center wirelessly.

[0009] Furthermore, the acquired environmental parameter data is preprocessed, including removing outliers and filling missing values ​​through data cleaning; filtering out interference noise in the data through data denoising; and mapping the data to a linear interval through linear normalization to facilitate feature extraction.

[0010] Furthermore, the construction of the wind speed control model for the tunnel air outlet based on deep learning includes constructing the wind speed control model for the tunnel air outlet based on a combination of a convolutional neural network and a recurrent neural network, wherein the convolutional neural network includes constructing an input layer, a convolutional layer, a pooling layer and an activation function; and the recurrent neural network includes an RNN layer and an output layer.

[0011] Furthermore, the wind speed control model of the tunnel air outlet is used to extract features of the data set, including performing a convolution operation on the input data based on sliding the convolution kernel in the time dimension, extracting local features in time, and using a pooling layer to reduce the dimension of the features, thereby reducing the amount of calculation and the risk of overfitting; the features extracted by the convolutional neural network are used as the input of the recurrent neural network, and the hidden state at the current moment is updated according to the input data and the hidden state at the previous moment, thereby capturing the dynamic features in the time series.

[0012] Furthermore, the method of using the long short-term memory network to predict the extracted features and obtain the prediction results includes determining the input layer and LSTM layer of the long short-term memory network, adding a fully connected layer after the LSTM layer to perform feature fusion, and performing prediction based on the feature fusion results.

[0013] Furthermore, the wind speed control according to the prediction result includes formulating a control strategy according to the prediction result, including control based on a fixed threshold, and taking corresponding control instructions when the prediction result is higher than the set threshold.

[0014] In a second aspect, a wind speed control system for a tunnel air outlet based on deep learning includes: The data acquisition module is configured to acquire environmental parameter data at different time points in the tunnel; A preprocessing module is configured to preprocess the acquired environmental parameter data; A data set partitioning module is configured to partition the data set based on the environmental parameter data to obtain a training set and a test set; A model building module, configured to build a wind speed control model for a tunnel air outlet based on deep learning; A feature extraction module is configured to extract features from a data set using a wind speed control model for a tunnel air outlet; A prediction module is configured to predict the extracted features using a long short-term memory network to obtain a prediction result; The control module is configured to control the wind speed according to the prediction results.

[0015] In a third aspect, the present invention provides a computer-readable storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device for a method for controlling wind speed for a tunnel air outlet based on deep learning.

[0016] In a fourth aspect, the present invention provides a terminal device comprising a processor and a computer-readable storage medium, the processor being used to implement various instructions; the computer-readable storage medium being used to store multiple instructions, the instructions being suitable for being loaded by the processor and executing the method for controlling wind speed for a tunnel air outlet based on deep learning.

[0017] In summary, the present invention has the following beneficial technical effects: The wind speed control model for tunnel air outlets constructed based on deep learning can effectively capture the complex nonlinear relationship between environmental parameters and wind speed. The combination of convolutional neural networks and recurrent neural networks fully utilizes the advantages of both, can extract spatial and temporal features in time series data, and improve the prediction accuracy of the model.

[0018] During the feature extraction process, the convolution kernel slides in the time dimension to extract local features, and the pooling layer reduces the feature dimension, reducing the amount of calculation and the risk of overfitting. The recurrent neural network updates the hidden state at the current moment based on the input data and the hidden state at the previous moment, capturing the dynamic features in the time series, enabling the model to better adapt to changes in the tunnel environment.

[0019] The extracted features are predicted using a long short-term memory network, which further improves the accuracy of the prediction results. LSTM can handle long-term dependencies in time series data, and by adding a fully connected layer for feature fusion, the prediction results are more reliable.

[0020] According to the prediction results, a control strategy is formulated, such as control based on a fixed threshold, so that corresponding control instructions can be taken in time to achieve precise wind speed control, ensuring that the air quality in the tunnel meets the standard while avoiding energy waste caused by excessive ventilation. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic diagram of a wind speed control method for a tunnel air outlet based on deep learning according to Example 1 of the present invention. DETAILED DESCRIPTION

[0022] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0023] Example 1 Reference Figure 1 , a wind speed control method for a tunnel air outlet based on deep learning in this embodiment includes: S1. Install multiple sensors in the tunnel, including temperature sensors, humidity sensors, vehicle flow sensors, pollutant concentration sensors, etc., to collect environmental parameters in the tunnel in real time.

[0024] The sensor transmits the collected data to the data processing center via wired or wireless means.

[0025] S2. Perform data preprocessing on the acquired environmental parameters: Removing outliers Analyze the value range and reasonable interval of each environmental parameter. Data points that are obviously beyond the normal range, such as too high or too low temperature, abnormally large or negative traffic volume, etc., are considered outliers.

[0026] Use statistical methods, such as the triple standard deviation method or the box plot method, to identify outliers. For the identified outliers, you can choose to delete them or replace them with a reasonable interpolation method. If the temperature sensor reading at a certain moment deviates significantly from the temperature range at other moments, you can first check whether the sensor is faulty. If the sensor is normal, you can replace the outlier by linear interpolation based on the temperature values ​​at the adjacent moments.

[0027] Handling missing values Environmental parameter data may be missing due to sensor failure, data transmission interruption, etc. For a small number of missing values, interpolation methods can be used to fill them, such as linear interpolation, spline interpolation, etc. If there are many missing values, you can consider using model-based methods, such as using other related parameters for regression prediction to fill the missing values. If the traffic flow sensor data is missing for a certain period of time, a regression model can be established based on the historical traffic flow data of the same period and other related parameters (such as time, weather, etc.) to predict the missing traffic flow values.

[0028] There may be noise interference in the environmental parameter data, which affects the quality of the data and the training effect of the model. Filtering methods can be used to remove noise. For temperature data, median filtering is used to sort the temperature value at a certain moment with the temperature values ​​at several moments before and after it, and the middle value is taken as the filtered temperature value, which can effectively remove sudden noise interference.

[0029] Data Normalization The data is mapped to the interval [0,1] or [-1,1]. For each environmental parameter, its maximum and minimum values ​​are calculated respectively.

[0030] Then, for each data point, use the formula Map the data points to the [0,1] interval and perform normalization.

[0031] For temperature data, let its maximum value be Tmax, its minimum value be Tmin, and the temperature value at a certain moment be T. Then the normalized temperature value is .

[0032] Standardization: Transform the data into a distribution with a mean of 0 and a standard deviation of 1. For each environmental parameter, calculate its mean and standard deviation.

[0033] Then, for each data point, use the formula Standardization is performed, where μ is the mean and σ is the standard deviation. For traffic flow data, calculate its mean μ flow Standard deviation flow , the traffic flow value at a certain moment is F, then the standardized traffic flow value is .

[0034] S3. Construct a wind speed control model for tunnel air outlets based on deep learning, including: Input layer: Determine the shape of the input data, usually multiple feature channels of time series data, and reshape the data into a shape suitable for CNN input, such as (batch_size, time_steps, features).

[0035] Convolutional layer: Add convolutional layers to extract spatial features from the input data. Different convolution kernel sizes and numbers can be used to capture features at different scales. Use a one-dimensional convolutional layer to process time series data. The convolution kernel slides in the time dimension to extract local features in time.

[0036] Pooling layer: A pooling layer is added after the convolution layer to reduce the dimension of features, reduce the amount of computation and the risk of overfitting. Pooling is performed in the time dimension through maximum pooling and average pooling.

[0037] Set activation function: Use activation function in convolutional layers and fully connected layers to introduce nonlinearity. Common activation functions include ReLU, Sigmoid, Tanh, etc.

[0038] RNN construction: Connect CNN and RNN: Use the features extracted by CNN as the input of RNN. The output of CNN can be flattened or processed appropriately before being input into RNN.

[0039] RNN layer: Choose an appropriate RNN type, such as Long Short-Term Memory (LSTM) or Gated Recurrent Unit (GRU). These RNN types are able to handle long-term dependencies in time series data. Set the number of hidden layers, number of neurons, and time step of the RNN.

[0040] Output layer: Determine the type of output layer based on the task requirements. For wind speed control problems, a linear output layer can be used to predict the wind speed value at the air outlet. S4. The wind speed control model of the tunnel air outlet is used to extract features from the dataset. The convolution layer in the model automatically learns the local features in the data. Different convolution kernel sizes and numbers can capture features of different scales.

[0041] The convolution kernel slides in the time dimension, performs convolution operations on the input data, and extracts local features in time. For temperature data, the convolution layer learns the temperature change pattern in different time periods. The pooling layer is used to reduce the dimension of features, reduce the amount of calculation and the risk of overfitting. Common pooling operations include maximum pooling and average pooling, which can be performed in the time dimension to extract more representative features. Maximum pooling can be used to extract the most significant temperature change features in the time series.

[0042] Connect the output of the convolutional neural network through a recurrent neural network: The features extracted by the convolutional neural network are used as the input of the recurrent neural network. The output of the convolutional neural network can be flattened or otherwise processed appropriately before being input into the recurrent neural network. The local features extracted by the convolutional neural network are used in combination with the recurrent neural network to model the long-term dependencies of time series data.

[0043] Recurrent neural networks (such as long short-term memory networks or gated recurrent units) can handle long-term dependencies in time series data. The recurrent layer updates the hidden state at the current moment based on the input data and the hidden state at the previous moment, thereby capturing the dynamic features in the time series. For traffic flow data, recurrent neural networks can learn the changing trends of traffic flow in different time periods and the long-term relationship with other parameters.

[0044] The last hidden state or output of the recurrent neural network can be used as the extracted features. These features contain the time series information and long-term dependencies in the input data, which can be used for subsequent tasks such as wind speed prediction or air outlet control.

[0045] Use visualization tools such as t-SNE or PCA to reduce the dimension of the extracted features and visualize them to better understand the distribution and structure of the features. Through visualization, you can observe the clustering of different parameter data in the feature space and the correlation between features.

[0046] S5. Use the long short-term memory network to predict the extracted features and obtain the prediction results, including determining the input layer and LSTM layer of the long short-term memory network, adding a fully connected layer after the LSTM layer to perform feature fusion, and making predictions based on the feature fusion results.

[0047] Here, the feature vectors extracted from different data sources are simply concatenated together. For example, if there are temperature feature vectors, humidity feature vectors, traffic flow feature vectors, etc., they can be concatenated in sequence into a longer feature vector.

[0048] It is necessary to ensure that the scales and value ranges of different features are relatively consistent, otherwise it may affect the performance of the model. You can normalize the features before splicing.

[0049] The temperature eigenvector is , the humidity feature vector is , the traffic flow feature vector is , then the concatenated feature vector is .

[0050] Assign weights to different features and then sum them up to get the fused features. The weights can be set empirically, adjusted manually, or automatically learned using an optimization algorithm.

[0051] The temperature feature vector is T, the humidity feature vector is H, the traffic flow feature vector is C, and the corresponding weights are , , , then the fused features are .

[0052] The attention mechanism is introduced to allow the model to automatically learn the importance weights of different features. The attention mechanism can dynamically adjust the weight distribution according to the contribution of different parts of the input features to the prediction target.

[0053] For temperature, humidity, and traffic flow features, the attention mechanism is used to calculate the attention weight of each feature, and then the feature is multiplied by the weight for fusion. This can highlight important features, suppress unimportant features, and improve the performance and robustness of the model.

[0054] S6. Clearly define the control objectives: The main objectives are to ensure good air quality in the tunnel, reduce energy consumption and ensure driving safety. Specifically, the air outlet speed should be adjusted to a suitable value according to the actual situation in the tunnel to meet the ventilation needs while avoiding energy waste caused by excessive ventilation.

[0055] Formulate control strategies: Use different control strategies, such as control based on fixed thresholds, control based on proportional-integral-derivative (PID) control, control based on model predictive control (MPC), etc. Select the appropriate control strategy based on actual conditions.

[0056] Analyze the predicted wind speed value: Based on the air outlet wind speed value predicted by the deep learning model, understand the required wind speed in the current tunnel environment. If the prediction result is a specific value, you can directly use this value as the target wind speed for regulation.

[0057] In addition to the predicted wind speed, other factors need to be considered, such as the real-time traffic volume in the tunnel, the trend of pollutant concentration changes, weather conditions, etc. These factors may affect the ventilation demand and need to be considered in wind speed control.

[0058] Example 2 This embodiment provides a wind speed control system for a tunnel air outlet based on deep learning, including: The data acquisition module is configured to acquire environmental parameter data at different time points in the tunnel; A preprocessing module is configured to preprocess the acquired environmental parameter data; A data set partitioning module is configured to partition the data set based on the environmental parameter data to obtain a training set and a test set; A model building module, configured to build a wind speed control model for a tunnel air outlet based on deep learning; A feature extraction module is configured to extract features from a data set using a wind speed control model for a tunnel air outlet; A prediction module is configured to predict the extracted features using a long short-term memory network to obtain a prediction result; The control module is configured to control the wind speed according to the prediction results.

[0059] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, for a method for controlling wind speed for a tunnel air outlet based on deep learning.

[0060] A terminal device includes a processor and a computer-readable storage medium, the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded and executed by the processor to describe a wind speed control method for a tunnel air outlet based on deep learning.

[0061] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for controlling wind speed at a tunnel air outlet based on deep learning, characterized in that: include: Obtain environmental parameter data at different time points in the tunnel; Preprocessing the acquired environmental parameter data; Divide the data set based on environmental parameter data to obtain training set and test set; Construct a wind speed control model for tunnel air outlets based on deep learning; The wind speed control model of the tunnel air outlet is used to extract features from the data set; The extracted features are predicted using the long short-term memory network to obtain the prediction results; The wind speed is controlled according to the forecast results.

2. According to a deep learning-based wind speed control method for a tunnel air outlet according to claim 1, it is characterized in that: The obtaining of environmental parameters at different time points in the tunnel includes collecting environmental parameters at different time points in the tunnel based on sensors in the tunnel, including temperature sensors, humidity sensors, vehicle flow sensors, and pollutant concentration sensors, and transmitting the environmental parameters to a data processing center wirelessly.

3. The method for controlling wind speed at a tunnel air outlet based on deep learning according to claim 2, characterized in that: The obtained environmental parameter data is preprocessed, including removing abnormal values ​​and filling missing values ​​through data cleaning; filtering out interference noise in the data through data denoising; Linear normalization is used to map the data to a linear interval to facilitate feature extraction.

4. The method for controlling wind speed at a tunnel air outlet based on deep learning according to claim 3, characterized in that: The method for constructing a wind speed control model for a tunnel air outlet based on deep learning includes constructing a wind speed control model for a tunnel air outlet based on a combination of a convolutional neural network and a recurrent neural network, wherein the convolutional neural network includes constructing an input layer, a convolutional layer, a pooling layer and an activation function; and the recurrent neural network includes an RNN layer and an output layer.

5. The method for controlling wind speed at a tunnel air outlet based on deep learning according to claim 4, characterized in that: The wind speed control model of the tunnel air outlet is used to extract features of the data set, including sliding the convolution kernel on the time dimension, performing a convolution operation on the input data, extracting local features in time, and using a pooling layer to reduce the dimension of the features, thereby reducing the amount of calculation and the risk of overfitting; the features extracted by the convolutional neural network are used as the input of the recurrent neural network, and the hidden state at the current moment is updated according to the input data and the hidden state at the previous moment, thereby capturing the dynamic features in the time series.

6. A method for controlling wind speed at a tunnel air outlet based on deep learning according to claim 5, characterized in that: The method of using the long short-term memory network to predict the extracted features and obtain the prediction results includes determining the input layer and LSTM layer of the long short-term memory network, adding a fully connected layer after the LSTM layer to perform feature fusion, and performing prediction based on the feature fusion results.

7. The method for controlling wind speed at a tunnel air outlet based on deep learning according to claim 6, characterized in that: The wind speed control according to the prediction result includes formulating a control strategy according to the prediction result, including control based on a fixed threshold, and taking corresponding control instructions when the prediction result is higher than the set threshold.

8. A wind speed control system for tunnel air outlets based on deep learning, characterized in that: include: The data acquisition module is configured to acquire environmental parameter data at different time points in the tunnel; A preprocessing module is configured to preprocess the acquired environmental parameter data; A data set partitioning module is configured to partition the data set based on the environmental parameter data to obtain a training set and a test set; A model building module, configured to build a wind speed control model for a tunnel air outlet based on deep learning; A feature extraction module is configured to extract features from a data set using a wind speed control model for a tunnel air outlet; A prediction module is configured to predict the extracted features using a long short-term memory network to obtain a prediction result; The control module is configured to control the wind speed according to the prediction result.

9. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instruction is suitable for being loaded and executed by a processor of a terminal device according to a deep learning-based wind speed control method for a tunnel air outlet as described in claim 1.

10. A terminal device, comprising a processor and a computer-readable storage medium, wherein the processor is used to implement each instruction; and the computer-readable storage medium is used to store multiple instructions, characterized in that: The instructions are suitable for being loaded by a processor and executed by a wind speed control method for a tunnel air outlet based on deep learning as described in claim 1.

Images