Tunnel ventilation fan control method and system based on data analysis

Abstract

The present application provides a tunnel ventilation fan control method and system based on data analysis. First, a tunnel scene detection data combination including the first and second tunnel scene detection data and the corresponding sample fan control strategy data are obtained, and then the tunnel scene parameter vectors of the first and second tunnel scene detection data are respectively extracted, and the corresponding ventilation fan control data are generated. Next, the first loss function value between the ventilation fan control data and the corresponding sample fan control strategy data under different tunnel scene definition modes, and the second loss function value between different tunnel scene parameter vectors are calculated, and the parameters of the ventilation fan control model are updated accordingly, so as to make ventilation fan control strategy decisions for the tunnel section corresponding to the tunnel scene detection data under any tunnel scene definition mode. As a result, the accuracy and efficiency of tunnel ventilation control are effectively improved.

Description

Technical Field

[0001] The present application relates to the technical field of tunnel ventilation, and in particular to a method and system for controlling a tunnel ventilation fan based on data analysis. Background Art

[0002] Tunnel ventilation system is an important part to ensure the safe and comfortable operation of the tunnel. In the existing technology, the control of tunnel ventilation system mainly relies on manual operation or preset strategies. However, due to the complex and changeable environment in the tunnel, factors such as vehicle flow, air quality, and climate conditions will affect the ventilation demand in the tunnel, and these factors are often difficult to be accurately captured and processed manually.

[0003] In addition, even if some automated control technologies are adopted, such as rule-based control systems or traditional feedback control algorithms, these methods often cannot meet the high requirements for ventilation control accuracy and efficiency in practical applications because their design and adjustment usually require a large number of people and have limited ability to respond to various complex situations.

[0004] In order to solve the above problems, some users have tried to introduce advanced technologies such as machine learning in recent years to learn ventilation control strategies in various tunnel scenarios through training models. However, these methods currently usually only consider a single type of tunnel scenario and lack the ability to comprehensively consider and process different tunnel scenarios. Therefore, they may not achieve the desired effect in practical applications.

[0005] Therefore, there is an urgent need for a ventilation fan control model that can effectively handle a variety of tunnel scenarios with high accuracy and efficiency. Summary of the invention

[0006] In view of this, the purpose of the present application is to provide a tunnel ventilation fan control method and system based on data analysis.

[0007] According to a first aspect of the present application, a tunnel ventilation fan control method based on data analysis is provided, which is applied to a tunnel ventilation fan control system based on data analysis, and the method comprises:

[0008] Acquire a tunnel scene detection data combination and sample fan control strategy data of the tunnel scene detection data combination, wherein the tunnel scene detection data combination includes first tunnel scene detection data and second tunnel scene detection data; the first tunnel scene detection data and the second tunnel scene detection data are different data obtained by performing tunnel scene parameter detection on a tunnel section in a same tunnel ventilation task using different tunnel scene definition modes, and the sample fan control strategy data is used to mark the control parameter data of the fan control strategy of the ventilation fan of the tunnel section corresponding to the corresponding tunnel scene detection data;

[0009] extracting tunnel scene parameter vectors of the first tunnel scene detection data and the second tunnel scene detection data respectively, and generating ventilation fan control data corresponding to the first tunnel scene detection data and the second tunnel scene detection data respectively based on the tunnel scene parameter vectors corresponding to the first tunnel scene detection data and the second tunnel scene detection data respectively;

[0010] Obtain a first loss function value between ventilation fan control data of tunnel scene detection data of different tunnel scene definition modes and corresponding sample fan control strategy data, and obtain a second loss function value between different tunnel scene parameter vectors based on the tunnel scene parameter vectors of different tunnel scene detection data;

[0011] The parameters of the ventilation fan control model are updated according to the training learning strategy of minimizing the first loss function value and the second loss function value; the ventilation fan control model after parameter update is used to make ventilation fan control strategy decisions for the tunnel section corresponding to the tunnel scene detection data under any tunnel scene definition mode.

[0012] In a possible implementation of the first aspect, the tunnel scene detection data combination is any one of a plurality of tunnel scene detection data combinations covered by a sample tunnel scene detection data sequence, the tunnel scene parameter vector and the ventilation fan control data are both obtained by calling a ventilation fan control model with initialized weight parameters, wherein the model performance values ​​of the ventilation fan control model used to make ventilation fan control strategy decisions on the first tunnel scene detection data and the second tunnel scene detection data are different, if the model performance value of the ventilation fan control model used to make ventilation fan control strategy decisions on the first tunnel scene detection data is greater than the model performance value of the ventilation fan control model used to make ventilation fan control strategy decisions on the second tunnel scene detection data;

[0013] The step of generating the ventilation fan control model with the initialization weight parameters comprises:

[0014] Extracting at least one first tunnel scene detection data from a plurality of tunnel scene detection data combinations covered by the sample tunnel scene detection data sequence, and acquiring sample fan control strategy data corresponding to each extracted first tunnel scene detection data;

[0015] Using the basic ventilation fan control model, the ventilation fan control strategy decision is made for each first tunnel scene detection data extracted;

[0016] Based on the ventilation fan control data extracted from each first tunnel scene detection data and the corresponding sample fan control strategy data, knowledge learning is performed on the basic ventilation fan control model to generate a ventilation fan control model with initialized weight parameters.

[0017] In a possible implementation of the first aspect, the method further includes:

[0018] If the sample fan control strategy data corresponding to any one of the extracted first tunnel scene detection data is an empty set, determining a target tunnel scene detection data combination corresponding to any one of the first tunnel scene detection data;

[0019] The second tunnel scene detection data associated with any one of the first tunnel scene detection data is obtained from the target tunnel scene detection data combination, and the sample fan control strategy data obtained from the associated second tunnel scene detection data is used as the ventilation fan control data of any one of the first tunnel scene detection data.

[0020] In a possible implementation manner of the first aspect, the ventilation fan control data for any one of the first tunnel scene detection data includes: fan control features of the ventilation fan of the tunnel section in the any one of the first tunnel scene detection data, and the sample fan control strategy data for any one of the first tunnel scene detection data includes: actual control features of the controlled ventilation fan of the tunnel section in the any one of the first tunnel scene detection data;

[0021] The method of performing knowledge learning on the basic ventilation fan control model based on the ventilation fan control data extracted from each first tunnel scene detection data and the corresponding sample fan control strategy data includes:

[0022] Calculating a feature comparison value based on the fan control feature and the actual control feature, the feature comparison value comprising a feature deviation calculated based on each fan control feature and the corresponding actual control feature;

[0023] Based on the characteristic comparison value, the parameters of the basic ventilation fan control model are updated according to a back propagation strategy.

[0024] In a possible implementation of the first aspect, obtaining a second loss function value between different tunnel scene parameter vectors includes:

[0025] Based on the ventilation category prediction network, ventilation category prediction is performed on the tunnel scene parameter vector of the first tunnel scene detection data and the tunnel scene parameter vector of the second tunnel scene detection data respectively, to generate a ventilation control category of the first tunnel scene detection data and a ventilation control category of the second tunnel scene detection data;

[0026] Based on the category loss function value between the ventilation control category of the first tunnel scene detection data and the ventilation control category of the second tunnel scene detection data, a second loss function value between corresponding different tunnel scene parameter vectors is determined.

[0027] In a possible implementation of the first aspect, the ventilation category prediction network is obtained through sample knowledge learning; and the method further includes:

[0028] Acquire a sample tunnel scene detection data sequence, wherein the sample tunnel scene detection data sequence includes a plurality of tunnel scene detection data combinations, each of which includes a first tunnel scene detection data and a second tunnel scene detection data;

[0029] Acquire tunnel scene parameter vectors of each tunnel scene detection data covered by the sample tunnel scene detection data sequence, and call a basic ventilation category prediction network to perform ventilation category prediction on the tunnel scene parameter vectors of each tunnel scene detection data covered by the sample tunnel scene detection data sequence, to generate a ventilation control category for each tunnel scene detection data in the sample tunnel scene detection data sequence;

[0030] The loss function value is calculated based on the ventilation control category of each tunnel scene detection data in the sample tunnel scene detection data sequence and the corresponding prior ventilation control category, and based on the loss function value, the parameters of the basic ventilation category prediction network are updated according to the back propagation strategy.

[0031] In a possible implementation manner of the first aspect, obtaining a first loss function value between ventilation fan control data of tunnel scene detection data of different tunnel scene definition modes and corresponding sample fan control strategy data includes:

[0032] Based on the ventilation fan control data corresponding to the first tunnel scene detection data and the sample fan control strategy data, obtaining a first loss function value between the first tunnel scene detection data and the sample fan control strategy data;

[0033] Based on the ventilation fan control data corresponding to the second tunnel scene detection data and the sample fan control strategy data, a first loss function value between the second tunnel scene detection data and the sample fan control strategy data is obtained.

[0034] In a possible implementation manner of the first aspect, updating the parameters of the ventilation fan control model according to the training learning strategy for minimizing the first loss function value and the second loss function value includes:

[0035] Generate a target loss function value based on a first loss function value between the first tunnel scene detection data and the sample fan control strategy data, a first loss function value between the second tunnel scene detection data and the sample fan control strategy data, and a loss function value of the second loss function value;

[0036] According to minimizing the objective loss function value, updating the parameters of the ventilation fan control model;

[0037] If the model performance value of the ventilation fan control model for making ventilation fan control strategy decisions on the first tunnel scene detection data is greater than the model performance value of the ventilation fan control strategy decisions on the second tunnel scene detection data;

[0038] The first loss function value based on the first tunnel scene detection data and the sample fan control strategy data, the first loss function value between the second tunnel scene detection data and the sample fan control strategy data, and the loss function value of the second loss function value and the generated target loss function value include:

[0039] Respectively obtain a first influence coefficient corresponding to a first loss function value between the first tunnel scene detection data and the sample fan control strategy data, a second influence coefficient corresponding to a first loss function value between the second tunnel scene detection data and the sample fan control strategy data, and a third influence coefficient corresponding to the second loss function value, wherein the second influence coefficient is greater than the first influence coefficient, and the second influence coefficient is greater than the third influence coefficient;

[0040] Based on the first influence coefficient, the first loss function value between the first tunnel scene detection data and the sample fan control strategy data is multiplied, based on the second influence coefficient, the first loss function value between the second tunnel scene detection data and the sample fan control strategy data is multiplied, and based on the third influence coefficient, the second loss function value is multiplied, the loss function values ​​after the product operation are summarized to generate a target loss function value.

[0041] In a possible implementation manner of the first aspect, the tunnel scene detection data combination is any one of a plurality of tunnel scene detection data combinations covered by the sample tunnel scene detection data sequence;

[0042] The updating of the parameters of the ventilation fan control model according to the training learning strategy that minimizes the first loss function value and the second loss function value includes:

[0043] Based on each tunnel scene detection data combination in the sample tunnel scene detection data sequence, calculating a first loss function value and a second loss function value for each tunnel scene detection data combination;

[0044] According to a training learning strategy for minimizing a first loss function value and a second loss function value corresponding to each combination of tunnel scene detection data, weight information of the ventilation fan control model is updated, and when the loss function value between tunnel scene parameter vectors corresponding to different tunnel scene detection data obtained by identifying the updated ventilation fan control model is less than a threshold value, it is determined that the parameter update of the ventilation fan control model is terminated;

[0045] Among them, when the tunnel scene parameter vectors corresponding to the different tunnel scene detection data all conform to the same characteristic distribution, it is determined that the loss function value between the tunnel scene parameter vectors corresponding to the different tunnel scene detection data is less than a threshold value.

[0046] According to the second aspect of the present application, a tunnel ventilation fan control system based on data analysis is provided, wherein the tunnel ventilation fan control system based on data analysis comprises a machine-readable storage medium and a processor, wherein the machine-readable storage medium stores machine-executable instructions, and when the processor executes the machine-executable instructions, the tunnel ventilation fan control system based on data analysis implements the aforementioned tunnel ventilation fan control method based on data analysis.

[0047] According to a third aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed, the aforementioned tunnel ventilation fan control method based on data analysis is implemented.

[0048] According to any of the above aspects, in the present application, first, a tunnel scene detection data combination including the first and second tunnel scene detection data and the corresponding sample fan control strategy data are obtained, and then the tunnel scene parameter vectors of the first and second tunnel scene detection data are respectively extracted, and the corresponding ventilation fan control data are generated. Next, the first loss function value between the ventilation fan control data and the corresponding sample fan control strategy data under different tunnel scene definition modes, and the second loss function value between different tunnel scene parameter vectors are calculated. Finally, the parameters of the ventilation fan control model are updated by a training learning strategy that minimizes the first loss function value and the second loss function value. The ventilation fan control model after the parameter update can be used to make ventilation fan control strategy decisions for the tunnel section corresponding to the tunnel scene detection data under any tunnel scene definition mode. Thereby, the accuracy and efficiency of tunnel ventilation control are effectively improved;

[0049] This application introduces a machine learning model and combines it with a large amount of actual tunnel scene detection data for training, which can more accurately predict and control the ventilation conditions in the tunnel, thereby improving the overall ventilation control accuracy. In addition, traditional manual or rule-based ventilation control methods often require a lot of manpower and time investment, and this application can significantly reduce these requirements, thereby improving the efficiency of ventilation control. Furthermore, this application takes into account different types of tunnel scene definition modes, so that the control model has strong adaptability and flexibility, and can effectively handle various complex and changeable tunnel environments. By optimizing ventilation control, the air quality and environmental conditions in the tunnel can be effectively improved, thereby improving the operating safety of the tunnel and the comfort of the passengers. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0051] Figure 1 A schematic diagram of a flow chart of a tunnel ventilation fan control method based on data analysis provided in an embodiment of the present application;

[0052] Figure 2 A schematic diagram of the component structure of a tunnel ventilation fan control system based on data analysis for implementing the above-mentioned tunnel ventilation fan control method based on data analysis provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0053] To make the purpose, technical scheme and advantages of the embodiment of the present application clearer, the technical scheme in the embodiment of the present application will be clearly and completely described below according to the drawings in the embodiment of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description, and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the embodiment of the present application. It should be understood that the operations of the flowchart can be implemented out of order, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art can add at least one other operation to the flowchart under the guidance of the content of the present application, or destroy at least one operation from the flowchart.

[0054] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. According to the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work correspond to the scope of protection of the present application.

[0055] Figure 1 The flow chart of the tunnel ventilation fan control method and system based on data analysis provided by the embodiment of the present application is shown. It should be understood that in other embodiments, the order of some steps of the tunnel ventilation fan control method based on data analysis of this embodiment can be shared with each other according to actual needs, or some steps can be omitted or maintained. The detailed steps of the tunnel ventilation fan control method based on data analysis include:

[0056] Step S110, obtaining a tunnel scene detection data combination and sample fan control strategy data of the tunnel scene detection data combination.

[0057] In this embodiment, the tunnel scene detection data combination includes first tunnel scene detection data and second tunnel scene detection data. The first tunnel scene detection data and the second tunnel scene detection data are different data obtained by performing tunnel scene parameter detection on a tunnel section in the same tunnel ventilation task using different tunnel scene definition modes, and the sample fan control strategy data is used to mark the control parameter data of the fan control strategy of the ventilation fan of the tunnel section corresponding to the corresponding tunnel scene detection data.

[0058] For example, suppose there is a subway tunnel and a control strategy decision for the ventilation fan is to be made. First, it is necessary to obtain tunnel scene detection data from two different tunnel scene definition modes. For example, data in the first mode is obtained from a tunnel, including tunnel length, width, height, curvature, terrain, geological structure, lighting intensity, noise level, number of vehicles, speed, type, etc.; data in the second mode is also obtained, including air flow rate, humidity, temperature, carbon dioxide content, PM2.5 concentration, oxygen concentration, concentration of harmful gases, wind direction, pressure, etc. There is also a preset fan control strategy data, which guides how to control the fan's switch status, speed, air volume, operation mode, start and stop time, operation frequency, etc. in various scenarios.

[0059] Step S120, respectively extract the tunnel scene parameter vectors of the first tunnel scene detection data and the second tunnel scene detection data, and based on the tunnel scene parameter vectors respectively corresponding to the first tunnel scene detection data and the second tunnel scene detection data, generate ventilation fan control data respectively corresponding to the first tunnel scene detection data and the second tunnel scene detection data.

[0060] For example, for the data obtained from the above two modes, the corresponding parameter vectors will be extracted from the above two sets of data. For the first mode, it may be a vector containing more than 10 physical and environmental parameters; for the second mode, it may be a vector containing more than 10 climate and air quality parameters. Then, based on these parameter vectors, the corresponding fan control data will be generated.

[0061] Step S130, obtain the first loss function value between the ventilation fan control data of the tunnel scene detection data of different tunnel scene definition modes and the corresponding sample fan control strategy data, and based on the tunnel scene parameter vectors of different tunnel scene detection data, obtain the second loss function value between different tunnel scene parameter vectors.

[0062] For example, in actual operation, the fan control data in the first mode may differ from the sample fan control strategy data, that is, there is a certain error, which is the first loss function value. Similarly, the fan control data in the second mode may also differ from the sample data, and there will also be a corresponding loss function value.

[0063] Step S140, updating the parameters of the ventilation fan control model according to the training learning strategy that minimizes the first loss function value and the second loss function value. The ventilation fan control model with updated parameters is used to make ventilation fan control strategy decisions for the tunnel section corresponding to the tunnel scene detection data in any tunnel scene definition mode.

[0064] For example, this step will adjust the parameters of the fan control model through a machine learning algorithm (such as the gradient descent method) so that the first loss function value and the second loss function value are as small as possible. This process may require multiple iterations to complete. After training, an optimized fan control model is obtained. In actual operation, no matter which tunnel scene definition mode is adopted, this model can be used to make fan control strategy decisions based on tunnel scene detection data, such as adjusting the fan speed, switch status, etc., to achieve the best ventilation effect.

[0065] Based on the above steps, first, a tunnel scene detection data combination including the first and second tunnel scene detection data and the corresponding sample fan control strategy data are obtained, and then the tunnel scene parameter vectors of the first and second tunnel scene detection data are respectively extracted, and the corresponding ventilation fan control data is generated. Next, the first loss function value between the ventilation fan control data and the corresponding sample fan control strategy data under different tunnel scene definition modes, as well as the second loss function value between different tunnel scene parameter vectors are calculated. Finally, the parameters of the ventilation fan control model are updated by the training learning strategy that minimizes the first loss function value and the second loss function value. The ventilation fan control model after the parameter update can be used to make ventilation fan control strategy decisions for the tunnel section corresponding to the tunnel scene detection data under any tunnel scene definition mode. Thereby, the accuracy and efficiency of tunnel ventilation control are effectively improved.

[0066] That is, by introducing a machine learning model and combining it with a large amount of actual tunnel scene detection data for training, this application can more accurately predict and control the ventilation conditions in the tunnel, thereby improving the overall ventilation control accuracy. In addition, traditional manual or rule-based ventilation control methods often require a lot of manpower and time investment, and this application can significantly reduce these requirements, thereby improving the efficiency of ventilation control. Furthermore, this application takes into account different types of tunnel scene definition modes, so that the control model has strong adaptability and flexibility, and can effectively handle various complex and changeable tunnel environments. By optimizing ventilation control, the air quality and environmental conditions in the tunnel can be effectively improved, thereby improving the operating safety of the tunnel and the comfort of the passengers.

[0067] In a possible implementation, the tunnel scene detection data combination is any one of a plurality of tunnel scene detection data combinations covered by a sample tunnel scene detection data sequence, and the tunnel scene parameter vector and the ventilation fan control data are both obtained by calling a ventilation fan control model with initialized weight parameters, wherein the model performance values ​​of the ventilation fan control model used to make ventilation fan control strategy decisions on the first tunnel scene detection data and the second tunnel scene detection data are different, if the model performance value of the ventilation fan control model used to make ventilation fan control strategy decisions on the first tunnel scene detection data is greater than the model performance value of the ventilation fan control strategy decisions on the second tunnel scene detection data.

[0068] In actual operation, the scene data in the tunnel may be collected continuously. For example, the physical conditions and air quality parameters of the tunnel are collected once a minute, which forms a sample tunnel scene detection data sequence. From this sequence, the data at any time point can be extracted, which is the tunnel scene detection data combination.

[0069] Therefore, assuming there is a pre-trained fan control model, the parameters of this fan control model (i.e., model weight information) have been initialized. When the tunnel scene detection data is input to this model, it will output the corresponding tunnel scene parameter vector and ventilation fan control data.

[0070] When a fan control model is used to make decisions based on the first tunnel scenario detection data (e.g., physical conditions and environmental parameters), one model performance value may be obtained; and when the same fan control model is used to make decisions based on the second tunnel scenario detection data (e.g., climate and air quality parameters), another model performance value may be obtained. These two performance values ​​are usually different.

[0071] The next step is to generate the ventilation fan control model with initialized weight parameters:

[0072] The step of generating the ventilation fan control model with the initialization weight parameters comprises:

[0073] Step S101, extracting at least one first tunnel scene detection data from a plurality of tunnel scene detection data combinations covered by the sample tunnel scene detection data sequence, and obtaining sample fan control strategy data corresponding to each extracted first tunnel scene detection data.

[0074] For example, the first tunnel scene detection data (ie, physical conditions and environmental parameters) at any time point can be extracted from the continuously collected tunnel scene detection data sequence, and the preset fan control strategy data corresponding to this time point can be found.

[0075] Step S102: using a basic ventilation fan control model, making a ventilation fan control strategy decision for each first tunnel scene detection data extracted.

[0076] For example, the extracted first tunnel scene detection data is input into the basic ventilation fan control model, and the model outputs a preliminary fan control strategy decision.

[0077] Step S103, based on the ventilation fan control data extracted from each first tunnel scene detection data and the corresponding sample fan control strategy data, the basic ventilation fan control model is subjected to knowledge learning to generate a ventilation fan control model with initialized weight parameters.

[0078] For example, the difference between the fan control strategy decision output by the model and the preset sample fan control strategy data will be compared, and this difference will be used to adjust the parameters of the model (i.e., knowledge learning). After multiple iterations, an optimized ventilation fan control model with initialized weight parameters is finally obtained.

[0079] In a possible implementation, the method further includes:

[0080] Step S104: if the sample fan control strategy data corresponding to any one of the extracted first tunnel scene detection data is an empty set, then determine the target tunnel scene detection data combination corresponding to any one of the first tunnel scene detection data.

[0081] Step S105, obtaining the second tunnel scene detection data associated with any one of the first tunnel scene detection data from the target tunnel scene detection data combination, and using the sample fan control strategy data obtained from the associated second tunnel scene detection data as the ventilation fan control data of any one of the first tunnel scene detection data.

[0082] For example, the first tunnel scene detection data (including physical conditions and environmental parameters) at a specific time point is selected from a continuous tunnel scene detection data sequence. However, when searching for the preset fan control strategy data corresponding to this time point, it is found that the data is empty. In this case, it is necessary to find the time point closest to this time point with complete data, that is, the target tunnel scene detection data combination.

[0083] After finding the target tunnel scene detection data combination, the second tunnel scene detection data (i.e., climate and air quality parameters) associated with the initially selected first tunnel scene detection data can be obtained. Then, the sample fan control strategy data at this time point is used as the ventilation fan control data of the initially selected first tunnel scene detection data. In this way, even if there are gaps in the original data, they can be filled by other related data to ensure that the training and optimization of the fan control model will not be affected.

[0084] In one possible implementation, the ventilation fan control data for any one of the first tunnel scene detection data includes: the fan control characteristics of the ventilation fan of the tunnel section in any one of the first tunnel scene detection data, and the sample fan control strategy data for any one of the first tunnel scene detection data includes: the actual control characteristics of the controlled ventilation fan of the tunnel section in any one of the first tunnel scene detection data.

[0085] For example, the first tunnel scene detection data (including physical conditions and environmental parameters) at a specific time point is selected and input into the basic ventilation fan control model. The model will generate the control characteristics that the fan should have, such as switch status, speed, air volume, etc.

[0086] At the same time, the actual control characteristics of the fan at the same time point can also be found from the preset sample fan control strategy data, such as the actual switch state, speed, air volume, etc.

[0087] Step S103 may include:

[0088] Step S1031 : calculating a feature comparison value based on the fan control feature and the actual control feature, wherein the feature comparison value includes a feature deviation calculated based on each fan control feature and the corresponding actual control feature.

[0089] For example, you can compare the fan control features generated by the model with the actual fan control features and calculate the difference between them. This is the feature comparison value. For example, if the model infers that the fan should run at 1000 rpm, but the fan actually runs at 800 rpm, then the feature deviation is 200 rpm.

[0090] Step S1032: Based on the characteristic comparison value, update the parameters of the basic ventilation fan control model according to the back propagation strategy.

[0091] For example, feature comparison values ​​will be used to adjust the parameters of the fan control model, and this process usually adopts a back-propagation strategy. In this example, the fan speed predicted by the model is higher than the actual speed, so it may be necessary to reduce the values ​​of some parameters in the model so that the next prediction can be closer to the actual speed. After multiple iterations, the prediction performance of the model will gradually improve and eventually reach a level that meets actual needs.

[0092] In a possible implementation, step S130 may include:

[0093] Step S131, based on the ventilation category prediction network, ventilation category prediction is performed on the tunnel scene parameter vector of the first tunnel scene detection data and the tunnel scene parameter vector of the second tunnel scene detection data respectively, to generate the ventilation control category of the first tunnel scene detection data and the ventilation control category of the second tunnel scene detection data.

[0094] For example, the ventilation category prediction network can predict the required ventilation control category based on the input tunnel scene parameter vector. For example, if the first tunnel scene detection data (including physical conditions and environmental parameters) at a time point is input to the network, it may predict that the "strong ventilation" strategy should be adopted at this moment; similarly, if the second tunnel scene detection data (i.e. climate and air quality parameters) at the same time point is input, the network may predict that the "energy-saving ventilation" strategy should be adopted at this moment.

[0095] Step S132, determining a second loss function value between corresponding different tunnel scene parameter vectors based on the category loss function value between the ventilation control category of the first tunnel scene detection data and the ventilation control category of the second tunnel scene detection data.

[0096] For example, the difference between the two predicted ventilation control categories can be compared. If the first tunnel scene detection data predicts "powerful ventilation" and the second tunnel scene detection data predicts "energy-saving ventilation", then there is a certain difference between the two, which is the category loss function value. This category loss function value can be used as the second loss function value for subsequent model optimization and training.

[0097] In a possible implementation, the ventilation category prediction network is obtained through sample knowledge learning. The method further includes:

[0098] Step A110, obtaining a sample tunnel scene detection data sequence, wherein the sample tunnel scene detection data sequence includes a plurality of tunnel scene detection data combinations, and each tunnel scene detection data combination includes a first tunnel scene detection data and a second tunnel scene detection data.

[0099] For example, scene data in a tunnel, including physical conditions and environmental parameters (i.e., first tunnel scene detection data), and climate and air quality parameters (i.e., second tunnel scene detection data), have been continuously collected. These data form a sample tunnel scene detection data sequence.

[0100] Step A120, obtaining the tunnel scene parameter vectors of each tunnel scene detection data covered by the sample tunnel scene detection data sequence, and calling the basic ventilation category prediction network to perform ventilation category prediction on the tunnel scene parameter vectors of each tunnel scene detection data covered by the sample tunnel scene detection data sequence, and generating the ventilation control category of each tunnel scene detection data in the sample tunnel scene detection data sequence.

[0101] For example, data at any time point can be extracted from this sample tunnel scene detection data sequence to obtain tunnel scene parameter vectors. Then, these tunnel scene parameter vectors are input into the basic ventilation category prediction network to output the corresponding ventilation control category.

[0102] Step A130, calculate the loss function value based on the ventilation control category of each tunnel scene detection data in the sample tunnel scene detection data sequence and the corresponding prior ventilation control category, and based on the loss function value, update the parameters of the basic ventilation category prediction network according to the back propagation strategy.

[0103] For example, the difference between the predicted ventilation control category and the prior (i.e. actual) ventilation control category can be compared, which is the loss function value. For example, if the model predicts that the "strong ventilation" strategy should be adopted, but the "energy-saving ventilation" strategy is actually adopted, there is a certain difference. This difference will be used to adjust the parameters of the basic ventilation category prediction network, and the parameters will be updated through the back-propagation strategy. After multiple iterations, the prediction performance of the model will gradually improve, and eventually it can reach a level that meets actual needs.

[0104] In a possible implementation, step S130 may further include:

[0105] Step S133, based on the ventilation fan control data corresponding to the first tunnel scene detection data and the sample fan control strategy data, obtain a first loss function value between the first tunnel scene detection data and the sample fan control strategy data.

[0106] Step S134, based on the ventilation fan control data corresponding to the second tunnel scene detection data and the sample fan control strategy data, obtain a first loss function value between the second tunnel scene detection data and the sample fan control strategy data.

[0107] For example, the fan control data generated by the basic ventilation fan control model based on the first tunnel scene detection data (including physical conditions and environmental parameters) and the preset sample fan control strategy data at the corresponding time point can be compared to calculate the difference between them, which is the first loss function value.

[0108] Similarly, the fan control data generated by the basic ventilation fan control model based on the second tunnel scene detection data (i.e., climate and air quality parameters) and the preset sample fan control strategy data at the corresponding time point will be compared to calculate the difference between them, which is also the first loss function value.

[0109] In a possible implementation, step S140 may include:

[0110] Step S141, based on the first loss function value between the first tunnel scene detection data and the sample fan control strategy data, the first loss function value between the second tunnel scene detection data and the sample fan control strategy data, and the loss function value of the second loss function value, a target loss function value is generated.

[0111] For example, the difference between the predicted ventilation fan control strategy and the actual sample strategy (i.e., the first loss function value) and the difference between the ventilation category predicted by the ventilation category prediction network and the actual ventilation category (i.e., the second loss function value) when processing the first tunnel scene detection data and the second tunnel scene detection data have been calculated. Next, a comprehensive target loss function value will be calculated based on these loss function values.

[0112] Step S142, updating the parameters of the ventilation fan control model according to minimizing the target loss function value.

[0113] For example, the objective loss function value will be used to adjust the parameters of the ventilation fan control model. The parameters are updated through the back-propagation strategy to minimize the difference between the model prediction results and the actual situation.

[0114] If the model performance value of the ventilation fan control model for making ventilation fan control strategy decisions on the first tunnel scene detection data is greater than the model performance value of the ventilation fan control strategy decisions on the second tunnel scene detection data, step S141 may include:

[0115] Step S1411, respectively obtain the first influence coefficient corresponding to the first loss function value between the first tunnel scene detection data and the sample fan control strategy data, the second influence coefficient corresponding to the first loss function value between the second tunnel scene detection data and the sample fan control strategy data, and the third influence coefficient corresponding to the second loss function value, the second influence coefficient is greater than the first influence coefficient, and the second influence coefficient is greater than the third influence coefficient.

[0116] Step S1412, based on the first influence coefficient, multiply the first loss function value between the first tunnel scene detection data and the sample fan control strategy data, based on the second influence coefficient, multiply the first loss function value between the second tunnel scene detection data and the sample fan control strategy data, and multiply the second loss function value based on the third influence coefficient, summarize the loss function values ​​after the product operation, and generate a target loss function value.

[0117] If the model performance value of the ventilation fan control model for making ventilation fan control strategy decisions on the first tunnel scene detection data is greater than the model performance value of the ventilation fan control strategy decisions on the second tunnel scene detection data, this means that the model performs better when processing the first tunnel scene detection data than when processing the second tunnel scene detection data. This may be because the characteristics of the first tunnel scene data are more consistent with the model's preset assumptions, or there is more understanding and knowledge of the data of the first tunnel scene.

[0118] Therefore, different influence coefficients are assigned to the first loss function value and the second loss function value according to their magnitude. Because the second tunnel scene detection data performs poorly in terms of model performance, it is given a larger influence coefficient, which means that the model is expected to pay more attention to this part of the data in the future learning process.

[0119] Then, the loss function values ​​are weighted according to their respective influence coefficients, and the results are summarized to obtain the final target loss function value. This value will be used to guide the optimization and training of the model so that the model can better handle various types of tunnel scene detection data.

[0120] In a possible implementation, the tunnel scene detection data combination is any one of a plurality of tunnel scene detection data combinations covered by the sample tunnel scene detection data sequence. For example, suppose there is a sequence containing a large amount of tunnel scene detection data, each of which includes first tunnel scene detection data (physical conditions and environmental parameters) and second tunnel scene detection data (climate and air quality parameters). During model training, any data combination can be randomly selected from this sequence as input.

[0121] The parameters of the ventilation fan control model are updated according to the training learning strategy of minimizing the first loss function value and the second loss function value, including: based on each tunnel scene detection data combination in the sample tunnel scene detection data sequence, calculating the first loss function value and the second loss function value of each tunnel scene detection data combination. The weight information of the ventilation fan control model is updated according to the training learning strategy of minimizing the first loss function value and the second loss function value corresponding to each tunnel scene detection data combination, and when the loss function value between the tunnel scene parameter vectors corresponding to the different tunnel scene detection data obtained by the updated ventilation fan control model is less than the threshold value, it is determined that the parameter update of the ventilation fan control model is terminated.

[0122] For example, for each combination of tunnel scene detection data in the sample tunnel scene detection data sequence, the first loss function value will be calculated based on the prediction results of the ventilation fan control model and the actual sample strategy data. At the same time, the second loss function value will also be calculated based on the prediction results of the ventilation category prediction network and the actual ventilation category. Then, adjust the parameters of the ventilation fan control model. If after the model parameters are updated, the difference between the new prediction results and the actual situation (i.e., the loss function value) is less than a preset threshold value, then it is considered that the ventilation fan control model has reached a satisfactory performance level and further parameter updates can be stopped.

[0123] Among them, when the tunnel scene parameter vectors corresponding to the different tunnel scene detection data all conform to the same characteristic distribution, it is determined that the loss function value between the tunnel scene parameter vectors corresponding to the different tunnel scene detection data is less than a threshold value.

[0124] For example, this means assuming that all tunnel scene detection data comes from the same type of tunnel, that is, they have the same feature distribution. Only when this assumption is true will the comparison result between the loss function value and the threshold value be used to decide whether to stop the model parameter update. If this assumption is not true, then it may be necessary to train and optimize the model for different types of tunnels separately.

[0125] Figure 2 A tunnel ventilation fan control system 100 based on data analysis which can be used to implement various embodiments described in this application is schematically shown.

[0126] For one embodiment, Figure 2 A data analysis-based tunnel ventilation fan control system 100 is shown, which has at least one processor 102, a control module (chip set) 104 coupled to at least one of the (at least one) processors 102, a memory 106 coupled to the control module 104, a non-volatile memory (NVY) / storage device 108 coupled to the control module 104, at least one input / output device 110 coupled to the control module 104, and a network interface 112 coupled to the control module 104.

[0127] The processor 102 may include at least one single-core or multi-core processor, and the processor 102 may include any combination of general-purpose processors or special-purpose processors (such as graphics processors, application processors, baseband processors, etc.). In an alternative implementation, the tunnel ventilation fan control system 100 based on data analysis can be used as a server device such as a gateway described in the embodiments of the present application.

[0128] Figure 2 A tunnel ventilation fan control system 100 based on data analysis which can be used to implement various embodiments described in this application is schematically shown.

[0129] For one embodiment, Figure 2A data analysis-based tunnel ventilation fan control system 100 is shown, which has at least one processor 102, a control module (chip set) 104 coupled to at least one of the (at least one) processors 102, a memory 106 coupled to the control module 104, a non-volatile memory (NVM) / storage device 108 coupled to the control module 104, at least one input / output device 110 coupled to the control module 104, and a network interface 112 coupled to the control module 104.

[0130] The processor 102 may include at least one single-core or multi-core processor, and the processor 102 may include any combination of general-purpose processors or special-purpose processors (such as graphics processors, application processors, baseband processors, etc.). In an alternative implementation, the tunnel ventilation fan control system 100 based on data analysis can be used as a server device such as a gateway described in the embodiments of the present application.

[0131] In an alternative embodiment, the data analysis-based tunnel ventilation fan control system 100 may include at least one computer-readable medium (e.g., memory 106 or NVM / storage device 108) having instructions 114 and at least one processor 102 that is integrated with the at least one computer-readable medium and configured to execute the instructions 114 to implement a module to perform the actions described in the present disclosure.

[0132] For one embodiment, the control module 104 may include any suitable interface controller to provide any suitable interface to at least one of the processor(s) 102 and / or any suitable device or component in communication with the control module 104 .

[0133] The control module 104 may include a memory controller module to provide an interface to the memory 106. The memory controller module may be a hardware module, a software module, and / or a firmware module.

[0134] The memory 106 may be used, for example, to load and store data and / or instructions 114 for the tunnel ventilation fan control system 100 based on data analysis. For one embodiment, the memory 106 may include any suitable volatile memory, such as a suitable DRAM. In an alternative embodiment, the memory 106 may include a double data rate type four synchronous dynamic random access memory (DDR4 SDRAM).

[0135] For one embodiment, the control module 104 may include at least one input / output controller to provide an interface to the NVM / storage device 108 and the (at least one) input / output device 110 .

[0136] For example, NVM / storage device 108 may be used to store data and / or instructions 114. NVM / storage device 108 may include any suitable non-volatile memory (e.g., flash memory) and / or may include any suitable (at least one) non-volatile storage device (e.g., at least one hard disk drive (HDD), at least one compact disk (CD) drive, and / or at least one digital versatile disk (DVD) drive).

[0137] The NVM / storage device 108 may include storage resources that are physically part of the device on which the data analysis-based tunnel ventilation fan control system 100 is installed, or it may be accessible to the device without being part of the device. For example, the NVM / storage device 108 may be accessed via (at least one) input / output device 110 over a network.

[0138] The (at least one) input / output device 110 may provide an interface for the data analysis-based tunnel ventilation fan control system 100 to communicate with any other appropriate device, and the input / output device 110 may include a communication component, a phonetic component, a sensor component, etc. The network interface 112 may provide an interface for the data analysis-based tunnel ventilation fan control system 100 to communicate according to at least one network, and the data analysis-based tunnel ventilation fan control system 100 may wirelessly communicate with at least one component of a wireless network according to any standard and / or protocol of at least one wireless network standard and / or protocol, such as accessing a wireless network according to a communication standard, or a combination thereof for wireless communication.

[0139] For one embodiment, at least one of the (at least one) processors 102 may be loaded together with the logic of at least one controller (e.g., a memory controller module) of the control module 104. For one embodiment, at least one of the (at least one) processors 102 may be loaded together with the logic of at least one controller of the control module 104 to form a system-level load. For one embodiment, at least one of the (at least one) processors 102 may be integrated on the same die with the logic of at least one controller of the control module 104. For one embodiment, at least one of the (at least one) processors 102 may be integrated on the same die with the logic of at least one controller of the control module 104 to form a system on chip (SoC).

[0140] In various embodiments, the tunnel ventilation fan control system 100 based on data analysis may be, but is not limited to, a terminal device such as a server, a desktop computing device or a mobile computing device (e.g., a laptop computing device, a handheld computing device, a tablet computer, a netbook, etc.). In various embodiments, the tunnel ventilation fan control system 100 based on data analysis may have more or fewer components and / or different architectures. For example, in an alternative embodiment, the tunnel ventilation fan control system 100 based on data analysis includes at least one camera, a keyboard, a liquid crystal display (LCD) screen (including a touch screen display), a non-volatile memory port, multiple antennas, a graphics chip, an application-specific integrated circuit (ASIC), and a speaker.

[0141] The embodiments of the present application are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A tunnel ventilation fan control method based on data analysis, characterized in that: Applied to a tunnel ventilation fan control system based on data analysis, the method comprises: Acquire a tunnel scene detection data combination and sample fan control strategy data of the tunnel scene detection data combination, wherein the tunnel scene detection data combination includes first tunnel scene detection data and second tunnel scene detection data; the first tunnel scene detection data and the second tunnel scene detection data are different data obtained by performing tunnel scene parameter detection on a tunnel section in a same tunnel ventilation task using different tunnel scene definition modes, and the sample fan control strategy data is used to mark the control parameter data of the fan control strategy of the ventilation fan of the tunnel section corresponding to the corresponding tunnel scene detection data; extracting tunnel scene parameter vectors of the first tunnel scene detection data and the second tunnel scene detection data respectively, and generating ventilation fan control data corresponding to the first tunnel scene detection data and the second tunnel scene detection data respectively based on the tunnel scene parameter vectors corresponding to the first tunnel scene detection data and the second tunnel scene detection data respectively; Obtain a first loss function value between ventilation fan control data of tunnel scene detection data of different tunnel scene definition modes and corresponding sample fan control strategy data, and obtain a second loss function value between different tunnel scene parameter vectors based on the tunnel scene parameter vectors of different tunnel scene detection data; The parameters of the ventilation fan control model are updated according to the training learning strategy of minimizing the first loss function value and the second loss function value; the ventilation fan control model after parameter update is used to make ventilation fan control strategy decisions for the tunnel section corresponding to the tunnel scene detection data under any tunnel scene definition mode.

2. The tunnel ventilation fan control method based on data analysis according to claim 1 is characterized in that: The tunnel scene detection data combination is any one of a plurality of tunnel scene detection data combinations covered by the sample tunnel scene detection data sequence, the tunnel scene parameter vector and the ventilation fan control data are both obtained by calling a ventilation fan control model with initialized weight parameters, wherein the model performance values ​​of the ventilation fan control model used to make ventilation fan control strategy decisions on the first tunnel scene detection data and the second tunnel scene detection data are different, if the model performance value of the ventilation fan control model used to make ventilation fan control strategy decisions on the first tunnel scene detection data is greater than the model performance value of the ventilation fan control model used to make ventilation fan control strategy decisions on the second tunnel scene detection data; The step of generating the ventilation fan control model with the initialization weight parameters comprises: Extracting at least one first tunnel scene detection data from a plurality of tunnel scene detection data combinations covered by the sample tunnel scene detection data sequence, and acquiring sample fan control strategy data corresponding to each extracted first tunnel scene detection data; Using the basic ventilation fan control model, the ventilation fan control strategy decision is made for each first tunnel scene detection data extracted; Based on the ventilation fan control data extracted from each first tunnel scene detection data and the corresponding sample fan control strategy data, knowledge learning is performed on the basic ventilation fan control model to generate a ventilation fan control model with initialized weight parameters.

3. The tunnel ventilation fan control method based on data analysis according to claim 2 is characterized in that: The method further comprises: If the sample fan control strategy data corresponding to any one of the extracted first tunnel scene detection data is an empty set, determining a target tunnel scene detection data combination corresponding to any one of the first tunnel scene detection data; The second tunnel scene detection data associated with any one of the first tunnel scene detection data is obtained from the target tunnel scene detection data combination, and the sample fan control strategy data obtained from the associated second tunnel scene detection data is used as the ventilation fan control data of any one of the first tunnel scene detection data.

4. The tunnel ventilation fan control method based on data analysis according to claim 2 is characterized in that: The ventilation fan control data for any one of the first tunnel scene detection data include: the fan control characteristics of the ventilation fan of the tunnel section in the any one of the first tunnel scene detection data, and the sample fan control strategy data for any one of the first tunnel scene detection data include: the actual control characteristics of the controlled ventilation fan of the tunnel section in the any one of the first tunnel scene detection data; The method of performing knowledge learning on the basic ventilation fan control model based on the ventilation fan control data extracted from each first tunnel scene detection data and the corresponding sample fan control strategy data includes: Calculating a feature comparison value based on the fan control feature and the actual control feature, the feature comparison value comprising a feature deviation calculated based on each fan control feature and the corresponding actual control feature; Based on the characteristic comparison value, the parameters of the basic ventilation fan control model are updated according to a back propagation strategy.

5. The tunnel ventilation fan control method based on data analysis according to claim 1 is characterized in that: The obtaining of the second loss function value between different tunnel scene parameter vectors includes: Based on the ventilation category prediction network, ventilation category prediction is performed on the tunnel scene parameter vector of the first tunnel scene detection data and the tunnel scene parameter vector of the second tunnel scene detection data respectively, to generate a ventilation control category of the first tunnel scene detection data and a ventilation control category of the second tunnel scene detection data; Based on the category loss function value between the ventilation control category of the first tunnel scene detection data and the ventilation control category of the second tunnel scene detection data, a second loss function value between corresponding different tunnel scene parameter vectors is determined.

6. The tunnel ventilation fan control method based on data analysis according to claim 5 is characterized in that: The ventilation category prediction network is obtained through sample knowledge learning; the method further includes: Acquire a sample tunnel scene detection data sequence, wherein the sample tunnel scene detection data sequence includes a plurality of tunnel scene detection data combinations, each of which includes a first tunnel scene detection data and a second tunnel scene detection data; Acquire tunnel scene parameter vectors of each tunnel scene detection data covered by the sample tunnel scene detection data sequence, and call a basic ventilation category prediction network to perform ventilation category prediction on the tunnel scene parameter vectors of each tunnel scene detection data covered by the sample tunnel scene detection data sequence, to generate a ventilation control category for each tunnel scene detection data in the sample tunnel scene detection data sequence; The loss function value is calculated based on the ventilation control category of each tunnel scene detection data in the sample tunnel scene detection data sequence and the corresponding prior ventilation control category, and based on the loss function value, the parameters of the basic ventilation category prediction network are updated according to the back propagation strategy.

7. The tunnel ventilation fan control method based on data analysis according to claim 1 is characterized in that: Obtaining a first loss function value between ventilation fan control data of tunnel scene detection data of different tunnel scene definition modes and corresponding sample fan control strategy data, including: Based on the ventilation fan control data corresponding to the first tunnel scene detection data and the sample fan control strategy data, obtaining a first loss function value between the first tunnel scene detection data and the sample fan control strategy data; Based on the ventilation fan control data corresponding to the second tunnel scene detection data and the sample fan control strategy data, a first loss function value between the second tunnel scene detection data and the sample fan control strategy data is obtained.

8. The tunnel ventilation fan control method based on data analysis according to claim 7 is characterized in that: The updating of the parameters of the ventilation fan control model according to the training learning strategy that minimizes the first loss function value and the second loss function value includes: Generate a target loss function value based on a first loss function value between the first tunnel scene detection data and the sample fan control strategy data, a first loss function value between the second tunnel scene detection data and the sample fan control strategy data, and a loss function value of the second loss function value; According to minimizing the objective loss function value, updating the parameters of the ventilation fan control model; If the model performance value of the ventilation fan control model for making ventilation fan control strategy decisions on the first tunnel scene detection data is greater than the model performance value of the ventilation fan control strategy decisions on the second tunnel scene detection data; The first loss function value based on the first tunnel scene detection data and the sample fan control strategy data, the first loss function value between the second tunnel scene detection data and the sample fan control strategy data, and the loss function value of the second loss function value and the generated target loss function value include: Respectively obtain a first influence coefficient corresponding to a first loss function value between the first tunnel scene detection data and the sample fan control strategy data, a second influence coefficient corresponding to a first loss function value between the second tunnel scene detection data and the sample fan control strategy data, and a third influence coefficient corresponding to the second loss function value, wherein the second influence coefficient is greater than the first influence coefficient, and the second influence coefficient is greater than the third influence coefficient; Based on the first influence coefficient, the first loss function value between the first tunnel scene detection data and the sample fan control strategy data is multiplied, based on the second influence coefficient, the first loss function value between the second tunnel scene detection data and the sample fan control strategy data is multiplied, and based on the third influence coefficient, the second loss function value is multiplied, the loss function values ​​after the product operation are summarized to generate a target loss function value.

9. The tunnel ventilation fan control method based on data analysis according to claim 1 is characterized in that: The tunnel scene detection data combination is any one of a plurality of tunnel scene detection data combinations covered by the sample tunnel scene detection data sequence; The updating of the parameters of the ventilation fan control model according to the training learning strategy that minimizes the first loss function value and the second loss function value includes: Based on each tunnel scene detection data combination in the sample tunnel scene detection data sequence, calculating a first loss function value and a second loss function value for each tunnel scene detection data combination; According to a training learning strategy for minimizing a first loss function value and a second loss function value corresponding to each combination of tunnel scene detection data, weight information of the ventilation fan control model is updated, and when the loss function value between tunnel scene parameter vectors corresponding to different tunnel scene detection data obtained by identifying the updated ventilation fan control model is less than a threshold value, it is determined that the parameter update of the ventilation fan control model is terminated; Among them, when the tunnel scene parameter vectors corresponding to the different tunnel scene detection data all conform to the same characteristic distribution, it is determined that the loss function value between the tunnel scene parameter vectors corresponding to the different tunnel scene detection data is less than a threshold value.

10. A tunnel ventilation fan control system based on data analysis, characterized in that: It comprises a processor and a computer-readable storage medium, wherein the computer-readable storage medium stores machine-executable instructions, and when the machine-executable instructions are executed by the processor, the tunnel ventilation fan control method based on data analysis according to any one of claims 1 to 9 is implemented.

Images