A freezing warning method for mountainous highways combined with terrain factors

By constructing the freezing feature data set and using the LSTM neural network, a short-term freezing warning model for mountain highways was established, which solved the refined problem of road freezing forecasts in mountainous areas, improved the early warning efficiency and accuracy, and achieved fine prediction of special road sections .

CN118095348BActive Publication Date: 2024-11-08GUIZHOU PROVINCIAL METEOROLOGICAL INFORMATION CENT (GUIZHOU PROVINCIAL METEOROLOGICAL ARCHIVES GUIZHOU PROVINCIAL METEOROLOGICAL STAFF EDUCATION & TRAINING CENT)
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410135062.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-01-31
Publication Date
2024-11-08
Estimated Expiration
2044-01-31

AI Technical Summary

Technical Problem

In mountainous areas, traditional road freezing forecasts are difficult to refine, especially the freezing conditions of special sections such as bridges, tunnels, and culverts are difficult to accurately predict, resulting in insufficient traffic safety warning efficiency and accuracy.

Method used

Construct a frozen feature data set, including experimental road data, highway meteorological data, meteorological site data, meteorological forecast data and geographical environment data, filter the data factors through principal component analysis method and entropy value method, and input them into the LSTM neural network for training. Establish a short-term freezing warning model for expressways to achieve fine prediction of freezing sections and time.

Benefits of technology

It improves the efficiency and accuracy of short-term freezing warnings on highways, can be fine to the road section and hourly level, provides multi-factor meteorological forecast information for the transportation department, and enhances road safety management.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118095348B_ABST
    Figure CN118095348B_ABST
Patent Text Reader

Abstract

The present invention provides a freezing warning method for mountain highways combined with terrain factors, including: constructing a freezing feature data set; the freezing feature data set includes experimental road data, highway meteorological data, meteorological station data, meteorological forecast data and geographical environment data; storing the freezing feature data set in a preset meteorological big data cloud platform column storage database, and establishing an associated view of the freezing feature data set to form a highway meteorological-geographic fusion data set; calculating and screening the highway meteorological-geographic fusion data set according to the principal component analysis method and the entropy value method to obtain data factors for model training; inputting the data factors into the LSTM neural network for training to obtain a trained highway short-term freezing warning model; inputting the data to be predicted into the highway short-term freezing warning model to obtain the predicted frozen road section pile number and freezing time. The present invention can improve the efficiency and accuracy of highway short-term freezing warning.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence and road safety monitoring, and in particular to a freezing early warning method for mountain highways combined with terrain factors. Background Art

[0002] There are two main types of road icing prediction algorithms. One is the model building method, which mainly uses the regression equation with road icing data as the output to predict the road icing state. The other is the data-driven icing prediction method, which mainly processes data and builds machine learning models to effectively predict the road icing state. The data-driven method has higher accuracy and can approximate more complex functions. The data mining method has the advantages of compatibility of data selection, accuracy of model construction, and convenience of parameter setting.

[0003] In view of the special geographical conditions in mountainous areas, a comprehensive short-term and refined road freezing warning system is needed to provide efficient and accurate road freezing safety warnings. Due to the special geographical conditions and geomorphological features of mountainous areas, the latitude and longitude, altitude and terrain changes have a very direct impact on road temperature, which will cause segmented road freezing. At the same time, the road freezing conditions of sections where road structures such as bridges, tunnels, culverts and retaining walls are located are even more difficult to control. Even if only meteorological conditions are considered, it is difficult to achieve relevant forecasts of road freezing conditions in these special sections. Therefore, in mountainous areas, it is necessary to develop short-term and refined road freezing forecast products, combine geographic environment information data, and use data-driven machine learning models to build a road freezing warning model to solve the problem of excessive granularity of traditional meteorological forecasts, and provide transportation departments with multi-factor hourly meteorological forecast information that is refined to road sections and pile numbers. Summary of the invention

[0004] In order to overcome the deficiencies of the prior art, the present invention aims to provide a freezing early warning method for mountainous highways in combination with terrain factors.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A freezing early warning method for mountainous highways combined with terrain factors, comprising:

[0007] Constructing a freezing characteristic data set; the freezing characteristic data set includes experimental road data, highway meteorological data, meteorological station data, meteorological forecast data and geographical environment data;

[0008] The freezing characteristic data set is stored in a preset meteorological big data cloud platform column storage database, and an associated view of the freezing characteristic data set is established to form a highway meteorological-geographic fusion data set;

[0009] The highway meteorological-geographic fusion data set is calculated and screened according to the principal component analysis method and the entropy value method to obtain data factors for model training;

[0010] The data factors are input into the LSTM neural network for training to obtain a trained highway short-term freezing warning model;

[0011] The data to be predicted is input into the short-term freezing warning model of the expressway to obtain the predicted frozen road section pile number and freezing time, and a warning prompt is given according to the frozen road section pile number and freezing time.

[0012] Preferably, the constructing of the freezing characteristic data set comprises:

[0013] Determine the target experimental section;

[0014] Extracting the experimental road data of the target experimental section from the highway freezing event log; the experimental road data includes: geographical location information of the target experimental section and a road friction coefficient measured for the target experimental section;

[0015] Obtaining the highway meteorological data monitored by the meteorological traffic stations around the target experimental road section; the highway meteorological data includes: road surface temperature, meteorological temperature, pressure, humidity and wind elements, road surface state and corresponding road surface friction coefficient;

[0016] Acquire the meteorological station data monitored by the meteorological stations around the target experimental section; the meteorological station data includes wind direction vector, wind speed, air pressure, ambient temperature, relative humidity, rainfall and snow depth;

[0017] Obtaining the weather forecast data using a grid forecast product;

[0018] The geographical environment data is constructed according to the target experimental road section; the geographical environment data includes: road section, stake number, longitude and latitude, altitude and whether it contains special road structure.

[0019] Preferably, the data factors include: ambient temperature, road surface temperature, rainfall, snow depth, wind speed, altitude, relative humidity, 1-hour forecast temperature and 1-hour forecast rainfall.

[0020] Preferably, after the data factors are input into the LSTM neural network for training to obtain a trained highway short-term freezing warning model, the method further includes:

[0021] The trained highway short-term freezing warning model is verified by using the root mean square error to obtain a verified highway short-term freezing warning model.

[0022] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0023] The present invention provides a freezing early warning method for mountain highways combined with terrain factors, including: constructing a freezing feature data set; the freezing feature data set includes experimental road data, highway meteorological data, meteorological station data, meteorological forecast data and geographical environment data; storing the freezing feature data set in a preset meteorological big data cloud platform column storage database, and establishing an associated view of the freezing feature data set to form a highway meteorological-geographic fusion data set; calculating and screening the highway meteorological-geographic fusion data set according to the principal component analysis method and the entropy value method to obtain data factors for model training; inputting the data factors into the LSTM neural network for training to obtain a trained highway short-term freezing early warning model; inputting the data to be predicted into the highway short-term freezing early warning model to obtain the predicted frozen road section pile number and freezing time. The present invention can improve the efficiency and accuracy of highway short-term freezing early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0025] Figure 1 A flow chart of a method provided by an embodiment of the present invention;

[0026] Figure 2 A schematic diagram of a technical route provided for an embodiment of the present invention;

[0027] Figure 3 A schematic diagram of the LSTM algorithm structure provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] The purpose of the present invention is to provide a freezing early warning method for mountainous highways in combination with terrain factors, which can improve the efficiency and accuracy of short-term freezing early warning for highways.

[0030] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] Figure 1 A flow chart of a method provided by an embodiment of the present invention, such as Figure 1 As shown, the present invention provides a freezing early warning method for mountainous highways combined with terrain factors, comprising:

[0032] Step 100: constructing a freezing characteristic data set; the freezing characteristic data set includes experimental road data, highway meteorological data, meteorological station data, meteorological forecast data and geographical environment data;

[0033] Step 200: storing the freezing characteristic data set in a preset meteorological big data cloud platform column storage database, and establishing an associated view of the freezing characteristic data set to form a highway meteorological-geographic fusion data set;

[0034] Step 300: Calculate and screen the highway meteorological-geographic fusion data set according to the principal component analysis method and the entropy value method to obtain data factors for model training;

[0035] Step 400: inputting the data factors into the LSTM neural network for training to obtain a trained highway short-term freezing warning model;

[0036] Step 500: input the data to be predicted into the highway short-term freezing warning model to obtain the predicted frozen road section pile number and freezing time, and issue a warning prompt based on the frozen road section pile number and freezing time.

[0037] like Figure 2 As shown, the technical route in this embodiment is as follows:

[0038] 1. Analyze and extract road sections prone to freezing.

[0039] The present invention needs to extract the location information of the prone to freezing sections (i.e., experimental sections) from the highway freezing event ledger from December 2020 to March 2023, including the section, pile number, longitude and latitude, altitude, and whether it contains special road structures, and correspond to the measured data of traffic meteorological stations and surrounding meteorological stations under simultaneous and spatial distribution, so as to facilitate the analysis of the impact of geographic environment information data on road surface temperature in the experiment.

[0040] 2. Linear change of road friction coefficient.

[0041] The adhesion characteristics between ice and different attachments can vary greatly. Considering that the asphalt material of the road surface has a certain impact on road freezing, it is necessary to measure the linear change of the friction coefficient of the road surface under freezing weather on the experimental road section. The change in friction coefficient is only related to the road surface state (dry, damp, wet, ice, snow, etc.) and is not affected by other factors. Therefore, the friction coefficient can further quantify the road surface state into a numerical form. Therefore, this project uses the road friction coefficient to test and score the early warning model, and further optimize the model algorithm and threshold adjustment parameters.

[0042] 3. Form a fusion data set of experimental road data, highway meteorological data, meteorological station data, meteorological forecast data and geographic environment data.

[0043] The freezing feature dataset consists of experimental road data, highway meteorological data, meteorological station data, meteorological forecast data, and geographic environment data. The experimental road data comes from the geographical location information of the experimental roads selected in the freezing ledger, as well as the measured road friction coefficient. The highway meteorological data comes from the meteorological traffic station of the Department of Transportation, the road surface temperature, meteorological temperature, pressure, humidity and wind elements, road surface state and corresponding road friction coefficient detected by the roadside equipment, and the meteorological station data comes from the meteorological stations around the expressway, including six meteorological elements: wind direction vector, wind speed, air pressure, ambient temperature, relative humidity, rainfall, and snow depth. In order to ensure that the data information is rich enough, 5 minutes is used as the benchmark for data collection and the target time interval for warning. Since the high-altitude temperature and near-ground temperature have a certain influence on road freezing, the ground temperature of the meteorological station, the temperature at 100m, 300m, and 500m above the ground, and the ground temperature at 5cm, 10cm, and 20cm underground are added. In addition, considering the influence of water vapor in the air, the relative humidity at 100m, 300m, and 500m above the meteorological station and the precipitable water data of GPS / MET water vapor data are added. The meteorological forecast data uses grid forecast products (elements include but are not limited to: temperature, air pressure, humidity, wind, rainfall), and the forecast time is within 6 hours. The road section, stake number, longitude and latitude, altitude, slope, slope direction, and whether there are special road structures of the constructed experimental section form the geographical environment data.

[0044] According to the designed database fields, the corresponding database tables are created in the column storage database Gbase of the meteorological big data cloud platform, and the experimental road data (T1), geographic information data (T2), highway meteorological data (T3), meteorological element data (T4), and meteorological forecast data (T5) of the screened experimental sections are stored, and associated views are established to form a highway meteorological-geographic fusion data set (T6).

[0045] 4. Screening of main data factors

[0046] The principal component analysis (PCA) and entropy method are used to calculate the weights of different data factors, and finally the data factors with the greatest influence are selected for algorithm training.

[0047] Determine the principal components F i About the original variable X j The expression of coefficient a ij .

[0048] Determine the principal components F i About the original variable X j The expression of coefficient a ij .

[0049]

[0050] The amount of information extracted by each principal component can be measured by its variance. For example, the larger the variance of F1, the more information F1 contains. It is often hoped that the first principal component F1 contains the largest amount of information. Therefore, the F1 selected from all linear combinations should be X1, X2, ..., X p F1 has the largest variance among all linear combinations, so it is called the first principal component.

[0051] The main data factors include: ambient temperature, road surface temperature, rainfall, snow depth, wind speed, altitude, relative humidity, 1-hour forecast temperature, and 1-hour forecast rainfall.

[0052] 5. Establish a short-term and refined highway freezing warning model based on machine learning

[0053] Based on the main data factors and taking advantage of the higher precision of data-driven methods, the experimental road data, highway meteorological data, meteorological station data, meteorological forecast data and geographical environment data are used as input parameters to establish an LSTM neural network highway short-term freezing warning model, and the prediction results are refined to the hourly level and specific road section pile numbers.

[0054] LSTM neural network. The LSTM algorithm structure is as follows Figure 3 As shown, the algorithm realizes the long and short-term memory of time series through cell states and hidden states, and inputs data in the form of sliding windows to complete each set of calculations.

[0055] Assume that in any set of data, the states of the two at the kth moment are c k 、h k , the calculation process of δ is as follows:

[0056] c k =f(h k-1 ,c k-1 ,a k )

[0057] =δ(W f ⊙[a k ,h k-1 ])⊙c k-1 +δ(W i ⊙[a k ,h k-1 ])⊙tan h(W⊙[a k ,h k-1 ])

[0058] h k =f(h k-1 ,c k ,a k )

[0059] =δ(W o ⊙[a k ,h k-1 ])⊙tan h(c k )

[0060] Among them, a k W represents the input vector, and tanh represents the sigmoid and activation functions respectively. f , W i , W o is the weight matrix of each gate control unit. Through iterative calculation, the implicit state vector hend obtained by this group of training can be obtained, and then the final prediction result can be obtained through the weight coefficient and activation function of the fully connected layer, that is:

[0061]

[0062] in, represents the predicted value, W D Represents the weight vector of the fully connected layer. a Represents the activation function.

[0063] 6. Use root mean square error to evaluate prediction results

[0064] The root mean square error (RMSE) is used to evaluate the prediction effect. The calculation method is as follows:

[0065]

[0066] Among them, m v Indicates the number of samples involved in the verification; Y i Represents the predicted value of road friction coefficient; y i Indicates the actual road friction coefficient.

[0067] The present invention uses interdisciplinary and interdepartmental data fusion of meteorology and transportation to identify significant relevant data factors of road freezing caused by complex weather changes and highway topography changes on mountain highways. A short-term refined road freezing warning model is established through machine learning methods, and the accuracy of the model is verified to provide a scientific basis for summarizing the laws of road freezing prediction. The effect of the short-term refined road freezing warning model is tested and evaluated, and the optimal model is selected to design and implement the short-term refined road freezing warning model system.

[0068] The beneficial effects of the present invention are as follows:

[0069] (1) The present invention takes the interdisciplinary research in the field of meteorology and transportation as the starting point, actively explores new ways for the meteorological department to deeply integrate into local economic development and protect lives, and expands ideas for the transportation department to understand the occurrence laws of highway freezing and establish a scientific prevention and response system.

[0070] (2) The present invention promotes the application of machine learning models in the field of meteorological big data and improves the effective allocation of traffic management resources.

[0071] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0072] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A freezing warning method for mountain highways combined with terrain factors, It is characterized in that include: Constructing a freezing characteristic data set; the freezing characteristic data set includes experimental road data, highway meteorological data, meteorological station data, meteorological forecast data and geographical environment data; The freezing characteristic data set is stored in a preset meteorological big data cloud platform column storage database, and an associated view of the freezing characteristic data set is established to form a highway meteorological-geographic fusion data set; The highway meteorological-geographic fusion data set is calculated and screened according to the principal component analysis method and the entropy value method to obtain data factors for model training; The data factors are input into the LSTM neural network for training to obtain a trained highway short-term freezing warning model; Input the data to be predicted into the highway short-term freezing warning model to obtain the predicted frozen road section pile number and freezing time, and issue a warning prompt based on the frozen road section pile number and freezing time; The step of constructing a freezing feature data set comprises: Determine the target experimental section; Extracting the experimental road data of the target experimental section from the highway freezing event log; the experimental road data includes: geographical location information of the target experimental section and a road friction coefficient measured for the target experimental section; Obtaining the highway meteorological data monitored by the meteorological traffic stations around the target experimental road section; the highway meteorological data includes: road surface temperature, meteorological temperature, pressure, humidity and wind elements, road surface state and corresponding road surface friction coefficient; Acquire the meteorological station data monitored by the meteorological stations around the target experimental section; the meteorological station data includes wind direction vector, wind speed, air pressure, ambient temperature, relative humidity, rainfall and snow depth; Obtaining the weather forecast data using a grid forecast product; The geographical environment data is constructed according to the target experimental road section; the geographical environment data includes: road section, stake number, longitude and latitude, altitude and whether it contains special road structure.

2. The freezing early warning method for mountainous highways combined with terrain factors according to claim 1, It is characterized in that The data factors include: ambient temperature, road surface temperature, rainfall, snow depth, wind speed, altitude, relative humidity, 1-hour forecast temperature and 1-hour forecast rainfall.

3. The freezing early warning method for mountainous highways combined with terrain factors according to claim 1, It is characterized in that After the data factors are input into the LSTM neural network for training to obtain a trained highway short-term freezing warning model, the method further includes: The trained highway short-term freezing warning model is verified by using the root mean square error to obtain a verified highway short-term freezing warning model.