ETC portal frame data-based expressway section differentiated charging pricing optimization method

Abstract

The invention relates to the technical field of pricing management, and discloses an ETC portal frame data-based expressway section differentiated charging pricing optimization method. Comprising the steps of combining a comprehensive traffic parameter with a preset charge value, generating a first comprehensive optimization parameter, predicting a road section selection probability, combining a differentiation parameter with the road section selection probability, generating a second comprehensive optimization parameter, predicting the total income of a road section, and making a charge pricing optimization strategy after charge pricing prompt information is generated; compared with the prior art, the method can accurately predict the probability that the vehicle selects the selected road section as the passing road section, can rapidly and accurately predict the total income of the road section when the vehicle passes on the basis of taking the selected road section as the passing road section, and specifically formulates specific strategies and measures for road section charging pricing optimization. The dynamic balance effect between the road section selection probability and the total income of the road section is realized, and the charging pricing optimization effect is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of pricing management, and more specifically, to a method for optimizing differentiated toll pricing for highway sections based on ETC gantry data. Background Art

[0002] Expressways are an important symbol of highway modernization. Since expressways cost a lot of money to build and require a lot of money for daily maintenance, vehicles need to pay certain tolls when passing through expressways. In order to improve the accuracy of expressway toll pricing, it is necessary to optimize and adjust the expressway toll pricing, so as to achieve an effective balance between expressway toll pricing and total revenue.

[0003] The patent application with reference publication number CN1 15800845A discloses a method and system for differentiated pricing of highways, including obtaining the traffic volume of each vehicle type between the starting and ending points of the highway; classifying each vehicle type obtained according to the travel purpose, and establishing travel benefit models for the classified vehicles respectively; using a multi-logit model to describe the probability of different vehicle types choosing different paths in different time periods; based on the travel benefit model and the probability of different vehicle types choosing different paths in different time periods, constructing a pricing model with the goal of maximizing highway revenue; based on the maximization of the revenue of the pricing model, obtaining the adjusted highway price; by constructing travel benefit functions for different vehicle types, based on the travel benefit function and the travel mode Logit selection model, a highway travel path selection model is established, which is used to predict the proportion of a certain vehicle type using highway travel according to different highway toll prices, and after obtaining the traffic volume data of each vehicle type between the starting and ending points, a differentiated pricing model that maximizes the toll revenue of highway investors can be constructed, and flexible differentiated toll standards can be formulated;

[0004] The existing technology has the following deficiencies: the existing differentiated pricing method for highways takes the total revenue of highway sections as the ultimate goal, and optimizes and adjusts the toll pricing according to the laws of market economic development, which limits the dimensions of factors considered when optimizing and adjusting the toll pricing of highway sections, and makes it impossible to integrate and analyze the multi-dimensional ETC gantry data on the highway sections. As a result, when optimizing and adjusting the toll pricing, it is easy for the toll pricing to be too high, resulting in an imbalance between the selection probability of the highway section and the total revenue, thereby reducing the total revenue of the highway.

[0005] In view of this, the present invention proposes a highway section differentiated toll pricing optimization method based on ETC gantry data to solve the above problems. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a method for optimizing differentiated toll pricing for highway sections based on ETC gantry data, comprising:

[0007] S1: extracting comprehensive traffic parameters from ETC gantry data, combining the comprehensive traffic parameters with preset charging values, and generating first comprehensive optimization parameters, wherein the comprehensive traffic parameters include vehicle type value, peak time value, travel time value, and road section congestion rate;

[0008] S2: inputting the first comprehensive optimization parameter into a first machine learning model trained in advance for predicting the probability of road section selection, and predicting the probability of road section selection;

[0009] S3: obtaining differentiated parameters of the selected road section, and combining the differentiated parameters with the road section selection probability to generate a second comprehensive optimization parameter, wherein the differentiated parameters include a passing vehicle value, a passing distance value, and a passing timeout value;

[0010] S4: inputting the second comprehensive optimization parameter into a second machine learning model trained in advance for predicting the total revenue of the road section to predict the total revenue of the road section;

[0011] S5: Compare the predicted total revenue of the road section with the preset total revenue threshold, generate charging pricing prompt information, the charging pricing prompt information includes insufficient charging pricing prompt information and sufficient charging pricing prompt information, and formulate a charging pricing optimization strategy, the charging pricing optimization strategy includes maintaining the preset charging value strategy and reducing the preset charging value strategy.

[0012] Further, the vehicle type values ​​include a class 1 vehicle, a class 2 vehicle, a class 3 vehicle, a class 4 vehicle, and a class 5 vehicle;

[0013] The methods for extracting peak hour values ​​include:

[0014] The driving time data of i vehicles on the selected road section in the ETC gantry data are queried one by one through the timestamp, and all the entry and exit times in the driving time data are identified through natural language processing technology to obtain i entry times and i exit times;

[0015] The period between the i entry time and the i exit time is recorded as the driving period, and i driving periods are obtained;

[0016] Compare one driving period with the standard peak period in turn, and record the period when one driving period overlaps with the standard peak period as a valid period, to obtain one valid period;

[0017] Compare the duration of a valid period with the duration of a driving period in sequence to obtain a sub-peak value;

[0018] The expression of the sub-peak value is:

[0019]

[0020] In the formula, GF zi is the peak value of the ith sub-peak, SC yxi is the length of the i-th valid period, SC xsi is the duration of the first driving period;

[0021] After removing the maximum and minimum sub-peak values, the remaining i-2 sub-peak values ​​are accumulated and averaged to obtain the peak period value;

[0022] The expression for the peak hour value is:

[0023]

[0024] In the formula, GF sd is the peak time value, GF za is the ath sub-peak value.

[0025] Furthermore, the method for extracting the travel time value includes:

[0026] The time when a vehicle first and last generated travel time data in the ETC gantry data is queried one by one through the timestamp, and a start time and an end time are obtained respectively;

[0027] The duration from a start time to an end time is recorded as the total duration, and a total duration is obtained;

[0028] The length of the selected section and the total length corresponding to the entire journey duration are measured by the Beidou positioning system, and the length of the selected section is compared with the total length corresponding to the entire journey duration to obtain the length ratio value;

[0029] The expression of length ratio is:

[0030]

[0031] In the formula, CD zb is the length ratio, CD xd is the length of the selected section, CD zc is the total length corresponding to the whole journey;

[0032] Multiply the length ratio by the i-th total duration to obtain i-th product values, and then add up the i-th product values ​​and calculate the average to obtain the travel duration value;

[0033] The expression of the travel time value is:

[0034]

[0035] In the formula, TX sc is the travel time value, SC qcb is the duration of the bth full journey.

[0036] Furthermore, the method for extracting the road section congestion rate includes:

[0037] Randomly mark two ETC gantries at adjacent positions on the selected road section and record them as the first gantry and the second gantry respectively;

[0038] Identify the time when a vehicle passes through the first gantry and the second gantry from the ETC gantry data, and obtain i first time points and 1 second time point;

[0039] The duration between a first moment and a second moment is recorded as the elapsed duration, and an elapsed duration is obtained;

[0040] The distance between the first gantry and the second gantry is measured by the Beidou positioning system to obtain the passing distance, and the passing time is compared with the passing distance in turn to obtain a passing speed;

[0041] The expression of the passing speed is:

[0042]

[0043] In the formula, SD jgi is the i-th passing speed, JL jg is the distance travelled, SC jgi is the i-th elapsed time;

[0044] The passing speed that is less than the standard speed threshold is recorded as the congestion speed, and the number of congestion speeds is counted. After comparing the number of congestion speeds with the number of passing speeds, the congestion rate of the road section is obtained;

[0045] The expression of road section congestion rate is:

[0046]

[0047] In the formula, YD ld is the road congestion rate, SL yd is the number of congestion speeds.

[0048] Furthermore, the training method of the first machine learning model includes:

[0049] Convert the vehicle type value to 1 for a type 1 vehicle, 2 for a type 2 vehicle, 3 for a type 3 vehicle, 4 for a type 4 vehicle, and 5 for a type 5 vehicle.

[0050] The first comprehensive optimization parameter is converted into a corresponding set of first eigenvectors, the first eigenvectors are used as the input of the first machine learning model, and the section selection probability corresponding to each set of first comprehensive optimization parameters is used as the output of the first machine learning model. The section selection probability is used as the prediction target, and minimizing the sum of prediction errors of all training data is used as the training target. The first machine learning model is trained until the sum of prediction errors converges, and the training is stopped to obtain the first machine learning model that predicts the section selection probability.

[0051] Furthermore, the method for obtaining the passing vehicle value includes:

[0052] Query the attribute data of p vehicles passing through the selected road section within a preset vehicle statistical period from the ETC gantry data to obtain the attribute data of p vehicles;

[0053] Through natural language processing technology, semantic recognition is performed on p vehicle attribute data one by one to obtain p key semantics;

[0054] Passenger and cargo category 1, passenger and cargo category 2, passenger and cargo category 3, passenger and cargo category 4 and passenger and cargo category 5 are recorded as standard semantics, and the key semantics are matched with the standard semantics one by one;

[0055] The vehicles corresponding to the vehicle attribute data with the key semantics of passenger and freight category 1, passenger and freight category 2, passenger and freight category 3, passenger and freight category 4, and passenger and freight category 5 are recorded as category 1 vehicles, category 2 vehicles, category 3 vehicles, category 4 vehicles, and category 5 vehicles, respectively, and the number of category 1 vehicles, category 2 vehicles, category 3 vehicles, category 4 vehicles, and category 5 vehicles are counted one by one;

[0056] The number of vehicles of category 1, category 2, category 3, category 4 and category 5 are assigned corresponding proportional coefficients respectively and added to obtain the passing vehicle value;

[0057] The expression of the passing vehicle value is:

[0058] TX cl =σ1*SL 1l +σ2*SL 2l +σ3*SL 3l +σ4*SL4 l +σ5*SL 5l ;

[0059] In the formula, TX cl is the value of passing vehicles, SL 1l is the number of vehicles of type -, SL 2l is the number of second-class vehicles, SL 3l is the number of three types of vehicles, SL 4l is the number of four types of vehicles, SL 5lis the number of five types of vehicles, σ1, σ2, σ3, σ4, and σ5 are proportional coefficients greater than 0.

[0060] Furthermore, the method for obtaining the travel distance value includes:

[0061] The positions of p vehicles when they first enter the selected road section are queried through the ETC gantry data to obtain p initial positions;

[0062] The real-time positions of p vehicles are monitored by the Beidou positioning system. When the real-time position of a vehicle does not coincide with the position of the selected road section, the real-time position of the vehicle is recorded as an abnormal position, and p abnormal positions are obtained;

[0063] The distances between p initial positions and p abnormal positions are measured one by one by the Beidou positioning system to obtain p sub-distance values, and the passing distance value is obtained by accumulating the p sub-distance values ​​one by one;

[0064] The expression of the travel distance value is:

[0065]

[0066] In the formula, TX jl is the travel distance value, JL zc is the cth sub-distance value.

[0067] Furthermore, the method for obtaining the pass timeout value includes:

[0068] The driving time data of p vehicles on the selected road section in the ETC gantry data are queried one by one through the timestamp, and the entry time and exit time in the driving time data are identified through natural language processing technology to obtain p entry times and p exit times;

[0069] The time between the p entry moments and the p exit moments is recorded as the travel time, and p travel times are obtained;

[0070] The travel time longer than the standard travel time is recorded as the target time, and s target time lengths are obtained;

[0071] Subtract the s target durations from the standard travel durations in turn to obtain s sub-timeout values, and then add up the s sub-timeout values ​​one by one to obtain the travel timeout value;

[0072] The expression for the pass timeout value is:

[0073]

[0074] In the formula, TX cs is the timeout value, SC mbd is the duration of the dth target, SC bzThis is the standard travel time.

[0075] Furthermore, the training method of the second machine learning model includes:

[0076] The second comprehensive optimization parameters are converted into a corresponding set of second eigenvectors, the second eigenvectors are used as the input of the second machine learning model, and the total revenue of the road section corresponding to each set of second comprehensive optimization parameters is used as the output of the second machine learning model. The total revenue of the road section is used as the prediction target, and minimizing the sum of prediction errors of all training data is used as the training target. The second machine learning model is trained until the sum of prediction errors converges, and the training is stopped to obtain a second machine learning model that predicts the total revenue of the road section.

[0077] Furthermore, the method for generating the charging pricing insufficient prompt information and the charging pricing sufficient prompt information includes:

[0078] The predicted total revenue SR of the road segment yc With the preset total income threshold SR ys Compare;

[0079] When SR yc Greater than or equal to SR ys When generating charging pricing sufficient prompt information;

[0080] When SR yc Less than SR ys When the charging price is insufficient, a prompt message is generated;

[0081] The method for formulating the strategy of maintaining the preset charging value and the strategy of reducing the preset charging value includes:

[0082] When generating sufficient reminder information for charging pricing, formulate a strategy to maintain the preset charging value;

[0083] When a reminder message of insufficient charging pricing is generated, a strategy for reducing the preset charging value is formulated.

[0084] Technical effects and advantages of the highway section differentiated toll pricing optimization method based on ETC gantry data of the present invention:

[0085] The present invention extracts comprehensive traffic parameters from ETC gantry data, combines the comprehensive traffic parameters with preset charging values, generates a first comprehensive optimization parameter, inputs the first comprehensive optimization parameter into a first machine learning model trained in advance for predicting the probability of section selection, predicts the section selection probability, obtains the differentiated parameters of the selected section, and generates a second comprehensive optimization parameter after combining the differentiated parameters with the section selection probability, inputs the second comprehensive optimization parameter into a second machine learning model trained in advance for predicting the total revenue of the section, predicts the total revenue of the section, compares the predicted total revenue of the section with a preset total revenue threshold, generates charging pricing prompt information, and formulates a charging pricing optimization strategy; compared with the prior art, by generating the first comprehensive optimization parameter and combining the first The machine learning model can accurately predict the probability that a vehicle will select a selected road section as a passable road section, thereby accurately predicting the probability of selecting a highway section. At the same time, combined with the second comprehensive optimization parameter and the second machine learning model, it can quickly and accurately predict the total revenue of the road section when the vehicle passes, based on the selected road section as the passable road section and combined with the differentiated influencing parameters of different road sections. According to the comparative analysis results of the total revenue of the road section, specific strategies and measures for optimizing the road section toll pricing are formulated in a targeted manner, so that the highway section can achieve a dynamic balance between the probability of section selection and the total revenue of the road section based on the differentiated influencing parameters, thereby ensuring that the highway section can achieve the purpose of differentiated toll pricing optimization, greatly improving the total revenue of the highway. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 A flow chart of a method for optimizing differentiated toll pricing for highway sections based on ETC gantry data provided in the first embodiment of the present invention;

[0087] Figure 2 A schematic diagram of a highway section differentiated toll pricing optimization system based on ETC gantry data provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0088] 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.

[0089] Example 1: Please refer to Figure 1 As shown, the highway section differentiated toll pricing optimization method based on ETC gantry data described in this embodiment includes:

[0090] S1: extracting comprehensive traffic parameters from ETC gantry data, combining the comprehensive traffic parameters with preset charging values, and generating first comprehensive optimization parameters;

[0091] ETC gantry data refers to the data collected by the ETC equipment on the gantry installed on the selected road section to represent the vehicle and road information on the highway, and serves as the basic data for subsequent toll pricing optimization. Since ETC gantry data represents many data dimensions and different dimensions represent different meanings, in order to accurately identify ETC gantry data, it is necessary to extract the required comprehensive traffic parameters from the integrated ETC gantry data;

[0092] Comprehensive traffic parameters include vehicle type value, peak hour value, travel time value and road section congestion rate;

[0093] The vehicle type value refers to the data in the ETC gantry data that can represent the types of different vehicles on the selected road section, thereby effectively distinguishing between different types of vehicles; the vehicle type value includes Class I vehicles, Class II vehicles, Class III vehicles, Class IV vehicles and Class V vehicles; the vehicle type value is obtained by identifying the vehicle license plate information in the ETC gantry data and matching it with the pre-stored data in the vehicle management platform;

[0094] The peak time value refers to the ratio of the time period when the vehicle passes through the selected road section to the corresponding peak time in the ETC gantry data, thereby indicating the ratio of the peak time period when the vehicle is traveling on the selected road section. The larger the peak time value, the larger the ratio of the time period when the vehicle passes through the selected road section to the corresponding peak time.

[0095] The methods for extracting peak hour values ​​include:

[0096] The driving time data of i vehicles on the selected road section in the ETC gantry data are queried one by one through the timestamp, and all the entry and exit times in the driving time data are identified through natural language processing technology to obtain i entry times and i exit times;

[0097] The period between 1 entry time and i exit times is recorded as the driving period, and i driving periods are obtained;

[0098] Compare the i driving periods with the standard peak periods in turn, and record the period in which a driving period overlaps with the standard peak period as a valid period, to obtain a valid period; the standard peak period refers to the time period of traffic that is identified as having extremely high traffic volume on the selected road section, thereby providing a comparison basis for whether the subsequent driving period is in a situation of extremely high traffic volume; the standard peak period is set according to the specific numerical value of the traffic volume on the historically selected road section;

[0099] Compare the duration of a valid period with the duration of a driving period in sequence to obtain a sub-peak value;

[0100] The expression of the sub-peak value is:

[0101]

[0102] In the formula, GF zi is the peak value of the ith sub-peak, SC yxi is the length of the i-th valid period, SC xsi is the duration of the i-th driving period;

[0103] After removing the maximum and minimum sub-peak values, the remaining i-2 sub-peak values ​​are accumulated and averaged to obtain the peak period value;

[0104] The expression for the peak hour value is:

[0105]

[0106] In the formula, GF sd is the peak time value, GF za is the ath sub-peak value.

[0107] The travel time value refers to the actual travel time taken by the vehicle when passing through the selected road section in the ETC gantry data, which can accurately represent the travel time of the vehicle on the selected road section. The larger the travel time value, the longer the actual travel time taken by the vehicle when passing through the selected road section.

[0108] Methods for extracting the travel time value include:

[0109] The time when a vehicle first and last generated travel time data in the ETC gantry data is queried one by one through the timestamp, and a start time and an end time are obtained respectively;

[0110] The duration from a start time to an end time is recorded as the total duration, and a total duration is obtained;

[0111] The length of the selected section and the total length corresponding to the entire journey duration are measured by the Beidou positioning system, and the length of the selected section is compared with the total length corresponding to the entire journey duration to obtain the length ratio value;

[0112] The expression of length ratio is:

[0113]

[0114] In the formula, CD zb is the length ratio, CD xd is the length of the selected section, CD zcis the total length corresponding to the whole journey;

[0115] The length ratio value is multiplied by the total journey time to obtain l product values, and the product values ​​are accumulated and averaged to obtain the travel time value;

[0116] The expression of the travel time value is:

[0117]

[0118] In the formula, TX sc is the travel time value, SC qcb is the duration of the bth full journey.

[0119] The road section congestion rate refers to the degree of congestion on the selected road section caused by vehicles passing through the selected road section in the ETC gantry data, which can be used to indicate the degree of traffic congestion on the selected route. The greater the road section congestion rate, the more serious the congestion on the selected road section caused by vehicles passing through the selected road section.

[0120] The methods for extracting the road section congestion rate include:

[0121] Randomly mark two ETC gantries at adjacent positions on the selected road section and record them as the first gantry and the second gantry respectively;

[0122] Identify the time when a vehicle passes through the first gantry and the second gantry from the ETC gantry data, and obtain l first time and l second time;

[0123] The duration between the i first moments and the i second moments is recorded as the elapsed duration, and 1 elapsed duration is obtained;

[0124] The distance between the first gantry and the second gantry is measured by the Beidou positioning system to obtain the passing distance, and the passing time is compared with the passing distance in turn to obtain a passing speed;

[0125] The expression of the passing speed is:

[0126]

[0127] In the formula, SD jgi is the i-th passing speed, JL jg is the distance travelled, SC jgi is the i-th elapsed time;

[0128] The passing speed less than the standard speed threshold is recorded as the congestion speed, and the number of congestion speeds is counted. After comparing the number of congestion speeds with the number of passing speeds, the congestion rate of the road section is obtained; the standard speed threshold refers to the minimum value of the passing speed in a non-congested state on the selected road section, which can limit the minimum value of the congestion speed, thereby realizing accurate identification of the passing speed; the standard speed threshold is obtained by collecting a large number of minimum values ​​of the passing speeds corresponding to the congestion speeds identified in history and calculating their average value;

[0129] The expression of road section congestion rate is:

[0130]

[0131] In the formula, YD ld is the road congestion rate, SL yd is the number of congestion speeds.

[0132] The preset toll value refers to the unit price of the highway for different sections and time periods set by the highway management department, and it affects the section selection probability of the selected section, so that preset toll values ​​of different sizes correspond to different section selection probabilities. By analyzing the proportional relationship between a large number of preset toll values ​​and section selection probabilities in the past, it can be seen that when the preset toll value is larger, the vehicle spends more money on the selected section. At this time, the probability that the vehicle chooses the selected section as the passing section is lower, and the total number of vehicles on the selected section is smaller, and the corresponding total revenue of the selected section will be reduced, and vice versa;

[0133] After obtaining the comprehensive traffic parameters, the comprehensive traffic parameters can be combined with the preset charging values, so that the vehicle type value, peak hour value, travel time value, road section congestion rate and preset charging value are combined together, thereby forming the first comprehensive optimization parameter for predicting the probability of road section selection.

[0134] S2: inputting the first comprehensive optimization parameter into a first machine learning model trained in advance for predicting the probability of road section selection, and predicting the probability of road section selection;

[0135] The first machine learning model is a prediction model for predicting the probability of the selected road section being selected as the highway for vehicle travel according to the first comprehensive optimization parameter, that is, the probability of whether the vehicle enters the selected road section for travel can be predicted, and it can also provide indirect data support for the subsequent toll pricing of the highway section;

[0136] The road section selection probability refers to the probability that a vehicle selects the selected road section as a passable road section based on the first comprehensive optimization parameter on the selected road section, that is, the numerical representation of the selected road section in the current state as the selected road section for vehicle travel; the road section selection probability is obtained by counting the probability that a sufficient number of vehicles select the selected road section as a travel section based on the first comprehensive optimization parameter on the selected road section;

[0137] The training method of the first machine learning model includes:

[0138] Convert the vehicle type value to a numerical value. For example, a type 1 vehicle is converted to 1, a type 2 vehicle is converted to 2, a type 3 vehicle is converted to 3, a type 4 vehicle is converted to 4, and a type 5 vehicle is converted to 5.

[0139] The first comprehensive optimization parameter is converted into a corresponding set of first eigenvectors, the first eigenvectors are used as the input of the first machine learning model, and the section selection probability corresponding to each set of first comprehensive optimization parameters is used as the output of the first machine learning model. The section selection probability is used as the prediction target, and minimizing the sum of prediction errors of all training data is used as the training target. The first machine learning model is trained until the sum of prediction errors converges, and the training is stopped to obtain the first machine learning model that predicts the section selection probability.

[0140] Exemplarily, the first machine learning model is any one of a CNN neural network model or AlexNet;

[0141] The calculation formula for the prediction error is:

[0142] zk=(ak-wk) 2 ;

[0143] Where zk is the prediction error, k is the group number of the first eigenvector; ak is the predicted state value corresponding to the first eigenvector of the kth group, and wk is the actual state value corresponding to the kth group of training data;

[0144] In the first machine learning model, the first feature vector is the first comprehensive optimization parameter, and the state value is the road section selection probability;

[0145] After the acquired first comprehensive optimization parameter is input into the first machine learning model, the road section selection probability corresponding to the first comprehensive optimization parameter can be predicted.

[0146] S3: obtaining the differentiated parameters of the selected road section, and combining the differentiated parameters with the road section selection probability to generate a second comprehensive optimization parameter;

[0147] Differentiation parameters refer to parameters that can affect the total revenue and expenses generated by the passage of vehicles on the selected road section, so as to represent the specific factors affecting the revenue and expenses on the selected road section and serve as the data basis for the subsequent prediction of the total revenue of the road section;

[0148] Differentiation parameters include passing vehicle value, passing distance value and passing overtime value;

[0149] The passing vehicle value refers to the numerical representation of the overall number of vehicles that choose the selected road section as the passing road section, that is, it can represent the macro quantity of vehicles on the selected road section. When the passing vehicle value is larger, it means that the specific number of vehicles that choose the selected road section as the passing road section is more, and the total revenue of the road section is greater;

[0150] Methods for obtaining the passing vehicle value include:

[0151] The attribute data of p vehicles passing through the selected road section within the preset vehicle statistical period are queried from the ETC gantry data to obtain p vehicle attribute data; the attribute data refers to data that can represent the type of the vehicle, thereby providing a basis for the accurate calculation of the subsequent vehicle value; the preset vehicle statistical period is used to calculate the time span corresponding to the statistical vehicle attribute data on the selected road section, so that the vehicle attribute data can be counted within a reasonable and limited time period to ensure that the number of vehicle attribute data is sufficient;

[0152] Through natural language processing technology, semantic recognition is performed on p vehicle attribute data one by one to obtain p key semantics; the key semantics is used to concisely express the true meaning of the vehicle attribute data and serve as the basis for subsequent standard semantic matching;

[0153] Passenger and cargo category 1, passenger and cargo category 2, passenger and cargo category 3, passenger and cargo category 4 and passenger and cargo category 5 are recorded as standard semantics, and the key semantics are matched with the standard semantics one by one;

[0154] The vehicles corresponding to the vehicle attribute data with the key semantics of passenger and cargo are recorded as a type of vehicle, and the number of vehicles in the type is counted;

[0155] The vehicles corresponding to the vehicle attribute data with the key semantics of passenger and freight are recorded as Class II vehicles, and the number of Class II vehicles is counted;

[0156] The vehicles corresponding to the vehicle attribute data with the key semantics of passenger and freight are recorded as three types of vehicles, and the number of the three types of vehicles is counted;

[0157] The vehicles corresponding to the vehicle attribute data with the key semantics of passenger and freight are recorded as four types of vehicles, and the number of the four types of vehicles is counted;

[0158] The vehicles corresponding to the vehicle attribute data with the key semantics of passenger and freight are recorded as five types of vehicles, and the number of vehicles in the five types is counted;

[0159] The number of vehicles of category 1, category 2, category 3, category 4 and category 5 are assigned corresponding proportional coefficients respectively and added to obtain the passing vehicle value;

[0160] The expression of the passing vehicle value is:

[0161] TX cl =σ1*SL 1l +σ2*SL 2l +σ3*SL 3l +σ4*SL 4l +σ5*SL 5l ;

[0162] In the formula, TX cl is the value of passing vehicles, SL 1l is the number of vehicles of type -, SL 2l is the number of second-class vehicles, SL 3l is the number of three types of vehicles, SL 4l is the number of four types of vehicles, SL 5l is the number of five types of vehicles, σ1, σ2, σ3, σ4, σ5 are proportional coefficients greater than 0;

[0163] Among them, σ1+σ2+σ3+σ4+σ5=1. For example, σ1 is 0.23, σ2 is 0.22, σ3 is 0.18, σ4 is 0.20, and σ5 is 0.17.

[0164] The travel distance value refers to the specific driving distance of the vehicle on the selected road section as the passable road section, which can be used to represent the mileage of the vehicle on the selected road section. The larger the travel distance value, the greater the specific driving distance of the vehicle on the selected road section as the passable road section, and the greater the total revenue of the road section;

[0165] Methods for obtaining the travel distance value include:

[0166] The positions of p vehicles when they first enter the selected road section are queried through the ETC gantry data to obtain p initial positions;

[0167] The real-time positions of p vehicles are monitored by the Beidou positioning system. When the real-time position of a vehicle does not coincide with the position of the selected road section, the real-time position of the vehicle is recorded as an abnormal position, and p abnormal positions are obtained;

[0168] The distances between p initial positions and p abnormal positions are measured one by one by the Beidou positioning system to obtain p sub-distance values, and the passing distance value is obtained by accumulating the p sub-distance values ​​one by one;

[0169] The expression of the travel distance value is:

[0170]

[0171] In the formula, TX jl is the travel distance value, JL zc is the cth sub-distance value.

[0172] The overtime value refers to the length of time that the driving time of the vehicle that selects the selected road section as the passable road section exceeds the standard driving time, which can be used to indicate the driving overtime situation of the vehicle on the selected road section. The larger the overtime value, the longer the driving time of the vehicle that selects the selected road section as the passable road section exceeds the standard driving time, and the greater the total revenue of the road section;

[0173] Methods for obtaining the pass timeout value include:

[0174] The driving time data of p vehicles on the selected road section in the ETC gantry data are queried one by one through the timestamp, and the entry time and exit time in the driving time data are identified through natural language processing technology to obtain p entry times and p exit times;

[0175] The time between the p entry moments and the p exit moments is recorded as the travel time, and p travel times are obtained;

[0176] The travel time longer than the standard travel time is recorded as the target time, and s target time lengths are obtained; the standard travel time refers to the maximum time length required for a vehicle to travel on the selected road section, which can provide a comparison basis for the identification of the target time length; the standard travel time is obtained by collecting the maximum time lengths required for a large number of vehicles to travel on the selected road section in history and calculating their average value;

[0177] Subtract the s target durations from the standard travel durations in turn to obtain s sub-timeout values, and then add up the s sub-timeout values ​​one by one to obtain the travel timeout value;

[0178] The expression for the pass timeout value is:

[0179]

[0180] In the formula, TX cs is the timeout value, SC mbd is the duration of the dth target, SC bz This is the standard travel time.

[0181] After obtaining the differentiated parameters, the obtained passing vehicle value, passing distance value and passing timeout value can be combined with the predicted section selection probability, so that the passing vehicle value, passing distance value, passing timeout value and predicted section selection probability can be combined together to form a second comprehensive optimization parameter as the predicted total section duration.

[0182] S4: inputting the second comprehensive optimization parameter into a second machine learning model trained in advance for predicting the total revenue of the road section to predict the total revenue of the road section;

[0183] The second machine learning model is a prediction model for predicting the highway toll that can be generated when the selected road section is used as a toll road section based on the second comprehensive optimization parameter, that is, the total cost of a vehicle entering the selected road section can be predicted, and it can also provide direct data support for the toll pricing of subsequent highway sections;

[0184] The total revenue of a road section refers to the total revenue of a vehicle when passing through the road section after obtaining the second comprehensive optimization parameter; the total revenue of a road section is obtained by querying the database of the highway management platform;

[0185] The training method of the second machine learning model includes:

[0186] The second comprehensive optimization parameters are converted into a corresponding set of second eigenvectors, the second eigenvectors are used as the input of the second machine learning model, and the total revenue of the road section corresponding to each set of second comprehensive optimization parameters is used as the output of the second machine learning model. The total revenue of the road section is used as the prediction target, and minimizing the sum of prediction errors of all training data is used as the training target. The second machine learning model is trained until the sum of prediction errors converges, and the training is stopped to obtain a second machine learning model that predicts the total revenue of the road section.

[0187] Exemplarily, the second machine learning model is any one of the CNN neural network model or AlexNet, and other relevant parameters and setting logic of the second machine learning model are consistent with other relevant parameters and setting logic of the above-mentioned first machine learning model.

[0188] After the acquired second comprehensive optimization parameters are input into the second machine learning model, the total revenue of the road section corresponding to the second comprehensive optimization parameters can be predicted.

[0189] S5: comparing the predicted total revenue of the road section with the preset total revenue threshold, generating toll pricing prompt information, and formulating a toll pricing optimization strategy;

[0190] The toll pricing prompt information is used to provide prompt information that meets the comparison result after the predicted total revenue of the road section is compared with the preset total revenue threshold, and serves as a guide for the subsequent toll pricing optimization strategy;

[0191] The toll pricing prompt information includes insufficient toll pricing prompt information and sufficient toll pricing prompt information; the insufficient toll pricing prompt information refers to prompt information corresponding to when the total revenue of the selected road section is low, and the sufficient toll pricing prompt information refers to prompt information corresponding to when the total revenue of the selected road section is high;

[0192] The method for generating the insufficient charging pricing prompt information and the sufficient charging pricing prompt information includes:

[0193] The predicted total revenue SR of the road segment yc With the preset total income threshold SR ys comparison; the preset total revenue threshold refers to the minimum value of the total revenue of toll collection pre-set for the selected road section and according to the preset toll value, and serves as a numerical basis for judging whether the total revenue of the selected road section is sufficient or insufficient;

[0194] When SR yc Greater than or equal to SR ys When , it indicates that the predicted total revenue of the road section is greater than or equal to the preset total revenue threshold. At this time, the total revenue of the selected road section is sufficient, and a prompt message of sufficient toll pricing is generated;

[0195] When SR yc Less than SR ys When , it means that the predicted total revenue of the road section is less than the preset total revenue threshold. At this time, the total revenue of the selected road section is low, and a reminder message of insufficient charging pricing is generated.

[0196] The toll pricing optimization strategy is a specific strategy that can be formulated to adjust the preset toll value according to different toll pricing prompt information, so as to ensure that the total revenue of the selected road section can be maintained at a reasonable and high level, thereby achieving the effect of optimizing the toll pricing strategy according to the differentiation of highway sections and achieving a dynamic balance between toll positioning and total revenue of the road section;

[0197] The charging pricing optimization strategy includes a strategy of maintaining a preset charging value and a strategy of reducing a preset charging value; the strategy of maintaining a preset charging value refers to an optimization scheme in which the preset charging value remains unchanged, and the strategy of reducing a preset charging value refers to an optimization scheme in which the preset charging value needs to be reduced;

[0198] The method for formulating the strategy of maintaining the preset charging value and the strategy of reducing the preset charging value includes:

[0199] When the charging pricing sufficient prompt information is generated, it means that the total revenue of the selected road section can reach the preset level, and the preset charging value does not need to be adjusted at this time, so a strategy for maintaining the preset charging value is formulated;

[0200] When a reminder message of insufficient toll pricing is generated, it indicates that the total revenue of the selected road section cannot reach the preset level. At this time, the preset toll value needs to be adjusted, and a strategy for reducing the preset toll value is formulated.

[0201] In this embodiment, comprehensive traffic parameters are extracted from ETC gantry data, the comprehensive traffic parameters are combined with preset charging values ​​to generate a first comprehensive optimization parameter, the first comprehensive optimization parameter is input into a first machine learning model trained in advance for predicting the probability of section selection, the section selection probability is predicted, the differentiated parameters of the selected section are obtained, and the differentiated parameters are combined with the section selection probability to generate a second comprehensive optimization parameter, the second comprehensive optimization parameter is input into a second machine learning model trained in advance for predicting the total revenue of the section, the total revenue of the section is predicted, the predicted total revenue of the section is compared with the preset total revenue threshold, to generate charging pricing prompt information, and formulate a charging pricing optimization strategy; compared with the prior art, by generating the first comprehensive optimization parameter and combining the first comprehensive optimization parameter with the first machine learning model, the second comprehensive optimization parameter is input into the second machine learning model trained in advance for predicting the total revenue of the section, the total revenue of the section is predicted, the predicted total revenue of the section is compared with the preset total revenue threshold, to generate charging pricing prompt information, and formulate a charging pricing optimization strategy; compared with the prior art, by generating the first comprehensive optimization parameter and combining the first comprehensive optimization parameter with the first machine learning model, the second comprehensive optimization parameter is input into the second machine learning model trained in advance for predicting the total revenue of the section, the total revenue of the section is predicted, the predicted total revenue of the section is compared with the preset total revenue threshold, the charging pricing prompt information is generated, and the charging pricing optimization strategy is formulated ... A machine learning model can accurately predict the probability that a vehicle selects a selected road section as a passable road section, thereby accurately predicting the probability of selecting a highway section. At the same time, combined with the second comprehensive optimization parameter and the second machine learning model, it can, on the basis of selecting a road section as a passable road section, combine the differentiated influencing parameters of different roads sections to quickly and accurately predict the total revenue of the road section when the vehicle passes through. Based on the comparative analysis results of the total revenue of the road section, specific strategies and measures for optimizing the road section toll pricing are formulated in a targeted manner, so that the highway section can achieve a dynamic balance between the probability of section selection and the total revenue of the road section based on the differentiated influencing parameters, thereby ensuring that the highway section can achieve the purpose of differentiated toll pricing optimization, greatly improving the total revenue of the highway.

[0202] Example 2: Please refer to Figure 2 As shown, the part not described in detail in this embodiment is described in the description of the embodiment =, and a highway section differentiated charging pricing optimization system based on ETC gantry data is provided, which is used to implement a highway section differentiated charging pricing optimization method based on ETC gantry data, including a first parameter generation module, a first model training module, a second parameter generation module, a second model training module and an optimization strategy formulation module, wherein each module is connected via a wired or wireless network;

[0203] A first parameter generation module is used to extract comprehensive traffic parameters from ETC gantry data, and generate first comprehensive optimization parameters after combining the comprehensive traffic parameters with preset charging values. The comprehensive traffic parameters include vehicle type values, peak time values, travel time values, and road section congestion rate;

[0204] A first model training module, used for inputting the first comprehensive optimization parameter into a first machine learning model trained in advance for predicting the probability of road section selection, to predict the probability of road section selection;

[0205] A second parameter generation module is used to obtain the differentiated parameters of the selected road section, and generate the second comprehensive optimization parameters after combining the differentiated parameters with the road section selection probability, wherein the differentiated parameters include a passing vehicle value, a passing distance value, and a passing timeout value;

[0206] A second model training module is used to input the second comprehensive optimization parameter into a second machine learning model trained in advance for predicting the total revenue of a road section, so as to predict the total revenue of the road section;

[0207] The optimization strategy formulation module is used to compare the predicted total revenue of the road section with the preset total revenue threshold, generate charging pricing prompt information, the charging pricing prompt information includes insufficient charging pricing prompt information and sufficient charging pricing prompt information, and formulate a charging pricing optimization strategy, the charging pricing optimization strategy includes maintaining the preset charging value strategy and reducing the preset charging value strategy.

[0208] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A highway section differentiated toll pricing optimization method based on ETC gantry data, characterized in that: include: S1: extracting comprehensive traffic parameters from ETC gantry data, combining the comprehensive traffic parameters with preset charging values, and generating first comprehensive optimization parameters, wherein the comprehensive traffic parameters include vehicle type value, peak time value, travel time value, and road section congestion rate; S2: inputting the first comprehensive optimization parameter into a first machine learning model trained in advance for predicting the probability of road section selection, and predicting the probability of road section selection; S3: obtaining differentiated parameters of the selected road section, and combining the differentiated parameters with the road section selection probability to generate a second comprehensive optimization parameter, wherein the differentiated parameters include a passing vehicle value, a passing distance value, and a passing timeout value; S4: inputting the second comprehensive optimization parameter into a second machine learning model trained in advance for predicting the total revenue of the road section to predict the total revenue of the road section; S5: Compare the predicted total revenue of the road section with the preset total revenue threshold, generate charging pricing prompt information, the charging pricing prompt information includes insufficient charging pricing prompt information and sufficient charging pricing prompt information, and formulate a charging pricing optimization strategy, the charging pricing optimization strategy includes maintaining the preset charging value strategy and reducing the preset charging value strategy.

2. The highway section differentiated toll pricing optimization method based on ETC gantry data according to claim 1 is characterized in that: The vehicle type values ​​include class 1 vehicle, class 2 vehicle, class 3 vehicle, class 4 vehicle and class 5 vehicle; The methods for extracting peak hour values ​​include: The driving time data of i vehicles on the selected road section in the ETC gantry data are queried one by one through the timestamp, and all the entry and exit times in the driving time data are identified through natural language processing technology to obtain i entry times and i exit times; The period between the i entry time and the i exit time is recorded as the driving period, and i driving periods are obtained; Compare the i driving periods with the standard peak periods in turn, and record the periods in which the i driving periods overlap with the standard peak periods as valid periods, to obtain i valid periods; Compare the duration of the i valid periods with the duration of the i driving periods in turn to obtain i sub-peak values; The expression of the sub-peak value is: In the formula, GF zi is the peak value of the ith sub-peak, SC yxi is the length of the i-th valid period, SC xsi is the duration of the i-th driving period; After removing the maximum and minimum sub-peak values, the remaining i-2 sub-peak values ​​are accumulated and averaged to obtain the peak period value; The expression for the peak hour value is: In the formula, GF sd is the peak time value, GF za is the ath sub-peak value.

3. The highway section differentiated toll pricing optimization method based on ETC gantry data according to claim 2 is characterized in that: The method for extracting the travel time value includes: Query the first and last time that the travel time data of each vehicle in the ETC gantry data is generated by the timestamp, and obtain the first and last start and end times respectively; The duration between l start times and l end times is recorded as the total duration, and l total duration is obtained; The length of the selected section and the total length corresponding to the entire journey duration are measured by the Beidou positioning system, and the length of the selected section is compared with the total length corresponding to the entire journey duration to obtain the length ratio value; The expression of length ratio is: In the formula, CD zb is the length ratio, CD xd is the length of the selected section, CD zc is the total length corresponding to the whole journey; Multiply the length ratio by the i-th total duration to obtain i-th product values, and then add up the i-th product values ​​and calculate the average to obtain the travel duration value; The expression of the travel time value is: In the formula, TX sc is the travel time value, SC qcb is the duration of the bth full journey.

4. The highway section differentiated toll pricing optimization method based on ETC gantry data according to claim 3 is characterized in that: The method for extracting the road section congestion rate includes: Randomly mark two ETC gantries at adjacent positions on the selected road section and record them as the first gantry and the second gantry respectively; Identify the moments when i vehicles pass through the first gantry and the second gantry from the ETC gantry data, and obtain a first moment and a second moment; Record the duration between the i first moments and the i second moments as the elapsed duration, and obtain i elapsed durations; The distance between the first gantry and the second gantry is measured by the Beidou positioning system to obtain the passing distance, and the i passing times are compared with the passing distances in turn to obtain i passing speeds; The expression of the passing speed is: In the formula, SD jgi is the i-th passing speed, JL jg is the distance travelled, SC jgi is the i-th elapsed time; The passing speed that is less than the standard speed threshold is recorded as the congestion speed, and the number of congestion speeds is counted. After comparing the number of congestion speeds with the number of passing speeds, the congestion rate of the road section is obtained; The expression of road section congestion rate is: In the formula, YD ld is the road congestion rate, SL yd is the number of congestion speeds.

5. The highway section differentiated toll pricing optimization method based on ETC gantry data according to claim 4 is characterized in that: The training method of the first machine learning model includes: Convert the vehicle type value to 1 for a type 1 vehicle, 2 for a type 2 vehicle, 3 for a type 3 vehicle, 4 for a type 4 vehicle, and 5 for a type 5 vehicle. The first comprehensive optimization parameter is converted into a corresponding set of first eigenvectors, the first eigenvectors are used as the input of the first machine learning model, and the section selection probability corresponding to each set of first comprehensive optimization parameters is used as the output of the first machine learning model. The section selection probability is used as the prediction target, and minimizing the sum of prediction errors of all training data is used as the training target. The first machine learning model is trained until the sum of prediction errors converges, and the training is stopped to obtain the first machine learning model that predicts the section selection probability.

6. The highway section differentiated toll pricing optimization method based on ETC gantry data according to claim 5 is characterized in that: The method for obtaining the passing vehicle value includes: Query the attribute data of p vehicles passing through the selected road section within a preset vehicle statistical period from the ETC gantry data to obtain the attribute data of p vehicles; Through natural language processing technology, semantic recognition is performed on p vehicle attribute data one by one to obtain p key semantics; Passenger and cargo category 1, passenger and cargo category 2, passenger and cargo category 3, passenger and cargo category 4 and passenger and cargo category 5 are recorded as standard semantics, and the key semantics are matched with the standard semantics one by one; The vehicles corresponding to the vehicle attribute data with the key semantics of passenger and freight category 1, passenger and freight category 2, passenger and freight category 3, passenger and freight category 4, and passenger and freight category 5 are recorded as category 1 vehicles, category 2 vehicles, category 3 vehicles, category 4 vehicles, and category 5 vehicles, respectively, and the number of category 1 vehicles, category 2 vehicles, category 3 vehicles, category 4 vehicles, and category 5 vehicles are counted one by one; The number of vehicles of category 1, category 2, category 3, category 4 and category 5 are assigned corresponding proportional coefficients respectively and added to obtain the passing vehicle value; The expression of the passing vehicle value is: TX cl =σ1*SL 1l +σ2*SL 2l +σ3*SL 3l +σ4*SL 4l +σ5*SL 5l ; In the formula, TX cl is the value of passing vehicles, SL 1l is the number of vehicles of a type, SL 2l is the number of second-class vehicles, SL 3l is the number of three types of vehicles, SL 4l is the number of four types of vehicles, SL 5l is the number of five types of vehicles, σ1, σ2, σ3, σ4, and σ5 are proportional coefficients greater than 0.

7. The method for optimizing highway section differentiated toll pricing based on ETC gantry data according to claim 6 is characterized in that: The method for obtaining the travel distance value includes: The positions of p vehicles when they first enter the selected road section are queried through the ETC gantry data to obtain p initial positions; The real-time positions of p vehicles are monitored by the Beidou positioning system. When the real-time position of a vehicle does not coincide with the position of the selected road section, the real-time position of the vehicle is recorded as an abnormal position, and p abnormal positions are obtained; The distances between p initial positions and p abnormal positions are measured one by one by the Beidou positioning system to obtain p sub-distance values, and the passing distance value is obtained by accumulating the p sub-distance values ​​one by one; The expression of the travel distance value is: In the formula, TX jl is the travel distance value, JL zc is the cth sub-distance value.

8. The highway section differentiated toll pricing optimization method based on ETC gantry data according to claim 7 is characterized in that: The method for obtaining the pass timeout value includes: The driving time data of p vehicles on the selected road section in the ETC gantry data are queried one by one through the timestamp, and the entry time and exit time in the driving time data are identified through natural language processing technology to obtain p entry times and p exit times; The time between the p entry moments and the p exit moments is recorded as the travel time, and p travel times are obtained; The travel time longer than the standard travel time is recorded as the target time, and s target time lengths are obtained; Subtract the s target durations from the standard travel durations in turn to obtain s sub-timeout values, and then add up the s sub-timeout values ​​one by one to obtain the travel timeout value; The expression for the pass timeout value is: In the formula, TX cs is the timeout value, SC mbd is the duration of the dth target, SC bz This is the standard travel time.

9. The highway section differentiated toll pricing optimization method based on ETC gantry data according to claim 8 is characterized in that: The training method of the second machine learning model includes: The second comprehensive optimization parameters are converted into a corresponding set of second eigenvectors, the second eigenvectors are used as the input of the second machine learning model, and the total revenue of the road section corresponding to each set of second comprehensive optimization parameters is used as the output of the second machine learning model. The total revenue of the road section is used as the prediction target, and minimizing the sum of prediction errors of all training data is used as the training target. The second machine learning model is trained until the sum of prediction errors converges, and the training is stopped to obtain a second machine learning model that predicts the total revenue of the road section.

10. The highway section differentiated toll pricing optimization method based on ETC gantry data according to claim 9 is characterized in that: The method for generating the charging pricing insufficient prompt information and the charging pricing sufficient prompt information comprises: The predicted total revenue SR of the road segment yc With the preset total income threshold SR ys Compare; When SR yc Greater than or equal to sR ys Generate sufficient reminder information for charging pricing; When SR yc Less than SR ys When the charging price is insufficient, a prompt message is generated; The method for formulating the strategy of maintaining the preset charging value and the strategy of reducing the preset charging value includes: When generating sufficient reminder information for charging pricing, formulate a strategy to maintain the preset charging value; When a reminder message of insufficient charging pricing is generated, a strategy for reducing the preset charging value is formulated.

Images