CN111125862B - Following model emission measuring and calculating method based on genetic algorithm and specific power - Google Patents

Abstract

The invention discloses a following model emission measuring and calculating method based on a genetic algorithm and specific power, which relates to the technical field of traffic environment, and the following model emission measuring and calculating method optimizes all parameters of the following model, ensures the comprehensiveness of model parameter optimization, utilizes the genetic algorithm to carry out cross pairing on all parameters, ensures that the model can obtain an optimal solution, reduces the combination number of model parameters by utilizing the characteristics of the genetic algorithm, and improves the model optimizing efficiency; meanwhile, the invention is an optimization method provided by utilizing a genetic algorithm, so that the method is suitable for a plurality of following models, in addition, after the genetic algorithm is utilized to optimize the parameters of the following models, compared with actual measurement, the time-speed and acceleration curve of the vehicle after the following is simulated and output by the model is obviously improved, the VSP distribution curve of the vehicle after the following is simulated and output by the model is obviously improved, the VSP mean square error is obviously reduced, and the estimated error of the total emission amount is also obviously reduced.

Description

An emission calculation method for car-following model based on genetic algorithm and specific power

Technical field

The invention relates to the technical field of traffic environment, and in particular to a car-following model emission measurement method based on genetic algorithm and specific power.

Background technique

Combining microscopic traffic simulation with emission models to analyze the impact of traffic management strategies on motor vehicle emissions has become a research hotspot in the field of transportation environment. The vehicle following model is used to describe the instantaneous speed and acceleration of the vehicle following the vehicle under different conditions. However, in current simulations, because the following behavior described by the vehicle following model is too radical, it often cannot be used to calculate the emissions of the vehicle following the vehicle. . In their research on the optimization of car-following models, many scholars try to adjust the model parameters to obtain a satisfactory result. For example, in Geng Zhongbo's optimization research on the Wiedemann74 and Fritzsche car-following models, he first conducted a sensitivity analysis on the model parameters, then set the threshold range and parameter value step size for the sensitivity parameters, and finally conducted a cross-over of the parameters. value, and finally determined 399 parameter combinations, which is currently the most commonly used method for optimizing the parameters of the car-following model.

Although the above research methods can optimize the model simulation effect, they have the following shortcomings:

(1) In the car-following model, there are many parameters that can be optimized. Simply extracting key parameters for optimization from parameter sensitivity analysis is not enough to ensure the comprehensiveness of model parameter optimization;

(2) When optimizing the parameters of the car-following model, what must be determined is the parameter threshold range and value step of the model, which is related to the final model parameter combination; for example, if 11 of the Wiedeman74 car-following models are optimized The parameters are optimized, and 10 combinations are obtained for each parameter by setting the threshold and step size. Then when the 11 parameters are paired and verified, there are a total of 10 ¹¹ parameter combinations. Due to the excessive calculation amount, even a high-performance computer It is also impossible to calculate the results in a short time;

(3) In the research on the optimization of car-following models for emission measurement, multiple sensitivity analyzes are often required for different car-following models to determine the optimization parameters, which reduces work efficiency.

Contents of the invention

The embodiment of the present invention provides a car-following model emission measurement method based on genetic algorithm and specific power to solve the problems existing in the existing technology.

A car-following model emission measurement method based on genetic algorithm and specific power, including:

S1. Obtain the measured second-by-second speed and acceleration data of the car ahead, and check and clean the data to ensure the validity of the data obtained;

S2. Program the car-following model. Based on the measured second-by-second speed and acceleration data of the car following the car ahead, simulate the second-by-second speed and acceleration data of the car following the car behind. Compare the motor vehicle ratio between the simulated car following the car and the measured vehicle ratio of the car following the car ahead. Power VSP is used as an emission evaluation index, and the simulated and measured mean square error MSE is calculated;

S3. Use the genetic algorithm to optimize the parameters of the car-following model and obtain the optimal parameter solution of the model;

S4. Evaluation of model effects based on vehicle speed, acceleration and VSP: Use the optimal parameter solution of the model obtained in S3, substitute it into S2, and obtain the time-speed, time-acceleration trajectory diagram and VSP distribution diagram after parameter optimization;

S5. Based on the speed-VSP emission model effect evaluation, calculate the actual vehicle emissions and the vehicle emissions of the car-following model before and after parameter optimization, and perform comparative analysis and verification.

Preferably, in S2, the second-by-second speed and acceleration data of the simulated following vehicle based on the measured second-by-second speed and acceleration data of the following vehicle include:

S21. First determine which following state the car behind is in at the current moment;

S22. Then determine the simulated acceleration of the car following the car based on the different acceleration calculation methods in each following state, thereby determining the simulated speed of the car following the car in the next second;

S23. Then, based on the simulated speed and simulated acceleration of the following car and the measured speed and acceleration of the following car, cycle the speed and acceleration of the following car one second after the simulation.

Preferably, in said S2,

The calculation formula of the motor vehicle specific power VSP is:

VSP=v(1.1a+0.132)+0.000302v ³

Among them, v is the velocity and a is the acceleration;

The calculation formula of the simulated and measured mean square error MSE is:

Among them: S _i is the simulated VSP at the i-th second; T _i is the measured VSP at the i-th second.

Preferably, in S3, using a genetic algorithm to optimize the parameters of the following model includes:

S31. Parameter selection and threshold setting: Select the parameters and value ranges that need to be optimized in the car-following model;

S32. Population binary coding: convert the optimized parameter values into binary coding;

S33. Determine the number of initial populations: the number of individuals randomly composed within the parameter range of the selected optimization parameters;

S34. Calculate the population fitness function: that is, the evaluation function of the optimization objective;

S35. Selection operation: According to the principle of "natural selection, survival of the fittest", select some of the best individuals as the next generation population. Corresponding to the fitness function above, the smaller the MSE value, the higher the simulated VSP and the measured VSP. The closer the distribution is, the greater the probability that the following model parameters corresponding to smaller MSE values will be retained in the next generation;

S36. Crossover and mutation operations: The optimization parameters are binary codes, just like a chromosome. Our crossover and mutation operations on the chromosome are equivalent to increasing the randomness of the next generation of individuals;

S37. Maximum number of iterations: that is, when the model iterates to the set termination number, it will automatically stop and output the optimal solution of the model;

S38. Convergence tolerance: that is, when the error of the optimal solution output by the model between the offspring and the parent is less than the set value, the model terminates itself and outputs the optimal parameter solution of the model.

Beneficial effects of the present invention: The present invention optimizes all parameters of the car-following model to ensure the comprehensiveness of model parameter optimization. It uses a genetic algorithm to cross-match all parameters, which not only ensures that the model can obtain the optimal solution, but also utilizes the characteristics of the genetic algorithm to reduce The number of combinations of model parameters improves the efficiency of model optimization; at the same time, because the present invention is an optimization method proposed by using a genetic algorithm, it is suitable for many car-following models. In addition, after the present invention uses a genetic algorithm to optimize the parameters of the car-following model, it is compared with Compared with the actual measurement, the model simulation output time-speed and acceleration curves of the car following the car have been significantly improved, the model simulation output VSP distribution curve of the car behind has been significantly improved, the VSP mean square error has been significantly reduced, and the total emissions The estimation error is also significantly reduced.

Description of drawings

Figure 1 is a flow chart of a car-following model emission measurement method based on genetic algorithm and specific power provided by an embodiment of the present invention;

Figure 2 is a car-following data collection driving route map and collection equipment for a car-following model emission measurement method based on genetic algorithm and specific power provided by an embodiment of the present invention;

Figure 3 is a Wiedemann74 car-following model simulation VSP distribution diagram before parameter optimization of a car-following model emission measurement method based on genetic algorithm and specific power provided by the embodiment of the present invention;

Figure 4 is a Fritzsche car-following model simulation VSP distribution diagram before parameter optimization of a car-following model emission measurement method based on genetic algorithm and specific power provided by the embodiment of the present invention;

Figure 5 shows the measured simulation time-speed and time-acceleration curves of the Wiedemann74 car-following model before and after parameter optimization of a car-following model emission measurement method based on genetic algorithm and specific power provided by the embodiment of the present invention;

Figure 6 is a VSP distribution diagram after parameter optimization of the Wiedemann74 car-following model simulation parameters of a car-following model emission measurement method based on genetic algorithm and specific power provided by the embodiment of the present invention;

Figure 7 shows the measured simulation time-speed and time-acceleration curves of the Fritzsche car-following model before and after parameter optimization of a car-following model emission measurement method based on genetic algorithm and specific power provided by the embodiment of the present invention;

Figure 8 is a VSP distribution diagram before and after parameter optimization of the Fritzsche car-following model simulation parameter optimization of a car-following model emission measurement method based on genetic algorithm and specific power provided by the embodiment of the present invention;

Figure 9 is a comparison chart of the actual measurement and the total emissions before and after the optimization of the parameters of the car-following model of a car-following model emission measurement method based on genetic algorithm and specific power provided by the embodiment of the present invention;

Figure 10 is a comparison chart of the overall errors before and after optimization of the car-following model parameters of a car-following model emission measurement method based on genetic algorithm and specific power provided by the embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. However, it should be understood that the protection scope of the present invention is not limited by the specific embodiments.

Referring to Figures 1-10, the present invention provides a car-following model emission measurement method based on genetic algorithm and specific power.

Refer to Figure 1, including:

S1. Obtain the measured second-by-second speed and acceleration data of the car ahead, and check and clean the data to ensure the validity of the data obtained;

S2. Use python to program the car-following model, and simulate the second-by-second speed and acceleration data of the car following the car based on the measured second-by-second speed and acceleration data of the car ahead, which includes:

S21. First determine which following state the car behind is in at the current moment;

S22. Then determine the simulated acceleration of the car following the car based on the different acceleration calculation methods in each following state, thereby determining the simulated speed of the car following the car in the next second;

S23. Then, based on the simulated speed and simulated acceleration of the following car and the measured speed and acceleration of the following car, cycle the speed and acceleration of the following car one second after the simulation.

Use the simulation to output the speed and acceleration of the car following the car, make time-speed curves and time-acceleration curves, and compare the vehicle specific power VSP of the simulated car following the car and the measured vehicle following the car in front, as an emission evaluation index. Related research shows , VSP and emission have significant physical meaning, and calculate the mean square error MSE between simulation and actual measurement, and compare the difference between simulation and actual measurement;

The calculation formula of the motor vehicle specific power VSP is:

VSP=v(1.1a+0.132)+0.000302v ³

Among them, v is the velocity and a is the acceleration;

The calculation formula of the simulated and measured mean square error MSE is:

Among them: S _i is the simulated VSP at the i-th second; T _i is the measured VSP at the i-th second.

S3. Use the genetic algorithm to optimize the parameters of the car-following model and obtain the optimal parameter solution of the model;

The parameter calibration flow chart of the following model is shown in Figure 2. In the genetic algorithm, some key parameters of the genetic algorithm need to be set. The specific parameters and description of the genetic algorithm are shown in Table 1 below:

Table 1 Optimization parameter settings based on genetic algorithm following model

Using genetic algorithms to optimize the parameters of the car-following model specifically includes:

S31. Parameter selection and threshold setting: that is, selecting the parameters and value ranges that need to be optimized in the car-following model;

S32. Population binary coding: Binary is generally selected as the individual coding method, that is, the optimized parameter values are converted into binary coding to facilitate subsequent cross-mutation operations;

S33. Determine the number of initial populations: that is, the number of individuals randomly formed by the selected optimization parameters within the parameter range; there is no clear value for the number of populations, and it is generally selected within 200. However, as the model becomes more complex, the parameters to be optimized will change. If there are more, the initial population will be larger.

S34. Calculate the population fitness function: that is, the evaluation function of the optimization target. In the present invention, the MSE of the measured VSP of the vehicle behind and the vehicle behind is calculated;

S35. Selection operation: According to the principle of "natural selection, survival of the fittest", select some of the best individuals as the next generation population, corresponding to the fitness function above, that is, the smaller the MSE value, the better the simulated and the measured ones. The closer the VSP distribution is, the greater the probability that the following model parameters corresponding to smaller MSE values will be retained in the next generation;

S36. Crossover and mutation operations: The optimization parameters are binary codes, just like a chromosome. Our crossover and mutation operations on the chromosome are equivalent to increasing the randomness of the next generation of individuals;

S37. Maximum number of iterations: that is, when the model iterates to the set termination number, it will automatically stop and output the optimal solution (minimum MSE) of the model;

S38. Convergence tolerance: that is, when the error of the optimal solution output by the model between the offspring and the parent is less than the set value, the model terminates itself and outputs the optimal parameter solution of the model.

S4. Evaluation of model effects based on vehicle speed, acceleration and VSP: Use the optimal parameter solution of the model obtained in S3, substitute it into S2, and obtain the time-speed, time-acceleration trajectory diagram and VSP distribution diagram after parameter optimization;

S5. Based on the speed-VSP emission model effect evaluation, calculate the actual vehicle emissions and the vehicle emissions of the car-following model before and after parameter optimization, and perform comparative analysis and verification.

Referring to Figures 2-8, the genetic algorithm of the present invention is suitable for parameter optimization of car-following models such as Wiedemann74, Wiedemann99, Fritzsche and FVDM. The Wiedemann74 and Fritzsche car-following models are used as examples below.

Step 1: Measured data collection

This article respectively selects the Second Ring Road in Xi'an, the entire section of Ziwu Avenue (expressway), the section from Aerospace City to Bell Tower of Chang'an Road (trunk road), the entire section of Ziqiang Road (secondary trunk road), and the entire section of Dongqi Road (branch road). Carry out a car-following test. The test vehicle is a light gasoline vehicle, a Ford Focus, and the driver has three years of driving experience. Data collection uses a Hi-drive car-following data collection instrument, which consists of a front-mounted laser, radar and built-in GPS. It can obtain data such as the distance between the front and rear cars, the speed difference between the front and rear cars, the speed of the own car, and acceleration.

The car-following test vehicle is equipped with Hi-drive equipment and can drive normally on urban roads of different grades. The car in front of it is a random vehicle on the driving route. Because Hi-drive equipment may suffer signal interference when encountering tunnels or overpasses, which may cause the vehicle's speed to be lost. Therefore, this article uses linear interpolation to complete the speed and acceleration where the data is lost within 5 seconds. If the data is lost for more than 5 seconds, it will be used as the dividing point between the front and rear sections of data; Hi-drive is prone to data loss of the vehicle ahead when driving on curves, so this test data only selects straight road sections and excludes turning pairs of data. In addition, due to vehicle fuel consumption and emissions, which are greatly affected by road gradient, the data on a few road sections with large road gradients were eliminated, and finally the second-by-second driving behavior data on different levels of test lines were obtained.

Step 2: Effect evaluation before parameter optimization

In order to more intuitively compare the effects of the two car-following models on vehicle following car simulation, the present invention selects 2500 seconds of continuous measured driving data as an example. During this period of time, the vehicle behind the measured vehicle not only accelerated to a near-free-flow state of driving at the speed limit, but also performed emergency braking, decelerating and stopping. Most of the time, the driving conditions of the vehicle were relatively stable, and the driving conditions of the vehicle also included Stop waiting time, the speed curve during this time includes all road conditions. On this basis, the present invention uses two physiological-psychological models, the Wiedemann74 car-following model and the Fritzsche car-following model, and uses the numerical simulation method designed by the present invention to simulate the car-following behavior of the virtual vehicle behind.

Step 3: Use genetic algorithm to optimize the parameters of the car-following model

(1) Optimize parameter selection and parameter value range settings, as shown in Table 2 and Table 3 below

Table 2 Wiedemann74 car-following model optimization parameter selection

Table 3 Fritzsche car-following model optimization parameter selection

(2). Set the parameters of the genetic algorithm itself as shown in Table 4 below.

Table 4 Optimization parameter settings based on genetic algorithm following model

(3). The optimal solution of output model parameters is shown in Table 5 and Table 6 below:

Table 5 Wiedemann74 car-following model parameter optimization results

Note: () inside is the mean square error of VSP before parameter optimization

Table 6 Parameter optimization results of Fritzsche car-following model

Note: () inside is the mean square error of VSP before parameter optimization

Step 4: Evaluation of V, a and VSP model effects

It can be seen in Table 5 and Table 6 that after parameter optimization, the mean square error of VSP has been significantly reduced. In order to compare the parameter optimization effect more intuitively, the present invention makes the Wiedemann74 car-following model and Fritzsche car-following model before and after parameter optimization. Model time-velocity, time acceleration and VSP distribution plots are evaluated,

Step 5: Evaluation of improvement effects based on speed and VSP emission calculation results

The emission algorithm based on speed and VSP is used to calculate the emissions of the measured rear vehicle and simulated vehicle. The total emissions of the measured vehicle and the simulated vehicle before and after parameter optimization were calculated respectively. Here, the total emission calculation formula in the emission algorithm based on the speed-VSP interval is as follows:

In the formula: Q _i is the corresponding emission rate in each partition, and time_Qi is the duration in each partition. Taking a motor vehicle as an example, by collecting actual emission data through on-board emission testing, the emission rate table corresponding to different speed-VSP intervals can be obtained, as shown in Table 7 below:

Table 7 Emission rates and V-VSP

Referring to Figure 9, the total emissions in each speed range before and after the parameters of the measured vehicle and car-following model are optimized.

The present invention conducts an overall error verification formula by comparing the simulated and measured overall emission totals, as shown below: The overall error diagram before and after the optimization of the car-following model parameters compared with the actual measurement is shown in Figure 10:

In the formula: X is the predicted total emissions of (CO ₂ , CO, NOx, HC); Y is the actual total emissions.

To sum up, the present invention optimizes all parameters of the car-following model to ensure the comprehensiveness of model parameter optimization. It uses a genetic algorithm to cross-match all parameters, which not only ensures that the model can obtain the optimal solution, but also uses the characteristics of the genetic algorithm to reduce The number of combinations of model parameters improves the efficiency of model optimization; at the same time, because the present invention is an optimization method proposed by using a genetic algorithm, it is suitable for many car-following models. In addition, after the present invention uses a genetic algorithm to optimize the parameters of the car-following model, it is compared with Compared with the actual measurement, the time-velocity and acceleration curves of the model simulation output following the car behind have been significantly improved, the VSP distribution curve of the model simulation output following the car behind has been significantly improved, and the VSP mean square error has been significantly reduced.

What is disclosed above is only a specific embodiment of the present invention. However, the embodiment of the present invention is not limited thereto. Any changes that can be thought of by those skilled in the art should fall within the protection scope of the present invention.

Claims (2)

1. A following model emission measuring and calculating method based on genetic algorithm and specific power is characterized by comprising the following steps:

s1, acquiring the second-by-second speed and acceleration data of an actual-measured vehicle before following a car, and checking and cleaning the data to ensure the effectiveness of the acquired data;

s2, programming a following model, simulating the second-by-second speed and acceleration data of a following vehicle according to the second-by-second speed and acceleration data of the following vehicle, comparing the motor vehicle specific power VSP of the simulated following vehicle and the actual following vehicle as emission evaluation indexes, and calculating simulated and actual mean square error MSE; in the step S2, the simulating the second-by-second speed and acceleration data of the following vehicle according to the actually measured second-by-second speed and acceleration data of the following vehicle includes:

s21, firstly judging which following state the following vehicle is in at the current moment;

s22, determining the simulation acceleration of the following vehicle according to different acceleration calculation modes of each following state, so as to determine the simulation speed of the following vehicle in the next second;

s23, according to the simulated speed and the simulated acceleration of the vehicle after the vehicle is followed and the actually measured speed and acceleration of the vehicle before the vehicle is followed, the speed and the acceleration of the vehicle after the vehicle is followed in one second are circularly simulated;

s3, optimizing the following model parameters by utilizing a genetic algorithm to obtain a model optimal parameter solution; in the step S3, the optimizing the following model parameters by using a genetic algorithm includes:

s31, parameter selection and threshold setting: selecting parameters and a value range which need to be optimized in a following model;

s32, group binary coding: converting the optimized parameter values into binary codes;

s33, determining the number of initial populations: the number of individuals whose selected optimization parameters are randomly composed within the parameter range;

s34, calculating a population fitness function: i.e. optimizing the evaluation function of the target;

s35, selecting: according to the principle of' object racing for the day and survival of the fittest, selecting a part of optimal individuals as a next generation population, wherein the smaller the MSE value is, the closer the VSP distribution of the simulated vehicle and the actually measured vehicle is, and the larger the probability that the following model parameter corresponding to the smaller MSE value is reserved in the next generation is;

s36, crossing and mutation operation: the optimized parameters are binary codes, and like a chromosome, the crossover and mutation operations of the chromosome are equivalent to the increase of the randomness of the next generation of individuals;

s37, maximum iteration times: the model is automatically stopped when iteration reaches the set termination times, and the optimal solution of the model is output;

s38, convergence tolerance: when the error of the optimal solution output by the model at the offspring and the father is smaller than a set value, the model automatically stops, and an optimal parameter solution of the model is output;

s4, evaluating effects based on vehicle speed, acceleration and VSP models: substituting the model optimal parameter solution obtained in the step S3 into the step S2 to obtain a time-speed, time-acceleration track graph and a VSP distribution graph after parameter optimization;

s5, calculating the actual vehicle emission and the vehicle emission of the following models before and after parameter optimization based on the speed-VSP emission model effect evaluation, and performing comparative analysis and verification.

2. A method for calculating emissions according to claim 1, wherein in said S2,

the calculation formula of the specific power VSP of the motor vehicle is as follows:

VSP＝v(1.1a+0.132)+0.000302v ³

wherein v is velocity and a is acceleration;

the calculation formula of the simulation and actual measurement mean square error MSE is as follows:

wherein: s is S _i S _i Simulating a VSP for the ith second; t (T) _i T _i VSP was measured for the ith second.