High-precision driving cycle construction method based on improved elitist genetic algorithm

Abstract

The application discloses a high-precision driving cycle construction method based on an improved elite genetic algorithm, relates to the technical field of vehicle energy consumption prediction, and specifically comprises the following steps: collecting vehicle driving parameters to form an original data set; based on the original data set, performing grid division on speed and acceleration and marking state IDs; counting the number of state ID transitions of adjacent sampling time points to construct a state transition matrix; performing random walk diffusion simulation on the state transition matrix to construct a statistical initial population; inputting the statistical initial population into an improved elite genetic model for iteration; screening a global optimal solution in the optimal individuals in the iteration process to obtain a driving cycle with the smallest deviation of target characteristics from a real working condition. Therefore, the problems that the operation data of new energy buses are often lost due to communication failures, that false acceleration is introduced by linear interpolation, and that oscillation errors are easily generated in the low-speed section by cubic spline interpolation are solved.

Description

Technical Field

[0001] This invention relates to the field of vehicle energy consumption prediction technology, and in particular to a high-precision driving cycle construction method based on an improved elite genetic algorithm. Background Technology

[0002] The intensifying global climate change has made transportation a key area for carbon emission control. According to the International Energy Agency, in 2023, the transportation sector accounted for 24% of global carbon emissions, with road vehicles contributing over 70% of that. Accurately assessing vehicle energy consumption and emissions, and developing effective emission reduction strategies, has become a top priority. Driving cycle data, as speed-time curves depicting the actual operating state of a vehicle, is a core tool connecting vehicle design, energy consumption simulation, and emissions assessment. It not only provides standard input for dynamometer testing, directly affecting the test results of key indicators such as fuel economy and driving range, but also forms the basis for the development of emission regulations such as the EU's WLTP (Worldwide Harmonized Test Cycle for Light Vehicles) and China's CLTC (China Light Vehicles Test Cycle). For electric and hybrid vehicles, driving cycle data also determines the optimization direction of energy management strategies, such as improving the efficiency of regenerative braking systems and accurately predicting battery driving range.

[0003] In existing technologies, the isolated analysis of single factors leads to a lack of coupled modeling of geography, transportation, and energy. For example, urban public transport operating conditions need to consider stop spacing, passenger capacity, and real-time traffic signals simultaneously, but current methods can only partially incorporate these variables. The range reduction of electric vehicles is affected by the interaction of temperature, battery aging, and operating condition type; existing operating conditions do not establish a multi-factor coupled energy consumption model. Secondly, due to limitations in data quality and processing technology, operational data of new energy buses is often missing due to communication failures. While linear interpolation is simple, it introduces spurious accelerations, and cubic spline interpolation is prone to oscillation errors at low speeds. Although hybrid interpolation methods combine linear and spline interpolation, they can control the operating condition characteristic error within 10%, but they do not solve the data synchronization problem. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a high-precision driving cycle construction method based on an improved elite genetic algorithm. The core of this algorithm lies in upgrading the inefficient single-point mutation algorithm in traditional genetic algorithms to a segment reconstruction algorithm based on Markov reachable domains. A two-stage elite selection strategy is designed, constructing a new, highly efficient genetic algorithm framework. This framework achieves global exploration through large-scale crossover and performs deep local optimization on elite individuals, thereby efficiently balancing the algorithm's crossover and mutation capabilities.

[0005] The high-precision driving cycle construction method based on an improved elite genetic algorithm provided by this invention includes the following steps: S10: Collect vehicle driving parameters to form the raw dataset; S20: Based on the original dataset, divide the velocity and acceleration into grids and label the state IDs; S30: Count the number of state ID transitions between adjacent sampling times and construct a state transition matrix; S40: Perform random walk diffusion simulation on the state transition matrix to construct a statistical initial population; S50: Input the statistical initial population into the improved elite genetic model for iteration; S60: Select the globally optimal solution from the optimal individuals in the iterative process to obtain the driving cycle with the smallest deviation from the target characteristics of the actual working conditions.

[0006] Furthermore, the specific steps for collecting vehicle driving parameters and forming the raw dataset are as follows: Real-time speed, location, and acceleration parameters of vehicles in highway scenarios are collected using the Global Positioning System and on-board diagnostic system, with a cumulative collection time of no less than 20 hours.

[0007] Furthermore, based on the original dataset, the specific steps for dividing the velocity and acceleration into grids and labeling the state IDs are as follows: The velocity and acceleration resolutions are meshed with unit step sizes to obtain a discrete mesh node sequence for the velocity. Discrete grid node sequence of acceleration ; in, The unit step size of the speed, The lower limit of the velocity for discrete meshes. The upper limit of the velocity of the discrete mesh. For the unit step size of acceleration, The lower limit of acceleration for discrete grids. This represents the upper limit of acceleration for the discrete grid. For any sample time t, the vehicle state ( , If satisfied and Then mark this state as a unique state ID. ; All (x, y) grid combinations constitute a finite set of states containing n elements. .

[0008] Furthermore, the specific steps for constructing the state transition matrix by counting the number of state ID transitions at adjacent sampling times are as follows: Statistical analysis of the state IDs of two adjacent sampling times from Transferred to The total number of times appears is denoted as . ; Based on each state Calculation by Transfer to Given the transition probabilities, construct the state transition matrix P. Where k is the index variable of the finite set of states.

[0009] Furthermore, the specific steps for constructing the statistical initial population by performing random walk diffusion simulation on the state transition matrix are as follows: Starting from a state where both velocity and acceleration are 0, a random walk simulation is performed based on the state transition matrix P to generate several state ID sequences. Each state ID sequence corresponds to a candidate driving cycle C, and all candidate driving cycles are used as the initial population.

[0010] Furthermore, the specific steps for constructing the statistical initial population by performing random walk diffusion simulation on the state transition matrix are as follows: Starting from a state where both velocity and acceleration are 0, a random walk simulation is performed based on the state transition matrix P to generate several state ID sequences. Each state ID sequence corresponds to a candidate driving cycle C, and all candidate driving cycles are used as the initial population.

[0011] Furthermore, the steps for constructing an improved elite genetic model are as follows: Optimize any candidate driving loop C, and optimize the objective function. The sum of the relative deviations between the key feature parameters of candidate driving cycle C and the target feature values ​​under actual operating conditions, i.e. in, It is the calculated value of the i-th feature parameter of the optimized candidate driving cycle. is the target feature value under actual working conditions, and h is the number of key feature parameters; From the initial population, parent candidate drivers are selected in a cycle, and a crossover operation is performed based on the state transition matrix P to generate a crossover pool; Calculate the optimization objective function value of all candidate driving cycles in the cross pool, and select the candidate driving cycles with the highest optimization objective function value as the elite pool; Reconstruct the mutation operators for all candidate driving loops in the elite pool to generate a mutation pool; Merge the crossover pool and the mutation pool, calculate the optimization objective function value of all candidate driving cycles after merging, and continue to select the candidate driving cycles with the highest optimization objective function value as the new generation population; Repeat the crossover and mutation process until the preset number of iterations is reached.

[0012] Furthermore, the specific steps for selecting the globally optimal solution from the optimal individuals in the iterative process to obtain the driving cycle with the smallest deviation from the target characteristics of the actual working conditions are as follows: The optimal individual with the smallest objective function value in each generation of the population is selected. Then, the optimal solution with the smallest objective function value is selected from all the optimal individuals and used as the optimal driving cycle.

[0013] Furthermore, the specific steps for reconstructing the mutation operator for all candidate driving loops in the elite pool are as follows: For a segment of a candidate driving cycle of length L to be reconstructed, let the initial state be... The termination status is Then the state at any intermediate time m in the segment to be reconstructed Must belong to the dispersal domain at the same time and convergence region ; The recursive definition of the divergence domain is as follows: The recursive definition of the region of convergence is: Then the reachable region at time m That is, the intersection of the divergence region and the convergence region: in, The state transition matrix P consists of states... arrive The transition probability.

[0014] In summary, the present invention has at least the following beneficial effects: 1. This method significantly improves the exploration capability and convergence performance of the mutation stage. The algorithm upgrades the inefficient single-point perturbation of the traditional GA mutation stage to a medium-step optimization, thereby reducing the dependence on the initial population diversity and effectively suppressing the phenomenon of prematurely falling into local optima due to insufficient population diversity, ensuring that the optimization process can continuously approach the global optimum.

[0015] 2. This method uses 16 key feature parameters to quantitatively evaluate the constructed highway driving cycle. The results show that the average relative deviation between the constructed driving condition and the measured data is only 0.59%. This result verifies the effectiveness of the method in accurately reproducing the real driving cycle. The fuel consumption deviation of the constructed driving cycle is as low as 6.24%. This result demonstrates that the multi-dimensional feature high-precision fitting framework achieved by this method ensures a high-reliability assessment of vehicle energy consumption. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a three-dimensional probability distribution diagram of velocity-acceleration according to the present invention; Figure 2 This is a two-dimensional probability distribution diagram of velocity-acceleration according to the present invention; Figure 3 This is the Markov state transition probability matrix diagram of the present invention; Figure 4 This is a schematic diagram of the 100 working condition features with the optimal initial population statistical bias in this invention. Figure 5 This is a schematic diagram of the bidirectional / unidirectional intersection of the working condition segment based on the transition probability constraint of the present invention; Figure 6 This is a schematic diagram of the reachable domain concept and intermediate state constraints of the segment to be reconstructed in this invention; Figure 7 This is a flowchart of the steps of the present invention; Figure 8 This is the iterative convergence curve of the elite genetic algorithm of this invention; Figure 9 This is the speed-time curve of the optimal typical driving cycle of the present invention; Figure 10 This is a three-dimensional deviation diagram of the 16 characteristic parameters of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The following is in conjunction with the appendix Figures 1 to 10 The present invention will be described in further detail below.

[0020] like Figure 7 As shown, this invention provides a high-precision driving cycle construction method based on an improved elite genetic algorithm, comprising: Data collection and analysis: The truck used for data collection is equipped with a GPS system and an on-board diagnostic system, and travels on the Shandong-Guangdong Expressway. It synchronously records the vehicle's real-time speed, location, acceleration, and other information at a frequency of 1Hz, accumulating a total of 60 hours of raw data to ensure data representativeness.

[0021] To construct a typical driving condition that accurately reflects actual driving behavior, statistical analysis was performed on the raw data using 16 key characteristic parameters. These parameters comprehensively describe the macroscopic and microscopic characteristics of the driving cycle. The specific characteristic parameters and their statistical results in this highway dataset are shown in Table 1. As can be seen from Table 1, the idling time proportion is extremely low in highway driving conditions, while the proportion of constant speed time is very high, significantly different from typical urban driving conditions. Furthermore, the absolute values ​​of average acceleration and deceleration are relatively small, reflecting the relatively gentle speed changes during high-speed driving.

[0022] Table 1. Characteristic parameter set of driving cycle To more intuitively understand the data distribution, the velocity-acceleration state is spatially discretized, and its distribution is plotted. For example... Figure 1 and Figure 2 As shown, unlike urban driving conditions, the probability distribution of high-speed driving conditions is significantly concentrated in the high-speed, low-acceleration region. Most data points are distributed in the range of speeds greater than 50 km / h, and the absolute values ​​of acceleration and deceleration fluctuate within a relatively small range.

[0023] Pre-construction of the candidate test case pool: Data from each actual working condition Each state is labeled according to its grid division. Among these, speed resolution is used to balance accuracy and computational efficiency. Set to 1 Acceleration resolution Set to 0.1 To determine the true speed range Divide the data into several grids to obtain a discrete grid of velocity nodes. ,in, and These represent the upper and lower bounds of velocity for the discrete mesh. Similarly, the acceleration range... use Divide into equal parts to obtain a discrete grid of acceleration nodes. ,in, and For the upper and lower bounds of acceleration in a discrete grid. When Then the state at that moment is recorded as .in, A single state ID is assigned to each row and column, thus representing all possible grid combinations. Construct a finite set of states containing n elements .

[0024] Statistical analysis of the state IDs of two adjacent sampling times from Transferred to The total number of times appears is denoted as . ,For example: in, This is an indicator function; if the condition inside the parentheses is satisfied, it is recorded as 1, otherwise it is recorded as 0. This represents the total sampling time.

[0025] Based on the transition count, for each state It can be obtained that it is transferred to Markov transition probabilities, for example: matrix This is the Markov state transition matrix for the actual operating conditions, as shown in the example below. Figure 3 As shown.

[0026] Generation of the initial population based on the Markov Monte Carlo (MCMC) method: Starting from a state where both velocity and acceleration are 0, a Monte Carlo random walk is performed using the state transition matrix P to generate 10,000 state ID sequences of 1800 seconds each, serving as candidate driving cycle sequences. These sequences initially possess the statistical characteristics of real-world driving conditions, forming the initial population for subsequent genetic algorithms.

[0027] To minimize the characteristic deviation between the constructed driving cycle and the actual operating conditions, a specific objective function Obj(C) is defined here. The objective function value for any candidate driving cycle C is calculated as follows: in, This is the calculated value of the i-th feature parameter in this loop. The target feature values ​​are statistically obtained from the parameters in Table 1, where h is the number of key feature parameters, h=16. The entire iterative process of the algorithm is to find the C that minimizes Obj(C).

[0028] like Figure 4 The 100 operating conditions with the smallest statistical bias in the initial population are shown.

[0029] Improvements to the Elite Genetic Algorithm: It should be noted that although the standard genetic algorithm has global search capabilities, when dealing with high-dimensional, strongly constrained driving loop construction problems, the mutation stage can only mutate the state in one place. This makes the algorithm highly dependent on the diversity of the working condition segments in the initial population, resulting in slow convergence speed and easy getting trapped in local optima.

[0030] To address the mismatch in importance between the crossover and mutation phases, this invention proposes an improved elite genetic algorithm after obtaining an optimized initial population. The core of this algorithm lies in its efficient balancing of the crossover and mutation strategies through a two-stage elite selection strategy.

[0031] The improved elite genetic algorithm framework includes: determining the fitness function and clearly defining several quantitative indicators to evaluate the performance of a driving cycle; a two-stage elite selection strategy, an efficient iterative mechanism, intelligent allocation of computing resources, and a framework that balances exploration and utilization; and a mutation operator based on reachability domain reconstruction, which is the core innovation of the algorithm, replacing the inefficient single-point mutation of traditional genetic algorithms with highly efficient and highly exploratory fragment-level optimization.

[0032] In the improved elite genetic algorithm, the crossover operation follows the transition constraints of Markov chains, meaning that after the exchange and recombination of segments between two parent individuals, the transition at their connection point must be a valid transition with a non-zero probability. For example... Figure 5 As shown, for the two segments and If in and If a crossover occurs at a certain point, the transition probability must be satisfied. At this point, a one-way crossing can be achieved if the following conditions are met simultaneously: This enables bidirectional crossing of state nodes.

[0033] It should be noted that the core of the mutation operator is to ensure that the mutated node conforms to the Markov state transition probability matrix. Although the traditional single-point mutation satisfies this premise, it suffers from high cost and low return, resulting in inefficiency.

[0034] This invention solves this problem through reachability domains. For example... Figure 6 As shown, for a length of The fragment to be reconstructed, its initial state is The termination status is At any intermediate time m in the segment ( ) state Two conditions must be met simultaneously: (1) can be obtained from Depart, and after m steps transfer to the destination.

[0035] (2) One can start from it and pass through m-step transfer to Da .

[0036] The set of states satisfying condition (1) is called the dispersion domain, denoted as The recursive definition is: The set of states satisfying condition (2) is the region of convergence, denoted as . The recursive definition is: Then the reachable region at time m That is, the intersection of the divergence region and the convergence region: in, The state transition matrix P consists of states... arrive The transition probability. The reachable domain provides a strict physical boundary for the entire fragment generation process, ensuring that all newly generated fragments conform to the Markov transition law statistically derived from real data.

[0037] Phase 1 Crossover: Select parent candidate driving loops from the initial population or the previous generation of iterative population. Perform a large-scale crossover operation based on the Markov state transition matrix P, which follows Markov transition constraints. That is, the state transition probability at the connection point after crossover is not zero, ensuring that the sequence after crossover conforms to the true state transition law, and generate a crossover pool. Each individual in the cross pool is a candidate driving cycle that has been recombined and optimized.

[0038] Phase Two Elite Mutation: Computing the Cross-Pool The objective function value Obj(C) of all candidate driving cycles is used to select the K individuals with the highest fitness to form an elite pool. For the elite pool In each candidate driving loop, a segment reconstruction mutation operator based on Markov reachability domains is applied to optimize the reconstruction of continuous segments from 30 seconds to 100 seconds in the loop, generating a mutation pool. During the mutation process, each mutated fragment generates 1000 candidate paths. The optimal path is selected based on the objective function value to ensure that the mutated individual still conforms to the characteristics of the actual working conditions.

[0039] New generation population generation: Merging crossover pools and mutation pools to generate a merged pool. , Calculate the merge pool Given the objective function value Obj(C) of all candidate driving cycles, select the R individuals with the optimal objective function value to form the next generation population. Repeat the iteration until the preset number of iterations is met.

[0040] Select the globally optimal solution from the optimal individuals of all iterations. The global optimal solution The corresponding candidate driving cycle is the highway driving cycle with the smallest deviation from the characteristics of real working conditions.

[0041] To verify the effectiveness of the improved elite genetic algorithm proposed in this invention in constructing highway driving cycles, a comprehensive analysis was conducted on the algorithm's convergence process and the accuracy of the construction results. The initial candidate pool size was 2000 individuals. In each iteration of the elite genetic algorithm, the crossover pool size was set to 5000, from which the 100 elite individuals with the highest fitness were selected for mutation. The length of the mutated fragment was dynamically selected between 30 and 100 seconds. During mutation, each fragment generated 1000 candidate paths for sorting and selection. The total number of iterations was set to 500 generations.

[0042] Figure 8 The graph illustrates the changes in the objective function, i.e., the sum of the biases of all feature parameters, during 10 algorithm iterations. As can be seen from the graph, the algorithm exhibits excellent convergence performance. In the early stages of iteration, the mean of the sum of biases rapidly decreases from over 300% to below 50%, indicating that the algorithm can quickly locate high-potential solution spaces. As iterations progress, the quality of the solutions steadily improves and converges after approximately 400 iterations, eventually stabilizing at an average total bias of around 10%. Simultaneously, the minimum or maximum range of biases in the graph also shrinks and converges, demonstrating that the algorithm not only converges quickly but also provides stable results and exhibits strong robustness.

[0043] Through deep optimization using an elite genetic algorithm, a series of candidate driving loops that highly fit the characteristics of actual data were finally obtained. Figure 9 The speed-time curves for 10 optimal typical driving cycles are presented. These curves exhibit typical characteristics of real high-speed driving conditions in both macro trends and micro details, such as long-term high-speed cruising, small speed fluctuations caused by changes in traffic flow, and relatively smooth acceleration and deceleration processes, intuitively demonstrating the effectiveness of the constructed driving conditions.

[0044] As a key evaluation indicator, the ultimate goal of the operating condition is to serve energy consumption and emissions assessment. To ensure the accuracy and comparability of the energy consumption assessment results, this invention is based on a light truck model built into the AVL-Cruise simulation platform, with the light truck having a mass of 4495 kg. As shown in Table 2, the fuel consumption simulation value based on the constructed operating condition deviates by only 6.24% compared to the fuel consumption simulation value based on actual driving data, fully demonstrating the accuracy and engineering practical value of this method in energy consumption assessment applications.

[0045] Table 2 shows the simulated fuel consumption values ​​for the constructed operating conditions and the simulated fuel consumption values ​​from actual driving data. Figure 10 The 3D bar chart provides a more intuitive view of the deviation distribution of various characteristic parameters in the results of 10 independent trials. The chart clearly shows that, except for a few parameters, such as the average acceleration / deceleration which exhibits slight fluctuations in relative deviation due to its small absolute value, the deviation values ​​of most parameters in different trials are stably concentrated at an extremely low level of less than 2%, statistically demonstrating the stability and high accuracy of the elite genetic algorithm.

[0046] The above are merely preferred embodiments of the invention and are not intended to limit the invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the protection scope of the invention.

Claims

1. A high-precision driving loop construction method based on an improved elite genetic algorithm, characterized in that, Includes the following steps: S10: Collect vehicle driving parameters to form the raw dataset; S20: Based on the original dataset, divide the velocity and acceleration into grids and label the state IDs; S30: Count the number of state ID transitions between adjacent sampling times and construct a state transition matrix; S40: Perform random walk diffusion simulation on the state transition matrix to construct a statistical initial population; S50: Input the statistical initial population into the improved elite genetic model for iteration; S60: Select the globally optimal solution from the optimal individuals in the iterative process to obtain the driving cycle with the smallest deviation from the target characteristics of the actual working conditions.

2. The high-precision driving loop construction method based on the improved elite genetic algorithm according to claim 1, characterized in that, The specific steps for collecting vehicle driving parameters and forming the raw dataset are as follows: Real-time speed, location, and acceleration parameters of vehicles in highway scenarios are collected using the Global Positioning System and on-board diagnostic system, with a cumulative collection time of no less than 20 hours.

3. The high-precision driving loop construction method based on the improved elite genetic algorithm according to claim 1, characterized in that, Based on the original dataset, the specific steps for dividing the velocity and acceleration into grids and labeling the state IDs are as follows: The velocity and acceleration resolutions are meshed with unit step sizes to obtain a discrete mesh node sequence for the velocity. Discrete grid node sequence of acceleration ; in, The unit step size of the speed. The lower limit of the velocity for discrete meshes. The upper limit of the velocity of the discrete mesh. For the unit step size of acceleration, The lower limit of acceleration for discrete grids. This represents the upper limit of acceleration for the discrete grid. For any sampling time t, the vehicle state ( , If satisfied and Then mark this state as a unique state ID. ; All (x, y) grid combinations constitute a finite set of states containing n elements. .

4. The high-precision driving loop construction method based on the improved elite genetic algorithm according to claim 3, characterized in that, The specific steps for constructing the state transition matrix by counting the number of state ID transitions between adjacent sampling times are as follows: Statistical analysis of the state IDs of two adjacent sampling times from Transferred to The total number of occurrences of is denoted as . ; Based on each state Calculation by Transfer to Given the transition probabilities, construct the state transition matrix P. Where k is the index variable of the finite set of states.

5. The high-precision driving loop construction method based on the improved elite genetic algorithm according to claim 4, characterized in that, The specific steps for constructing a statistical initial population by performing a random walk diffusion simulation on the state transition matrix are as follows: Starting from a state where both velocity and acceleration are 0, a random walk simulation is performed based on the state transition matrix P to generate several state ID sequences. Each state ID sequence corresponds to a candidate driving cycle C, and all candidate driving cycles are used as the initial population.

6. The high-precision driving loop construction method based on the improved elite genetic algorithm according to claim 5, characterized in that, The steps for constructing an improved elite genetic model are as follows: Optimize any candidate driving loop C, and optimize the objective function. The sum of the relative deviations between the key feature parameters of candidate driving cycle C and the target feature values ​​under actual operating conditions, i.e. in, It is the calculated value of the i-th feature parameter of the optimized candidate driving cycle. is the target feature value under actual working conditions, and h is the number of key feature parameters; From the initial population, parent candidate drivers are selected in a cycle, and a crossover operation is performed based on the state transition matrix P to generate a crossover pool; Calculate the optimization objective function value of all candidate driving cycles in the cross pool, and select the candidate driving cycles with the highest optimization objective function value as the elite pool; Reconstruct the mutation operators for all candidate driving loops in the elite pool to generate a mutation pool; Merge the crossover pool and the mutation pool, calculate the optimization objective function value of all candidate driving cycles after merging, and continue to select the candidate driving cycles with the highest optimization objective function value as the new generation population; Repeat the crossover and mutation process until the preset number of iterations is reached.

7. The high-precision driving loop construction method based on the improved elite genetic algorithm according to claim 6, characterized in that, The specific steps for selecting the globally optimal solution from the optimal individuals in the iterative process to obtain the driving cycle with the smallest deviation from the target characteristics of the actual working conditions are as follows: The optimal individual with the smallest objective function value in each generation of the population is selected. Then, the optimal solution with the smallest objective function value is selected from all the optimal individuals and used as the optimal driving cycle.

8. The high-precision driving loop construction method based on the improved elite genetic algorithm according to claim 6, characterized in that, The specific steps for reconstructing the mutation operators for all candidate driving loops in the elite pool are as follows: For a segment of a candidate driving cycle of length L to be reconstructed, let the initial state be... The termination status is Then the state at any intermediate time m in the segment to be reconstructed Must belong to the dispersal domain at the same time and convergence region ; The recursive definition of the divergence domain is as follows: The recursive definition of the region of convergence is: Then the reachable region at time m That is, the intersection of the divergence region and the convergence region: in, The state transition matrix P consists of states... arrive The transition probability.

Images