Intelligent optimization method for electric truck transmission gear ratio

Abstract

The application relates to the technical field of new energy automobile power system design and optimization, and discloses an intelligent optimization method for transmission speed ratio of an electric truck, which comprises the following steps: 1.1. According to the whole vehicle parameters, performance indexes and working condition load characteristics of the electric truck, the power, torque and rotating speed of the driving motor are matched and parameter configured; 1.2. Based on the whole vehicle energy consumption characteristics and the endurance mileage demand, the energy, capacity and voltage parameters of the power battery are optimized and configured; 1.3. The constraint conditions that the transmission speed ratio needs to meet are analyzed, including the maximum vehicle speed, climbing ability and acceleration performance. The transmission speed ratio of the electric truck is optimized through a multi-objective evolutionary algorithm improved by a deep Q network, the power performance and economic performance indexes can be comprehensively considered, and multi-objective optimization of the transmission speed ratio parameters is realized. Compared with the traditional optimization method, the application can not only effectively balance the power performance indexes such as the acceleration time and the climbing ability, but also significantly reduce the energy consumption per 100 kilometers, and improve the economy of the whole vehicle.

Description

Technical Field

[0001] This invention relates to the field of powertrain design and optimization technology for new energy vehicles, specifically to an intelligent optimization method for the transmission ratio of electric trucks. Background Technology

[0002] In the design and application of electric trucks, optimizing the transmission ratio is one of the key aspects of improving vehicle performance. Traditional methods mainly rely on empirical formulas and trial-and-error approaches, which have significant limitations in practical applications. Empirical formulas are usually based on past design experience and experimental data. While they can guide the initial design of transmission ratios to some extent, they often lack comprehensive consideration of complex operating conditions and multi-objective optimization problems. For example, when determining the transmission ratios for each gear, traditional methods may only consider a single performance indicator, such as maximum speed or maximum climbing ability, while neglecting the balance of other important indicators. Furthermore, trial-and-error methods require repeated parameter adjustments and numerous experiments to verify the design's rationality. This method is not only time-consuming and labor-intensive but also demands a high level of experience from the designer, making precise optimization difficult.

[0003] In recent years, with the development of computing technology, some methods based on mathematical optimization models have been introduced into transmission speed ratio optimization. These methods establish objective functions and constraints, and then use mathematical algorithms to solve for the optimal solution. However, when dealing with large-scale complex problems, the optimization efficiency and stability of the results are still not ideal. For example, some studies have attempted to optimize transmission speed ratios using genetic algorithms or particle swarm optimization algorithms. Although these methods have improved optimization efficiency to some extent, their global search capabilities and convergence speed still need improvement. Furthermore, existing methods also exhibit shortcomings in dynamic adjustment capabilities and adaptability, making it difficult to quickly respond to changes in operating conditions and make dynamic adjustments.

[0004] To address this issue, those skilled in the art have proposed an intelligent optimization method for the transmission ratio of electric trucks. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent optimization method for the transmission ratio of electric trucks, which solves the problems of existing electric truck transmission ratio optimization methods in terms of multi-objective balance, dynamic adaptability, and optimization efficiency.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent optimization method for the transmission ratio of an electric truck, comprising the following steps:

[0007] 1.1 Match and configure the power, torque, and speed of the drive motor according to the overall vehicle parameters, performance indicators, and working load characteristics of the electric truck;

[0008] 1.2. Based on the vehicle's energy consumption characteristics and range requirements, the energy, capacity, and voltage parameters of the power battery are optimized.

[0009] 1.3 Analyze the constraints that the transmission ratio must meet, including maximum vehicle speed, climbing ability, and acceleration performance;

[0010] 1.4 Construct a multi-objective optimization model for transmission ratio parameters with power and economic indicators as optimization objectives;

[0011] 1.5. Encode the transmission ratio parameters into binary form and design corresponding decoding and initialization strategies;

[0012] 1.6 Construct and train a deep Q-network agent, and design the state space, action space, reward function, selection policy, and neural network architecture;

[0013] 1.7 Introduce a population-individual collaborative restart mechanism to avoid the algorithm getting trapped in local optima;

[0014] 1.8 Construct a multi-objective evolutionary algorithm improved by deep Q-network to solve for the optimal combination of transmission ratio parameters for electric trucks;

[0015] The optimization results were verified by simulation to ensure that they meet the performance requirements of electric trucks under complex working conditions.

[0016] Preferably, in step 1.1, the power matching of the drive motor must simultaneously meet the requirements of maximum vehicle speed, maximum climbing ability, and acceleration performance. Its maximum output power must achieve an overall balance under multiple performance constraints. The specific calculation formula is as follows:

[0017] Determine the maximum output power based on the top vehicle speed:

[0018]

[0019] Determine the maximum output power based on the maximum climbing ability:

[0020]

[0021] Determine the maximum output power based on acceleration performance:

[0022]

[0023] Where g represents gravitational acceleration, v max This represents the preset maximum speed, θ represents the car's climbing angle, and v grade The constant speed of the car climbing the hill is represented by 'a', and the target acceleration is represented by 'v'. acc F represents the average velocity during the acceleration time. roll F represents rolling resistance.air η represents air resistance. mech This indicates the efficiency of mechanical transmission.

[0024] Preferably, in step 1.2, the energy design of the power battery needs to meet the maximum range target of the vehicle. Its energy consumption includes the energy required for motor drive, the energy required for vehicle electronic components, and the energy lost during discharge. The specific calculation formula is as follows:

[0025] E battery =E motor +E electronics +E loss

[0026] in,

[0027]

[0028] E electronics =α×E battery

[0029] E loss =(1-η discharge )×E battery

[0030] P motor D represents the power required to drive the motor, and η represents the driving range. motor η represents the motor efficiency. discharge α represents the discharge efficiency, and α represents the energy consumption ratio coefficient of automotive electronic components.

[0031] Preferably, in step 1.3, the constraints that the transmission ratio must satisfy include:

[0032] The maximum gear ratio must be sufficient for the vehicle to operate under the maximum gradient conditions:

[0033]

[0034] The minimum gear ratio must be sufficient for the vehicle to travel at its maximum speed:

[0035]

[0036] Among them, F grade_max T represents the maximum climbing resistance. motor_max The value r represents the maximum output torque of the motor. wheel V represents the wheel radius. grade_max F represents the maximum climbing speed. roll_max and F air_max These represent rolling resistance and air resistance at the maximum vehicle speed, respectively.

[0037] Preferably, in step 1.4, the multi-objective optimization model for the transmission speed ratio parameter, with power and economic indicators as optimization goals, uses a weighted method to normalize the objectives, as shown in the following formula:

[0038] F=ω×F performance +(1-ω)×F economy

[0039] Where ω is the weighting factor, F performance As a dynamic indicator, F economy For economic indicators, the performance indicators are mainly based on acceleration time, while the economic indicators are mainly based on energy consumption per 100 kilometers.

[0040] Preferably, in step 1.5, the transmission ratio parameters are encoded in binary. Each gear ratio is encoded using a 10-bit binary vector. The entire optimization variable is a binary string of length 10. The decoding process converts the binary string into the corresponding real-valued combination of the transmission ratio parameters.

[0041] Preferably, in step 1.6, the state space of the deep Q-network agent includes gear ratio, main reducer ratio, acceleration time, and energy consumption per 100 kilometers, and the action space includes mutation operations for different gear ratio parameters; the reward function determines the reward value based on whether the acceleration time and energy consumption per 100 kilometers of the new solution are better than the solution to be optimized.

[0042] Preferably, in step 1.7, the population individual collaborative restart mechanism determines whether it has fallen into a local optimum by checking the number of consecutive times an individual in the population has not been updated, and regenerates a new individual to replace the original individual through population collaborative operation.

[0043] Preferably, in step 1.8, the multi-objective evolutionary algorithm improved by deep Q-network improves the global search capability and convergence efficiency of the algorithm by dynamically adjusting the evolutionary strategy, and finally solves the optimal combination of electric truck transmission ratio parameters.

[0044] This invention provides an intelligent optimization method for the transmission ratio of electric trucks. It has the following beneficial effects:

[0045] 1. This invention optimizes the transmission ratio of electric trucks using a multi-objective evolutionary algorithm improved by deep Q-networks. It comprehensively considers both power and economic indicators, achieving multi-objective optimization of the transmission ratio parameters. Compared with traditional optimization methods, this invention not only effectively balances power indicators such as acceleration time, maximum speed, and climbing ability, but also significantly reduces energy consumption per 100 kilometers, improving the overall vehicle economy. This optimization method meets the performance requirements of electric trucks under complex operating conditions while greatly improving the vehicle's range and energy efficiency, providing strong support for the widespread application of electric trucks.

[0046] 2. This invention introduces a population-based collaborative restart mechanism, effectively avoiding the problem of the algorithm getting trapped in local optima during the optimization process. By dynamically adjusting the evolutionary strategy, the algorithm's global search capability and convergence efficiency are improved, ensuring that a globally optimal or near-optimal solution can be found quickly in large-scale complex problems. This not only improves the stability and reliability of the optimization results but also significantly shortens the optimization computation time, enhancing the efficiency and practicality of optimizing the transmission ratio of electric trucks.

[0047] 3. This invention employs a combination of binary encoding and a deep Q-network agent to achieve efficient encoding and decoding of transmission ratio parameters, simplifying the computational complexity of the optimization process. Simultaneously, by designing a reasonable state space, action space, and reward function, the agent can autonomously learn and make decisions under complex operating conditions, dynamically adjusting the transmission ratio parameters. This method not only reduces human intervention and improves the automation level of optimization but also enhances the adaptability and flexibility of electric trucks under different operating conditions, further improving the overall performance of the vehicle. Attached Figure Description

[0048] Figure 1 This is the overall flowchart of the present invention;

[0049] Figure 2 This is a flowchart of the intelligent optimization algorithm of the present invention. Detailed Implementation

[0050] The technical solutions in 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.

[0051] Please see the appendix Figure 1 - Appendix Figure 2 This invention provides an intelligent optimization method for the transmission ratio of an electric truck, comprising the following steps:

[0052] Based on the overall vehicle parameters, performance indicators, and working load characteristics of the electric truck, the power, torque, and speed of the drive motor are matched and configured.

[0053] The power matching of the drive motor must simultaneously meet the requirements of maximum vehicle speed, maximum climbing ability, and acceleration performance. Its maximum output power needs to achieve an overall balance under multiple performance constraints. The specific calculation formula is as follows:

[0054] Determine the maximum output power based on the top vehicle speed:

[0055]

[0056] Determine the maximum output power based on the maximum climbing ability:

[0057]

[0058] Determine the maximum output power based on acceleration performance:

[0059]

[0060] Where g represents gravitational acceleration, v max This represents the preset maximum speed, θ represents the car's climbing angle, and v grade The constant speed of the car climbing the hill is represented by 'a', and the target acceleration is represented by 'v'. acc F represents the average velocity during the acceleration time. roll F represents rolling resistance. air η represents air resistance. mech This indicates the efficiency of mechanical transmission.

[0061] Generally, the maximum power of a motor is 2 to 3 times its rated power. This depends on the motor's overload coefficient λ (taken as 3 in this article). A larger coefficient indicates stronger overload power, allowing the drive motor to operate at a higher permissible power. Conversely, if the power is too low, it will affect the vehicle's power performance. The relationship is as follows:

[0062]

[0063] Based on the vehicle's maximum speed and the motor's power, and considering a certain redundancy, the motor is selected in conjunction with economic indicators. Assume the maximum speed of the selected motor is n. max The relationship between the rated speed and the maximum speed of the motor is as follows:

[0064]

[0065] Where, n e This represents the rated speed; typically, 2 < β < 4. Once the rated power and rated speed of the motor are determined, the rated torque can be calculated, and the relationship is as follows:

[0066]

[0067] T max =λT e

[0068] Among them, T e T represents the rated torque of the motor. max This indicates the maximum torque of the motor.

[0069] The energy design of the power battery must meet the maximum range target of the entire vehicle. Its energy consumption includes the energy required for motor drive, the energy required for vehicle electronic components, and the energy lost during discharge. The specific calculation formula is as follows:

[0070] E battery =E motor +E electronics +E loss

[0071] in,

[0072]

[0073] E electronics =α×E battery

[0074] E loss =(1-η discharge )×E battery

[0075] P motor D represents the power required to drive the motor, and η represents the driving range. motor η represents the motor efficiency. discharge α represents the discharge efficiency, and α represents the energy consumption ratio coefficient of automotive electronic components.

[0076] The constraints that the transmission speed ratio must meet include:

[0077] The maximum gear ratio must be sufficient for the vehicle to operate under the maximum gradient conditions:

[0078]

[0079] The minimum gear ratio must be sufficient for the vehicle to travel at its maximum speed:

[0080]

[0081] Among them, F grade_max T represents the maximum climbing resistance. motor_max The value r represents the maximum output torque of the motor. wheel V represents the wheel radius. grade_max F represents the maximum climbing speed. roll_max and F air_max These represent rolling resistance and air resistance at the maximum vehicle speed, respectively.

[0082] The multi-objective optimization model for transmission speed ratio parameters, with power and economic indicators as optimization goals, uses a weighted method to normalize the objectives. The specific formula is as follows:

[0083] F=ω×F performance +(1-ω)×F economy

[0084] Where ω is the weighting factor, F performance As a dynamic indicator, F economy As an economic indicator,

[0085] Performance metrics are primarily measured by acceleration time, while economic metrics are primarily measured by energy consumption per 100 kilometers.

[0086] Based on the vehicle's energy consumption characteristics and range requirements, the energy, capacity, and voltage parameters of the power battery are optimized.

[0087] Based on the vehicle's energy consumption characteristics and range requirements, the energy, capacity, and voltage parameters of the power battery are optimized.

[0088] The energy design of the power battery needs to meet the maximum range target of the entire vehicle. Energy consumption mainly includes the energy required for motor drive, the energy required for vehicle electronic components, and the energy lost during discharge. The relationship between these factors is as follows:

[0089]

[0090] Where W represents the battery energy required to drive the motor, u a Indicates constant speed driving speed, S represents driving range, DOD represents depth of discharge, and η represents the driving speed. mc η represents the motor efficiency. dis Indicates discharge efficiency, η α This represents the energy consumption ratio of automotive electronic components, where E represents the total battery energy, and C represents the energy consumption ratio of automotive electronic components. E Indicates battery capacity, U m This is the rated voltage of the battery pack.

[0091] The constraints that the transmission ratio must meet include maximum vehicle speed, climbing ability, and acceleration performance.

[0092] Specifically, in selecting the transmission ratio parameters for electric trucks, the maximum transmission ratio must ensure the vehicle can operate under the steepest gradient, while the minimum transmission ratio must ensure the vehicle can operate at its maximum speed. The formula for calculating the total transmission ratio is:

[0093] i = i0i g ;

[0094] Determine the minimum gear ratio i based on the maximum vehicle speed. min Maximum:

[0095]

[0096] n max Indicates the maximum speed of the motor, u maxThis indicates the maximum vehicle speed. The minimum transmission ratio i is determined based on the maximum output torque at the motor's maximum speed and the driving resistance corresponding to the maximum stable vehicle speed. min The lower limit:

[0097]

[0098] Among them, T max C represents the maximum output torque of the motor. D represents the air drag coefficient, and A represents the frontal area. The maximum gear ratio i is determined based on the maximum gradeability. max The lower limit:

[0099]

[0100] α represents the slope angle. Under maximum gear ratio, the maximum driving force must be less than or equal to the adhesion force between the ground and the tires to prevent slippage.

[0101]

[0102] F z This represents the reaction force of the ground on the drive wheels (typically 60% of the total vehicle load). This indicates the adhesion coefficient (0.7).

[0103] Construct a multi-objective optimization model for transmission ratio parameters with power and economic indicators as optimization objectives;

[0104] The transmission ratio parameters are binary encoded, and corresponding decoding and initialization strategies are designed.

[0105] The transmission ratio parameters are encoded in binary. Each gear ratio is encoded using a 10-bit binary vector. The entire optimization variable is a 10-bit binary string. The decoding process converts the binary string into the corresponding real-valued combination of the gear ratio parameters.

[0106] In the optimization design of the transmission ratio of electric trucks, the selection of the objective function should, under the premise of meeting the power response requirements and energy utilization efficiency under actual vehicle operating conditions, take into account performance, economy, and engineering feasibility. Power performance, as the core indicator of a vehicle's power response capability, directly affects the user's driving experience and the vehicle's performance under dynamic conditions such as acceleration and hill climbing; while economic performance reflects the vehicle's energy utilization level and is an important quantitative basis for range and operating costs. Given that these two often constrain each other in actual operation, optimizing a single aspect can easily lead to performance imbalance. Therefore, this invention uses both power performance and economic performance as optimization objectives and employs a weighted sum method to normalize the objectives. The relationship expression is as follows:

[0107] F(x)=ω1×f1(x)+ω2×f2(x) (17)

[0108] Where ω is the weighting factor, f1(x) is the dynamic indicator, and f2(x) is the economic indicator.

[0109] The main performance indicators of electric trucks include acceleration, maximum speed, climbing ability, and traction. Compared to other indicators, acceleration is not only closely related to the matching of the vehicle's drive system, but also covers the requirements of maximum speed and climbing ability to a certain extent, making it representative and prioritized in performance optimization. Therefore, this invention selects the vehicle's acceleration time as the performance objective f1(x). The vehicle's driving dynamics equations are as follows:

[0110] F t =F r +F w +F j

[0111] Right now:

[0112]

[0113] Can be converted to:

[0114]

[0115] The formula for calculating acceleration time is as follows:

[0116]

[0117] Where t is the acceleration time, u0 represents the initial velocity of the vehicle, and u1 represents the final velocity of the vehicle.

[0118] The economic performance indicators of electric trucks mainly include energy consumption per unit distance, energy utilization rate, motor efficiency, and overall vehicle range. Among these, energy consumption per 100 kilometers is a crucial indicator for measuring the overall energy efficiency of the vehicle, representing the amount of electrical energy consumed per 100 kilometers of driving. This indicator not only reflects the synergistic effect of the drive system efficiency and the vehicle matching strategy but also directly relates to the vehicle's operating costs and energy economy, thus holding significant engineering importance in vehicle design and performance optimization.

[0119] Therefore, this paper uses the average energy consumption per 100 kilometers as the economic optimization objective f2(x). Typically, when testing this indicator, the vehicle travels at a constant speed on a straight and good road, recording the total electrical energy consumed and the distance traveled, thus calculating the energy consumption per 100 kilometers. In this process, the vehicle's power balance formula is:

[0120]

[0121] The driving range can be expressed as:

[0122]

[0123] The total energy of the motor drive can be expressed as:

[0124] W = EDODη mc η dis (1-η a )

[0125] In summary, the formula for calculating energy consumption per 100 kilometers is:

[0126]

[0127] Construct and train a deep Q-network agent, and design the state space, action space, reward function, selection policy, and neural network architecture;

[0128] Specifically, in this invention, the transmission speed ratio parameter of an electric truck is used as the optimization variable. For an electric truck with n gears, its speed ratio can be expressed as X. i =[i1,...,i g ,...,i n [i0] T i g Let represent the gear ratio of the g-th gear, and i0 represent the final drive ratio. Each gear ratio is encoded using a 10-bit binary vector, making the entire optimization variable a 10(n+1)-bit binary string. The first 10 bits represent the final drive ratio, the next 10 bits represent the first gear ratio, and so on. Decoding involves converting the binary string into the corresponding real-valued gear ratio parameter combination. Furthermore, at the start of the algorithm, a random initialization method is used, randomly generating PS 10(n+1)-bit binary encoded strings, each corresponding to a solution, where PS is the population size. After initialization, each individual in the population is decoded sequentially, and its fitness function value is calculated based on the corresponding dynamic and economic indicators and weights, providing an evaluation basis for subsequent algorithm evolution operations.

[0129] The state space of a deep Q-network agent includes gear ratio, main reducer ratio, acceleration time, and energy consumption per 100 kilometers. The action space includes mutation operations for different gear ratio parameters. The reward function determines the reward value based on whether the acceleration time and energy consumption per 100 kilometers of the new solution are better than the solution to be optimized.

[0130] Specifically, the optimization variables in this invention consist of the speed ratios of each gear and the speed ratio of the main reducer. Conventional evolutionary methods have significant randomness, making it impossible to control the range of variation when dealing with such structured parameters. Therefore, this paper employs DQN to intelligently select appropriate perturbation strategies and effective ranges based on the real-time state characteristics of the current solution, thereby achieving directional perturbation of the speed ratio parameters of the current solution and enhancing algorithm performance.

[0131] The detailed design is as follows:

[0132] State space: The solution to be optimized contains n gear ratios, the final drive ratio, and (n+3) characteristic parameters calculated based on the solution, including acceleration time and energy consumption per 100 kilometers. Therefore, these features are integrated into a one-dimensional vector of length (n+3) as the state input of DQN.

[0133] Action Space: The solution uses binary encoding, and this invention uses mutation operations as a perturbation strategy. However, conventional mutation methods have significant randomness and cannot control the mutation range when processing such structured parameters. For example, in a two-gear transmission system, when the first gear ratio and the final drive ratio are already at a good level, and only the second gear ratio needs further optimization, conventional mutation methods cannot target only the second gear ratio, nor can they determine when to mutate it. Therefore, this paper uses DQN to intelligently select the appropriate mutation interval based on the current solution's state characteristics in real time. In the above example, the effective range of the mutation operation can be divided into three types: first gear ratio mutation, second gear ratio mutation, and final drive ratio mutation. Furthermore, considering the case where mutation operations act on multiple ratio parameters simultaneously, four combinations can be formed: joint mutation of first and second gear ratios, joint mutation of first and final drive ratios, joint mutation of second and final drive ratios, and simultaneous mutation of all three. In summary, the action space contains a total of 7 different mutation methods. Therefore, in an n-gear transmission system, there are 2... n -1 mutation methods (ignoring the invariant case). DQN selects the optimal mutation range based on the state characteristics of the current solution to achieve targeted optimization of key parameters.

[0134] Network Structure: The DQN constructed in this invention adopts a five-layer fully connected network structure consisting of an input layer, four hidden layers, and an output layer. The network input is a state feature vector of length (n+3), and the output corresponds to the Q-value of the action in the action space. To enhance the model's ability to model complex feature relationships, all hidden layers use the ReLU activation function. The number of neurons in each hidden layer is set to 128, 256, 128, and 64, respectively, compressing the feature dimension layer by layer to improve the network's non-linear expressive power and training efficiency.

[0135] Reward function: After the solution to be optimized is mutated to generate a new solution, the acceleration time and energy consumption per 100 kilometers of the new solution are calculated. If the acceleration time and energy consumption per 100 kilometers of the new solution are both better than the solution to be optimized, the reward is 10; if the new solution is better than the solution to be optimized in only one of acceleration time or energy consumption per 100 kilometers, the reward is 5; otherwise, the reward is 0.

[0136] Selection Strategy: In DQN training, selecting only the action with the highest Q-value leads to over-reliance on specific state-action combinations, limiting the exploration of other potential optimal solutions. Therefore, this paper employs an ε-greedy selection strategy to balance the selection probability of each action during training. First, a small number `rand` within the range of 0 to 1 is generated. If `rand` is less than the threshold `ε`, the action with the highest Q-value is selected. Otherwise, actions are randomly selected with equal probability. The threshold `ε` changes linearly during training. Where E represents the total number of training iterations, and e represents the current training iteration.

[0137] Agent training: First, initialize the experience replay buffer buffer_m and the reward buffer buffer_r, and construct the evaluation network Q and the target network Q. * Q = Q * Next, we begin cyclic training. In each training round, the agent must first reset the environment and initialize the cumulative reward r. c At each decision point t, the agent bases its decision on the current state s. t Select action a t Execute action a t To make the agent enter the next state s t+1 Calculate the instant reward r t and the state transition (s t ,a t ,r t ,s t+1 The data is stored in the experience replay buffer. Once buffer_m reaches a predefined threshold, a batch of data (s) is randomly sampled from it. t ,a t ,r t ,s t+1 Using the evaluation network Q and the target network Q), * Calculate the Q-value and loss function Δloss. Then, minimize Δloss using the Adam optimizer and update the evaluation network Q. Finally, at the end of each training iteration, calculate the average reward r over the most recent 10 epochs. ave and compared it with the best recorded average reward r best Compare. If r ave >r best Then save the current target network Q. * .

[0138] A population-individual collaborative restart mechanism is introduced to prevent the algorithm from getting stuck in local optima;

[0139] The population-individual collaborative restart mechanism determines whether a population is trapped in a local optimum by checking the number of consecutive times an individual has not been updated, and then regenerates a new individual to replace the original individual through population collaborative operations.

[0140] A multi-objective evolutionary algorithm improved by deep Q-network is constructed to solve the optimal combination of transmission ratio parameters for electric trucks.

[0141] Specifically, at the end of each iteration of the algorithm, the number of consecutive times an individual in the population has not been updated is checked. If an individual in the population has not been updated for C consecutive times, it is considered to be trapped in a local optimum, and a new individual will be generated to replace the original individual using a population cooperative operation, as shown in Algorithm 1:

[0142]

[0143]

[0144] Based on the methods described above, a deep Q-network is constructed to improve the multi-objective evolutionary algorithm. The algorithm flow is as follows: Figure 2 As shown, this invention addresses the optimization problem of transmission ratio parameters for electric trucks by simultaneously optimizing acceleration time and energy consumption per 100 kilometers. Therefore, the weights of the two optimization objectives must first be determined based on factors such as the actual application, operating conditions, and target requirements of the electric truck. During algorithm execution, the determined weights are used to calculate the fitness function, thereby driving the algorithm's evolution. Finally, the solution with the best fitness value is selected as the optimal combination of transmission ratio parameters for the electric truck.

[0145] To further verify the effectiveness of the proposed method for optimizing the transmission ratio of electric trucks based on a deep Q-network-based improved multi-objective evolutionary algorithm, this invention conducted a case study based on the actual parameters of a certain model of electric truck. Table 1 shows the vehicle parameters, and Table 2 shows the performance index requirements.

[0146] Table 1 Basic Parameters of the Vehicle

[0147]

[0148]

[0149] Table 2 Performance Indicators

[0150]

[0151] The vehicle adopts a three-section axle housing drive structure, with square axle housings on both sides and the motor and gearbox assembly integrated in the middle, connected by flanges. Based on step 1, we can calculate: P1 = 175.12 kW, P2 = 192.7 kW, P3 = 167.4 kW. With the motor overload coefficient λ set to 3, and considering a certain amount of redundancy, the maximum power can be set to 210 kW, and the rated power to 70 kW. Therefore, the maximum speed n can be selected. max =12000r / min, rated speed is n eA permanent magnet synchronous motor (PMSM) with a speed of 4000 r / min is used as the vehicle's drive source, and the rated torque of the motor can be calculated to be 167.125 N·m, and the maximum torque of the motor is 501.375 N·m.

[0152] Assuming the maximum driving range is met, calculations from step 2 yield: W = 542.85 kWh, E ≥ 827.4 kWh. If a battery pack with a rated voltage of 640V is selected, the battery capacity C... E ≥1292.81A·h. Considering factors such as vehicle power requirements, battery specifications, lifespan, and economic cost, lithium iron phosphate (LFP) batteries can be selected as the energy supply unit for the vehicle. Specifically, LFP cells with a rated voltage of 3.2V and a rated capacity of 220Ah are selected, and six cells are connected in parallel to form a battery module. 200 such modules are then connected in series to construct the entire power battery system. The total system voltage is 3.2 × 200 = 640V, and the total energy is 640 × 220 × 6 / 1000 = 844.8 kWh.

[0153] The maximum and minimum transmission ratios must satisfy the constraint: 7.54 ≤ i min ≤27.14≤93.96≤i max ≤136.96. The calculation results show that the value ranges corresponding to the maximum and minimum transmission ratios do not overlap, indicating that it is difficult to simultaneously meet all the vehicle's performance requirements in terms of power using a single fixed transmission ratio. This results in a conflict between the vehicle's acceleration performance, climbing ability, and high-speed cruising. Therefore, this case study adopts a 3-speed ratio transmission scheme to distribute the load requirements under different operating conditions, achieving synergistic optimization of power and economy.

[0154] Combining vehicle parameters, drive motor, power battery, and transmission constraints, a multi-objective optimization model was constructed with acceleration time and energy consumption per 100 kilometers as objectives. The proposed deep Q-network improved multi-objective evolutionary algorithm was then used for optimization. In this case, the power performance and economic performance indicators were assigned equal weights (ω1=ω2=0.5), and the optimal transmission ratio parameter combination was obtained as follows: first gear ratio 16.37, second gear ratio 7.2, third gear ratio 3.8, and final drive ratio 6.26. Finally, the electric truck transmission ratio combination obtained by the algorithm was simulated and verified using ADVISOR2022 under the CYC_UDDS condition. The results are shown in Table 3.

[0155] Table 3 Simulation Results

[0156]

[0157] As shown in Table 3, the transmission ratio parameters obtained by the algorithm fully meet the requirements of electric trucks in terms of acceleration, range, hill climbing, and energy saving under complex cyclic operating conditions. Therefore, the method proposed in this invention can effectively solve the transmission ratio parameters of electric trucks, and can provide theoretical support and technical means for the intelligent and efficient design of electric truck transmission systems.

[0158] A multi-objective evolutionary algorithm improved by deep Q-networks enhances the algorithm's global search capability and convergence efficiency by dynamically adjusting the evolutionary strategy, ultimately solving for the optimal combination of transmission ratio parameters for electric trucks. Simulations are used to verify the optimization results, ensuring they meet the performance requirements of electric trucks under complex operating conditions.

[0159] Specifically, the Deep Q-Network Improved Multi-Objective Evolutionary Algorithm (MOEA) is an optimization method that integrates the reinforcement learning mechanism of Deep Q-Network (DQN) with the global optimization capability of the MOEA. This method significantly improves the algorithm's global search capability and convergence efficiency by dynamically adjusting the evolutionary strategy, ultimately enabling efficient solution of the optimal combination of transmission ratio parameters for electric trucks.

[0160] In its implementation, the DQN-IMOE algorithm utilizes a deep Q-network to model the state space and action space of the electric truck transmission ratio optimization problem. The state space includes the current operating state of the electric truck, such as vehicle speed, battery status, and load conditions; the action space corresponds to the transmission ratio adjustment strategy. By designing a reasonable reward function, the algorithm can dynamically adjust the transmission ratio parameters based on the electric truck's performance under different operating conditions, such as performance indicators (acceleration time, maximum speed, climbing ability) and economic indicators (energy consumption, driving range), to achieve multi-objective optimization.

[0161] Furthermore, the DQN-IMOE algorithm introduces a population-wide collaborative restart mechanism, effectively preventing the algorithm from getting trapped in local optima during optimization. This mechanism determines whether the algorithm is trapped in a local optimum by monitoring the number of consecutive times an individual in the population has not been updated. Once a local optimum is detected, the algorithm will regenerate new individuals through population-wide collaborative operations to replace the original locally optimal individuals, thereby ensuring the algorithm's global search capability.

[0162] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent optimization of electric truck drive gear ratio, characterized in that, Comprising the following steps: 1.

1. According to the whole vehicle parameters, performance indicators and working condition load characteristics of the electric truck, the power, torque and speed of the drive motor are matched and parameter configured; 1.

2. Based on the whole vehicle energy consumption characteristics and the endurance mileage demand, the energy, capacity and voltage parameters of the power battery are optimized and configured; 1.

3. Analyzing the constraint conditions that the transmission ratio needs to meet, including the maximum vehicle speed, climbing ability and acceleration performance; 1.

4. Building a transmission ratio parameter multi-objective optimization model with power and economy indicators as optimization objectives; 1.

5. Binary coding the transmission ratio parameters, and designing the corresponding decoding and initialization strategies; 1.

6. Constructing and training a deep Q network agent, designing state space, action space, reward function, selection strategy and neural network architecture; 1.

7. Introducing a population individual cooperative restart mechanism to avoid the algorithm falling into local optimum; 1.

8. Building a deep Q network improved multi-objective evolutionary algorithm to solve the optimal electric truck transmission ratio parameter combination, which improves the global search ability and convergence efficiency of the algorithm through dynamic adjustment of the evolutionary strategy, and finally solves the optimal electric truck transmission ratio parameter combination; The simulation verification is carried out on the optimization result to ensure that it meets the performance requirements of the electric truck under complex working conditions.

2. A method of intelligent optimization of transmission gear ratios of an electric truck as claimed in claim 1 wherein, In the step 1.1, the power matching of the drive motor needs to meet the maximum vehicle speed, maximum climbing ability and acceleration performance at the same time, and the maximum output power needs to achieve overall balance under multiple performance constraint conditions, and the specific calculation formula is as follows: According to the maximum vehicle speed, the maximum output power is determined: ； According to the maximum climbing ability, the maximum output power is determined: ； According to the acceleration performance, the maximum output power is determined: ； Wherein, g represents the acceleration of gravity, vmax represents the preset maximum speed, θ represents the climbing angle of the automobile, vgrade represents the constant speed of the automobile climbing, a represents the target acceleration of acceleration, vacc represents the average speed in the acceleration time, Froll represents the rolling resistance, Fair represents the air resistance, and ηmech represents the mechanical transmission efficiency.

3. A method of intelligent optimization of transmission gear ratios of an electric truck as claimed in claim 1, wherein, In the step 1.2, the energy design of the power battery needs to meet the maximum endurance index of the whole vehicle, and the energy consumption includes the energy required by the motor drive, the energy required by the vehicle electronic components and the energy loss in the discharge process, and the specific calculation formula is as follows: ； Wherein, ； ； ； Pmotor represents the power required by the motor drive, D represents the endurance mileage, ηmotor represents the motor efficiency, ηdischarge represents the discharge efficiency, and α represents the energy consumption proportion coefficient of automobile electronic components.

4. A method of intelligent optimization of transmission gear ratios of an electric truck as claimed in claim 1, wherein, In the step 1.3, the constraint conditions that the transmission ratio needs to meet include: The maximum transmission ratio needs to meet the vehicle driving under the maximum slope condition: ； The minimum transmission ratio needs to meet the vehicle driving at the maximum vehicle speed: ； Wherein, Fgrade_max represents the maximum climbing resistance, Tmotor_max represents the maximum output torque of the motor, rwheel represents the wheel radius, vgrade_max represents the maximum climbing speed, Froll_max and Fair_max represent the rolling resistance and air resistance at the highest speed respectively.

5. A method of intelligent optimization of transmission gear ratios of an electric truck as claimed in claim 1, wherein, In the 1.4 step, the multi-objective optimization model of the transmission ratio parameter with the power performance and the economic performance as the optimization targets adopts a weighting method to normalize the targets, and the specific formula is as follows: ； Wherein, ω is the weight factor, Fperformance is the power performance index, and Feconomy is the economic performance index. The power performance index is mainly the acceleration time, and the economic performance index is mainly the energy consumption per 100 kilometers.

6. A method of intelligent optimization of transmission gear ratios of an electric truck as claimed in claim 1, wherein, In the 1.5 step, the transmission ratio parameter adopts binary coding, each gear ratio is coded by a 10-bit binary vector, the whole optimization variable is a binary string with a length of 10, and the decoding process converts the binary string into the corresponding real ratio parameter combination.

7. A method of intelligent optimization of transmission gear ratios of an electric truck as claimed in claim 1, wherein, In the 1.6 step, the state space of the deep Q network agent includes the gear ratio, the main reducer ratio, the acceleration time and the energy consumption per 100 kilometers, and the action space includes the mutation operation for different ratio parameters; the reward function determines the reward value according to whether the acceleration time and the energy consumption per 100 kilometers of the new solution are better than the to-be-optimized solution.

8. A method of intelligent optimization of transmission gear ratios of an electric truck as claimed in claim 1, wherein, In the 1.7 step, the population individual cooperative restart mechanism judges whether it is trapped in the local optimum by checking the continuous number of times of non-updating of the individuals in the population, and generates new individuals to replace the original individuals through the population cooperative operation.

Images