New energy automobile gear decision-making method and control system based on deep reinforcement learning

Abstract

The invention relates to the technical field of intelligent control, in particular to a new energy automobile gear decision-making method based on deep reinforcement learning and a control system. The method comprises the following steps that a vehicle speed sensor, a motor rotating speed sensor and a battery state sensor are arranged on a new energy vehicle, vehicle state real-time collection and vehicle data fusion processing are carried out, and a new energy vehicle driving fusion data set is obtained; the current gear of the new energy automobile is obtained, state space vector determination is carried out in combination with the driving speed, the motor rotating speed, the battery electric quantity, the road gradient and the curvature corresponding to the current time, the state space vector is used as state input corresponding to a preset deep reinforcement learning model, and gear action space definition is carried out; and generating new energy vehicle gear switching action space selection, designing a corresponding vehicle gear decision reward function, performing vehicle driving gear decision control, and generating an optimal gear decision control strategy. According to the invention, more intelligent and efficient gear decision can be realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to a new energy vehicle gear decision method and control system based on deep reinforcement learning. Background Art

[0002] As an important direction for the development of future transportation, the improvement of intelligent control technology is crucial for new energy vehicles. In the power system of new energy vehicles, the choice of gear has a crucial impact on energy efficiency, driving performance and the life of the power system. In recent years, deep reinforcement learning (DRL), as a powerful machine learning method, has gradually been applied in the field of intelligent control. Deep reinforcement learning can continuously optimize the decision-making strategy according to feedback signals by simulating the interaction between the environment and the intelligent agent. In the gear decision of new energy vehicles, deep reinforcement learning can obtain information such as vehicle speed, acceleration, battery status, etc. in real time, and adaptively adjust the gear selection based on this information, so as to achieve the best balance between energy efficiency and driving performance. However, the existing gear decision method mainly relies on preset control rules and lacks intelligence and adaptive capabilities. This method cannot make personalized and real-time gear adjustments for different driving environments, resulting in energy waste. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide a new energy vehicle gear decision method and control system based on deep reinforcement learning to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a new energy vehicle gear decision method based on deep reinforcement learning includes the following steps: Step S1: by deploying a vehicle speed sensor, a motor speed sensor and a battery status sensor on the new energy vehicle, and using the vehicle speed sensor, the motor speed sensor and the battery status sensor to collect the vehicle status of the real-time driving process of the new energy vehicle in real time, so as to obtain the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle; obtaining the road slope and curvature corresponding to the new energy vehicle at different driving road locations, and performing vehicle data fusion processing on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle based on the road slope and curvature corresponding to the new energy vehicle at different driving road locations, so as to obtain a new energy vehicle driving fusion data set; Step S2: obtaining the current gear of the new energy vehicle and determining the state space vector in combination with the corresponding driving speed, motor speed, battery power, road slope and curvature in the new energy vehicle driving fusion data set at the current time, so as to obtain the state space vector of the new energy vehicle; using the state space vector of the new energy vehicle as the state input corresponding to the preset deep reinforcement learning model, and using the deep reinforcement learning model to define the gear action space of the state space vector of the new energy vehicle, so as to generate the gear switching action space selection of the new energy vehicle, including the corresponding upshift, downshift and keeping current gear actions of the new energy vehicle; Step S3: Designing a corresponding vehicle gear decision reward function by selecting the gear switching action space of the new energy vehicle corresponding to the deep reinforcement learning model; Step S4: Based on the vehicle gear decision reward function, the deep reinforcement learning model is used to perform vehicle gear decision control on the corresponding new energy vehicle state space vector, so as to select the optimal gear switching action from the action space corresponding to the model according to the real-time collected vehicle state information, and generate the optimal gear decision control strategy corresponding to the new energy vehicle under different driving conditions.

[0005] Further, step S1 includes the following steps: Step S11: by deploying a vehicle speed sensor, a motor speed sensor and a battery status sensor on the new energy vehicle, and using the vehicle speed sensor to collect the driving speed of the new energy vehicle in real time during the real-time driving process, so as to obtain the driving speed of the new energy vehicle; Step S12: using a motor speed sensor to collect the motor speed of the new energy vehicle in real time during the real-time driving process to obtain the motor speed of the new energy vehicle; Step S13: using a battery status sensor to collect the battery power of the new energy vehicle in real time during the real-time driving process to obtain the battery power of the new energy vehicle; Step S14: Obtaining the road slope and curvature corresponding to different driving road locations of the new energy vehicle; Step S15: Based on the road slope and curvature corresponding to different driving road locations of the new energy vehicle, the vehicle data fusion processing is performed on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle to obtain a new energy vehicle driving fusion data set.

[0006] Further, step S15 includes the following steps: Step S151: Perform vehicle data Kalman filtering on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle to obtain the driving noise reduction speed of the vehicle, the motor noise reduction speed of the vehicle and the battery noise reduction power of the vehicle; Step S152: constructing a vehicle dynamics model and a battery model corresponding to the new energy vehicle, and performing driving state correlation analysis on the vehicle driving noise reduction speed, the vehicle motor noise reduction speed, and the vehicle battery noise reduction power based on the vehicle dynamics model and the battery model, to obtain a new energy vehicle driving state correlation data set; Step S153: Obtain the corresponding vehicle driving road space location through the road slope and curvature corresponding to the new energy vehicle at different driving road locations, and perform driving space synchronization mapping on the new energy vehicle driving state associated data set based on the vehicle driving road space location to obtain the corresponding vehicle driving state data set at the same driving road location, including the driving speed, motor speed and battery power; Step S154: Based on the road slope and curvature corresponding to the new energy vehicle at different driving road locations, vehicle data fusion processing is performed on the vehicle driving state data set corresponding to the same driving road location to obtain a new energy vehicle driving fusion data set.

[0007] Further, step S2 includes the following steps: Step S21: obtaining the current gear position of the new energy vehicle; Step S22: determining the state space vector of the new energy vehicle by combining the current gear position of the new energy vehicle with the corresponding driving speed, motor speed, battery power, road slope and curvature in the new energy vehicle driving fusion data set at the current time, so as to obtain the state space vector of the new energy vehicle; Step S23: dividing the new energy vehicle state space vector into individual vehicle state discrete space vectors according to corresponding time intervals, and determining the corresponding vehicle speed state value, rotation speed state value, power state value, slope state value and curvature state value in each time range according to each vehicle state discrete space vector; Step S24: using the vehicle speed state value, rotation speed state value, power state value, slope state value and curvature state value corresponding to each time range as the state input corresponding to the preset deep reinforcement learning model, and performing gear power and energy consumption evaluation calculation on the current gear of the new energy vehicle based on the vehicle speed state value, rotation speed state value, power state value, slope state value and curvature state value corresponding to each time range, to obtain the gear power performance state value of the new energy vehicle and the gear energy loss of the new energy vehicle; Step S25: Use the deep reinforcement learning model to define the gear action space of the new energy vehicle gear power performance state value and the new energy vehicle gear energy loss to generate the new energy vehicle gear switching action space selection, including the corresponding upshift, downshift and maintain current gear actions of the new energy vehicle.

[0008] Further, the gear power and energy consumption evaluation calculation for the current gear of the new energy vehicle based on the corresponding vehicle speed state value, rotation speed state value, power state value, slope state value and curvature state value in each time range in step S24 includes the following steps: Based on the corresponding vehicle speed state value, rotation speed state value and power state value in each time range, the gear power performance evaluation calculation formula is used to perform gear power evaluation calculation on the current gear of the new energy vehicle to obtain the gear power performance state value of the new energy vehicle; The vehicle gear power output of the new energy vehicle in the corresponding time range is simulated and calculated according to the gear power performance state value of the new energy vehicle, and the vehicle gear power peak value corresponding to the new energy vehicle is calculated according to the vehicle gear power output; Based on the slope state value and the curvature state value, the energy consumption impact of the current gear of the new energy vehicle is analyzed to obtain the energy consumption impact coefficient of the gear of the new energy vehicle; Based on the vehicle gear power peak and the new energy vehicle gear energy consumption impact coefficient, the vehicle energy loss evaluation calculation formula is used to evaluate and calculate the energy consumption of the current gear of the new energy vehicle, and the gear energy loss of the new energy vehicle is obtained; Among them, the vehicle energy loss assessment calculation formula is as follows: ; In the formula, The energy loss of new energy vehicles during gear shifting. is the initial value of the time range, is the end value of the time range, is the time variable parameter, For new energy vehicles in time The vehicle speed state value at For new energy vehicles in time The speed state value at For new energy vehicles in time The power state value at is the weight coefficient of power energy loss, is the peak power of the vehicle gear, is the energy consumption impact coefficient of the gear position of new energy vehicles, It is the correction coefficient for energy loss during gear shift of new energy vehicles.

[0009] Furthermore, the gear dynamic performance evaluation calculation formula is specifically as follows: ; In the formula, is the gear power performance status value of new energy vehicles, is the initial value of the time range, is the end value of the time range, is the time variable parameter, For new energy vehicles in time The vehicle speed state value at is the maximum speed of the new energy vehicle in the current gear, is the vehicle speed power contribution coefficient, For new energy vehicles in time The speed state value at is the maximum motor speed of the new energy vehicle in the current gear, is the motor speed power contribution coefficient, For new energy vehicles in time The power state value at It is the maximum battery capacity of the new energy vehicle in the current gear. is the battery power contribution coefficient, is the power attenuation time constant of the new energy vehicle in the current gear, is the time exponential decay factor, It is the correction coefficient of the gear power performance status value of new energy vehicles.

[0010] Furthermore, the gear action space definition described in step S25 is specifically that when the power performance state value is large and the gear energy loss is in an acceptable range, the deep reinforcement learning model is used to define the action space corresponding to the new energy vehicle as an upshift action; when the power performance state value is insufficient and the gear energy loss is large, the deep reinforcement learning model is used to define the action space corresponding to the new energy vehicle as a downshift action; when the power performance state value and the gear energy loss are in a balanced state, the deep reinforcement learning model is used to define the action space corresponding to the new energy vehicle as a maintain current gear action.

[0011] Furthermore, the vehicle gear decision reward function described in step S3 is specifically that when the action gear selected by the deep reinforcement learning model results in reduced energy consumption and enhanced power performance, a positive reward is given to the gear decision; when the action gear selected by the deep reinforcement learning model results in excessive energy consumption and weakened power performance to affect the driving smoothness of the new energy vehicle, a negative reward is given to the gear decision.

[0012] Furthermore, the optimal gear decision control strategy corresponding to different driving conditions described in step S4 is specifically that when the vehicle is traveling on congested roads, the model determines whether it is necessary to downshift to maintain the corresponding power output and smoothness of the vehicle based on the vehicle speed and insufficient motor speed; when the vehicle is traveling on high-speed roads, the model determines whether it is necessary to upshift to reduce the motor speed and improve energy utilization efficiency based on sufficient battery power and excessive energy consumption; and when the vehicle is traveling on smooth roads, the model determines whether it is necessary to maintain the current gear to maintain the smooth operation of the vehicle and avoid unnecessary power fluctuations and motor losses based on normal battery power consumption and the fact that gear switching will not bring about obvious power improvement or energy consumption reduction.

[0013] Furthermore, the present invention also provides a new energy vehicle gear decision control system based on deep reinforcement learning, which is used to execute the new energy vehicle gear decision method based on deep reinforcement learning as described above. The new energy vehicle gear decision control system based on deep reinforcement learning includes: The vehicle driving status data fusion module is used to deploy a vehicle speed sensor, a motor speed sensor and a battery status sensor on the new energy vehicle, and use the vehicle speed sensor, the motor speed sensor and the battery status sensor to collect the vehicle status of the new energy vehicle in real time during the real-time driving process, so as to obtain the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle; obtain the road slope and curvature corresponding to the new energy vehicle at different driving road locations, and perform vehicle data fusion processing on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle based on the road slope and curvature corresponding to the new energy vehicle at different driving road locations, so as to obtain the new energy vehicle driving fusion data set; The vehicle gear action model analysis module is used to obtain the current gear of the new energy vehicle and determine the state space vector in combination with the corresponding driving speed, motor speed, battery power, road slope and curvature in the new energy vehicle driving fusion data set at the current time to obtain the new energy vehicle state space vector; the new energy vehicle state space vector is used as the state input corresponding to the preset deep reinforcement learning model, and the deep reinforcement learning model is used to define the gear action space of the new energy vehicle state space vector to generate the new energy vehicle gear switching action space selection, including the corresponding upshift, downshift and keep current gear actions of the new energy vehicle; The model decision reward function design module is used to design the corresponding vehicle gear decision reward function through the gear switching action space selection of the new energy vehicle corresponding to the deep reinforcement learning model; The vehicle gear decision control module is used to perform vehicle gear decision control on the corresponding new energy vehicle state space vector based on the vehicle gear decision reward function using a deep reinforcement learning model, so as to select the optimal gear switching action from the action space corresponding to the model according to the real-time collected vehicle state information, thereby generating the optimal gear decision control strategy corresponding to the new energy vehicle under different driving conditions.

[0014] Beneficial effects of the present invention: 1. The gear decision method for new energy vehicles based on deep reinforcement learning proposed in the present invention has the beneficial effect of providing key data support for subsequent vehicle control by deploying vehicle speed sensors, motor speed sensors and battery status sensors on new energy vehicles and performing real-time data collection. The vehicle speed sensor can monitor the vehicle's driving speed in real time, the motor speed sensor can reflect the running state of the motor, and the battery status sensor provides the remaining power and health status of the battery. Through the data of these sensors, the current operating state of the new energy vehicle can be accurately grasped. These data not only reflect the current driving state of the vehicle, but can also be further combined with road information, especially changes in road slope and curvature. This step can obtain a comprehensive vehicle driving data set by real-time monitoring of the road slope and curvature and fusing them with the vehicle's driving data. The fused data set not only includes the vehicle speed, motor speed and battery power, but can also be dynamically adjusted according to changes in the road, so that the vehicle can better adapt to different driving conditions, thereby improving driving efficiency and energy efficiency, and improving the adaptability and safety of the vehicle under complex conditions. Secondly, by obtaining the current gear of new energy vehicles and combining it with real-time driving data and road information, accurate state input can be provided for the deep reinforcement learning model. The gear of new energy vehicles directly determines the characteristics of power output and the response speed of the vehicle. Reasonable gear selection can effectively improve the power performance and fuel efficiency of the vehicle. By vectorizing the current state space, multi-dimensional data such as vehicle speed, motor speed, battery power, road slope, curvature, etc. can be integrated into a unified input as the state input of the deep reinforcement learning model. This method can not only cover all the key state variables of the vehicle, but also enable the model to automatically select the most appropriate gear action according to different working conditions, thereby reducing human intervention and gradually optimizing its decision-making ability for gear selection. As the training proceeds, the model will gradually master the optimal gear switching strategy under different driving conditions, improving the driving comfort, power response and energy efficiency of the vehicle. Then, by designing a reasonable vehicle gear decision reward function, it is the key to realize the effective application of deep reinforcement learning. The design of the reward function directly affects the training effect and decision-making accuracy of the model. A reasonable reward function can help the model clarify the optimization goals, such as improving driving efficiency, reducing energy consumption, extending battery life and ensuring driving safety. By defining the reward function, the deep reinforcement learning model can be guided towards the optimal decision-making direction, encouraging the model to make gear switching decisions that are most beneficial to vehicle performance under changing working conditions. When designing the reward function, multiple factors are usually considered comprehensively, such as the vehicle's energy consumption and driving stability. A well-designed reward function can not only enable the model to learn how to make reasonable gear choices based on the real-time vehicle status, but also provide real-time feedback in complex environments to optimize vehicle performance.Finally, through the deep reinforcement learning model and the designed reward function, new energy vehicles can perform precise gear decision control. The vehicle continuously selects the optimal gear switching strategy from the action space defined by the model based on real-time status information. This process can not only maximize the vehicle's energy efficiency, but also ensure the smooth operation of the vehicle under different driving conditions. Through the deep reinforcement learning model, the vehicle can make intelligent decisions based on different road conditions, driving styles and energy requirements, thereby achieving the most optimized driving experience and energy efficiency management. Under different driving conditions, the model can determine whether it is necessary to upshift, downshift or maintain the current gear based on the input state space information. Each gear switching decision is based on a comprehensive analysis of the current environment and vehicle status. Through this intelligent control strategy, new energy vehicles can achieve lower energy consumption and higher safety during actual driving, enabling them to make personalized and real-time gear adjustments under different driving environments, thereby avoiding the loss of gear energy in new energy vehicles.

[0015] 2. The new energy vehicle gear decision control system based on deep reinforcement learning proposed in the present invention is composed of a vehicle driving status data fusion module, a vehicle gear action model analysis module, a model decision reward function design module and a vehicle driving gear decision control module as a whole. It can realize any new energy vehicle gear decision method based on deep reinforcement learning described in the present invention, and is used to combine the operations between computer programs running on each module to realize the new energy vehicle gear decision method based on deep reinforcement learning. The internal structures of the system cooperate with each other, which can greatly reduce duplication of work and manpower investment, and can quickly and effectively provide a more accurate and efficient new energy vehicle gear decision control process based on deep reinforcement learning, thereby simplifying the operation process of the new energy vehicle gear decision control system based on deep reinforcement learning. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings: Figure 1 It is a schematic diagram of the steps of the gear decision method for new energy vehicles based on deep reinforcement learning of the present invention; Figure 2 for Figure 1 Detailed step flow diagram of step S1; Figure 3 for Figure 2 Detailed step flow chart of step S15 in FIG. DETAILED DESCRIPTION

[0017] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0018] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0019] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0020] To achieve this, please refer to Figures 1 to 3 The present invention provides a new energy vehicle gear decision method based on deep reinforcement learning, the method comprising the following steps: Step S1: by deploying a vehicle speed sensor, a motor speed sensor and a battery status sensor on the new energy vehicle, and using the vehicle speed sensor, the motor speed sensor and the battery status sensor to collect the vehicle status of the real-time driving process of the new energy vehicle in real time, so as to obtain the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle; obtaining the road slope and curvature corresponding to the new energy vehicle at different driving road locations, and performing vehicle data fusion processing on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle based on the road slope and curvature corresponding to the new energy vehicle at different driving road locations, so as to obtain a new energy vehicle driving fusion data set; Step S2: obtaining the current gear of the new energy vehicle and determining the state space vector in combination with the corresponding driving speed, motor speed, battery power, road slope and curvature in the new energy vehicle driving fusion data set at the current time, so as to obtain the state space vector of the new energy vehicle; using the state space vector of the new energy vehicle as the state input corresponding to the preset deep reinforcement learning model, and using the deep reinforcement learning model to define the gear action space of the state space vector of the new energy vehicle, so as to generate the gear switching action space selection of the new energy vehicle, including the corresponding upshift, downshift and keeping current gear actions of the new energy vehicle; Step S3: Designing a corresponding vehicle gear decision reward function by selecting the gear switching action space of the new energy vehicle corresponding to the deep reinforcement learning model; Step S4: Based on the vehicle gear decision reward function, the deep reinforcement learning model is used to perform vehicle gear decision control on the corresponding new energy vehicle state space vector, so as to select the optimal gear switching action from the action space corresponding to the model according to the real-time collected vehicle state information, and generate the optimal gear decision control strategy corresponding to the new energy vehicle under different driving conditions.

[0021] In the embodiment of the present invention, please refer to Figure 1 As shown, it is a schematic diagram of the step flow of the new energy vehicle gear decision method based on deep reinforcement learning of the present invention. In this example, the new energy vehicle gear decision method based on deep reinforcement learning includes the following steps: Step S1: by deploying a vehicle speed sensor, a motor speed sensor and a battery status sensor on the new energy vehicle, and using the vehicle speed sensor, the motor speed sensor and the battery status sensor to collect the vehicle status of the real-time driving process of the new energy vehicle in real time, so as to obtain the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle; obtaining the road slope and curvature corresponding to the new energy vehicle at different driving road locations, and performing vehicle data fusion processing on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle based on the road slope and curvature corresponding to the new energy vehicle at different driving road locations, so as to obtain a new energy vehicle driving fusion data set; In an embodiment of the present invention, a vehicle speed sensor, a motor speed sensor and a battery status sensor are installed to collect the operation data of the new energy vehicle in real time. The gearbox output shaft is monitored by the vehicle speed sensor, and the vehicle speed is calculated in combination with the wheel speed sensor, so that the driving speed of the new energy vehicle can be accurately monitored; the motor speed is monitored by the motor speed sensor, and the rotary signal corresponding to the monitored motor speed is decoded into the motor speed for real-time collection of the motor speed; the battery status sensor (such as BMS) is installed inside the battery pack to monitor the battery power status, charge and discharge rate and battery health. Through these sensors, real-time data of vehicle speed, motor speed and battery power can be collected during vehicle driving. In addition to real-time vehicle data collection, it is also necessary to obtain the slope and curvature information of the current driving section through external measurement equipment, such as ground slope sensors or navigation systems. The slope reflects the degree of rise or fall of the road, and the curvature reflects the curvature of the road. These factors directly affect the vehicle's power demand. According to the corresponding vehicle driving road position, this information is fused with the vehicle's driving data to generate a fused data set including vehicle speed, motor speed, battery power, road slope and curvature, which provides important input data for subsequent gear decisions and finally obtains a new energy vehicle driving fusion data set.

[0022] Step S2: obtaining the current gear of the new energy vehicle and determining the state space vector in combination with the corresponding driving speed, motor speed, battery power, road slope and curvature in the new energy vehicle driving fusion data set at the current time, so as to obtain the state space vector of the new energy vehicle; using the state space vector of the new energy vehicle as the state input corresponding to the preset deep reinforcement learning model, and using the deep reinforcement learning model to define the gear action space of the state space vector of the new energy vehicle, so as to generate the gear switching action space selection of the new energy vehicle, including the corresponding upshift, downshift and keeping current gear actions of the new energy vehicle; In an embodiment of the present invention, by obtaining the current gear state of the new energy vehicle, the current gear information is usually read through the gear control system of the vehicle, and combined with the real-time driving data previously obtained, the state space vector of the new energy vehicle is constructed. The state space vector contains input data of multiple dimensions, such as the current vehicle speed, motor speed, battery power, road slope and curvature, etc. By combining this information, a complete state description vector is formed as the input of the deep reinforcement learning model. The deep reinforcement learning model is trained and inferred through these state space vectors. The goal of the model is to generate a gear selection suitable for the current driving condition by observing the state space. According to the current vehicle driving state (such as low speed or low battery power), the model will evaluate the matching degree between the current gear and the vehicle performance, and select the most appropriate action according to the set reward mechanism. The action space includes upshifting, downshifting and maintaining the current gear. The deep reinforcement learning model finally generates the gear switching action space selection of the new energy vehicle by finding the best decision among these choices.

[0023] Step S3: Designing a corresponding vehicle gear decision reward function by selecting the gear switching action space of the new energy vehicle corresponding to the deep reinforcement learning model; In an embodiment of the present invention, the merits of each decision action are evaluated according to the performance of the gear switching action space selection of the new energy vehicle under different states. Specifically, when the gear action selected by the deep reinforcement learning model can reduce energy consumption and improve power performance, the reward function will give positive rewards. Positive rewards appear, for example, when the vehicle speed is low and it is necessary to downshift to improve motor efficiency, or when the vehicle is in an efficient operating state, upshift to reduce the motor speed, thereby improving energy efficiency. On the contrary, when the gear action selected by the deep reinforcement learning model leads to excessive energy consumption or power performance decline, the reward function will give negative rewards. For example, during high-speed driving, due to low battery power or high motor speed, the model chooses to downshift, which will lead to excessive energy consumption and affect power output, thereby bringing negative rewards. Through this positive and negative reward mechanism, the model can gradually learn how to optimize gear selection under different driving conditions, and finally design and generate the corresponding vehicle gear decision reward function.

[0024] Step S4: Based on the vehicle gear decision reward function, the deep reinforcement learning model is used to perform vehicle gear decision control on the corresponding new energy vehicle state space vector, so as to select the optimal gear switching action from the action space corresponding to the model according to the real-time collected vehicle state information, and generate the optimal gear decision control strategy corresponding to the new energy vehicle under different driving conditions.

[0025] In the embodiment of the present invention, the actual gear decision is made by using a deep reinforcement learning model based on the real-time state information of the vehicle and a designed reward function. The model evaluates possible gear selections based on the current vehicle state vector and the reward function, and selects the optimal gear switching action. For example, in congested road conditions, due to the low vehicle speed, the model will determine whether to downshift based on the vehicle speed and the insufficient motor speed to ensure sufficient power output and maintain driving smoothness. When the vehicle is traveling on a highway, the vehicle speed is high, the battery power is sufficient, and the energy consumption is also high, the model will choose to upshift to reduce the motor speed, thereby improving the performance. Improve the energy utilization efficiency of the vehicle and avoid unnecessary energy consumption. Under smooth road conditions, when the vehicle speed is moderate, the battery consumption is normal, and the gear switching does not have a significant impact on the power output, the model may choose to maintain the current gear to maintain the smooth operation of the vehicle and avoid unnecessary power fluctuations and motor losses. Through this gear decision control strategy based on deep reinforcement learning, new energy vehicles can automatically select the optimal gear according to the real-time driving environment and status information, thereby optimizing power performance and energy efficiency, and finally generating the optimal gear decision control strategy corresponding to new energy vehicles under different driving conditions.

[0026] Further, step S1 includes the following steps: Step S11: by deploying a vehicle speed sensor, a motor speed sensor and a battery status sensor on the new energy vehicle, and using the vehicle speed sensor to collect the driving speed of the new energy vehicle in real time during the real-time driving process, so as to obtain the driving speed of the new energy vehicle; Step S12: using a motor speed sensor to collect the motor speed of the new energy vehicle in real time during the real-time driving process to obtain the motor speed of the new energy vehicle; Step S13: using a battery status sensor to collect the battery power of the new energy vehicle in real time during the real-time driving process to obtain the battery power of the new energy vehicle; Step S14: Obtaining the road slope and curvature corresponding to different driving road locations of the new energy vehicle; Step S15: Based on the road slope and curvature corresponding to different driving road locations of the new energy vehicle, the vehicle data fusion processing is performed on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle to obtain a new energy vehicle driving fusion data set.

[0027] As an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 Detailed step flow diagram of step S1 in FIG. 1 , in this embodiment, step S1 includes the following steps: Step S11: by deploying a vehicle speed sensor, a motor speed sensor and a battery status sensor on the new energy vehicle, and using the vehicle speed sensor to collect the driving speed of the new energy vehicle in real time during the real-time driving process, so as to obtain the driving speed of the new energy vehicle; In an embodiment of the present invention, corresponding vehicle speed sensors, motor speed sensors and battery status sensors are installed and deployed on the new energy vehicle to collect corresponding data of the new energy vehicle during driving in real time, and the deployed vehicle speed sensors are used to collect real-time driving speed data. For example, the gearbox output shaft is monitored by the vehicle speed sensor, and the vehicle speed is calculated in combination with the wheel speed sensor. The driving speed of the new energy vehicle can be accurately monitored. The vehicle speed sensor transmits the data to the vehicle control system through the CAN bus, and provides the driving speed data to the vehicle system every second. During the driving of the vehicle, the sensor can update the vehicle speed value in real time, and transmit the data to the central computing unit for processing, and finally obtain the driving speed of the new energy vehicle.

[0028] Step S12: using a motor speed sensor to collect the motor speed of the new energy vehicle in real time during the real-time driving process to obtain the motor speed of the new energy vehicle; In an embodiment of the present invention, the motor speed is monitored by a motor speed sensor, and the resolver signal corresponding to the monitored motor speed is decoded into the motor speed for real-time acquisition of the motor speed. The real-time speed of the motor is acquired by monitoring the change in the magnetic field generated by the rotation of the motor shaft. The data acquisition system transmits the real-time motor speed to the control unit via the on-board CAN bus. The real-time data of the motor speed is an important basis for evaluating the power output, driving efficiency and energy consumption of the motor. Through the motor speed information collected by the sensor, the control system can evaluate the working status of the motor in real time, thereby providing a basis for gear adjustment or energy optimization, and finally obtaining the motor speed of the new energy vehicle.

[0029] Step S13: using a battery status sensor to collect the battery power of the new energy vehicle in real time during the real-time driving process to obtain the battery power of the new energy vehicle; In an embodiment of the present invention, in order to monitor the battery power of a new energy vehicle in real time, a battery status sensor (BMS, battery management system) is used to collect real-time battery power. The battery status sensor evaluates the remaining power of the battery by monitoring key parameters such as the voltage, current and temperature of the battery pack. The battery management system processes these parameters through an algorithm, calculates the current remaining power percentage of the battery, and transmits it to the central control unit through the on-board CAN bus. The power data can not only reflect the battery charging status in real time, but also evaluate the battery health status under different working conditions, and finally obtain the battery power of the new energy vehicle.

[0030] Step S14: Obtaining the road slope and curvature corresponding to different driving road locations of the new energy vehicle; In an embodiment of the present invention, the slope and curvature information of the driving road is obtained through vehicle-mounted sensors. The road slope can be calculated by the installed accelerometer and gyroscope combined with the geographical location information in the vehicle-mounted navigation system, and the curvature is obtained through the vehicle-mounted GPS system and the real-time positioning data of the road marking points combined with the sensor to evaluate the turning radius and road curvature during vehicle driving. This information is integrated into the vehicle-mounted computing platform through a high-precision map and road condition data interface. As the vehicle travels on different road sections, the real-time acquired road slope and curvature data can reflect the driving difficulty of different road sections, and finally the road slope and curvature corresponding to the new energy vehicle at different driving road locations are obtained.

[0031] Step S15: Based on the road slope and curvature corresponding to different driving road locations of the new energy vehicle, the vehicle data fusion processing is performed on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle to obtain a new energy vehicle driving fusion data set.

[0032] In an embodiment of the present invention, vehicle data fusion processing is performed on data collected by multiple sensors (including vehicle speed, motor speed, battery power, road slope and curvature) to obtain a more comprehensive new energy vehicle driving fusion data set. First, the on-board control system will perform time synchronization processing on the data from each sensor to ensure that the data collected by each sensor can correspond to the driving state at the same time. Then, through a fusion algorithm (such as Kalman filtering or weighted averaging), the collected data is weighted and integrated to obtain a fused vehicle driving data set. The data set contains information such as the vehicle's driving speed, motor speed, battery power, and road slope and curvature, which is used for decision-making training by a deep reinforcement learning model. Data fusion processing not only improves the accuracy of the data, but also effectively eliminates noise interference, providing accurate input data for subsequent gear decision optimization, and finally obtaining a new energy vehicle driving fusion data set.

[0033] Further, step S15 includes the following steps: Step S151: Perform vehicle data Kalman filtering on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle to obtain the driving noise reduction speed of the vehicle, the motor noise reduction speed of the vehicle and the battery noise reduction power of the vehicle; Step S152: constructing a vehicle dynamics model and a battery model corresponding to the new energy vehicle, and performing driving state correlation analysis on the vehicle driving noise reduction speed, the vehicle motor noise reduction speed, and the vehicle battery noise reduction power based on the vehicle dynamics model and the battery model, to obtain a new energy vehicle driving state correlation data set; Step S153: Obtain the corresponding vehicle driving road space location through the road slope and curvature corresponding to the new energy vehicle at different driving road locations, and perform driving space synchronization mapping on the new energy vehicle driving state associated data set based on the vehicle driving road space location to obtain the corresponding vehicle driving state data set at the same driving road location, including the driving speed, motor speed and battery power; Step S154: Based on the road slope and curvature corresponding to the new energy vehicle at different driving road locations, vehicle data fusion processing is performed on the vehicle driving state data set corresponding to the same driving road location to obtain a new energy vehicle driving fusion data set.

[0034] As an embodiment of the present invention, refer to Figure 3 As shown, Figure 2 Detailed step flow diagram of step S15 in FIG. 1 , in this embodiment, step S15 includes the following steps: Step S151: Perform vehicle data Kalman filtering on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle to obtain the driving noise reduction speed of the vehicle, the motor noise reduction speed of the vehicle and the battery noise reduction power of the vehicle; In an embodiment of the present invention, a Kalman filter is performed on the driving speed, motor speed and battery power of a new energy vehicle. To this end, various data of the new energy vehicle during driving are first collected. These data will be affected by environmental factors, sensor errors and other noises. Kalman filtering is an optimal recursive filtering method, which can suppress noise from these data and accurately estimate the actual driving state of the vehicle. The specific operation is to first establish a Kalman filter model, input the vehicle speed, motor speed and battery power as state variables into the filter, and perform weighted update on each input observation value through the Kalman gain matrix to filter out the noise, so as to obtain more accurate prediction values ​​of the vehicle speed, motor speed and battery power. At this time, the filter continuously updates the prediction results according to the current state and measurement data of the input, and the output vehicle speed, motor speed and battery power are "denoised" data, that is, smoothed data with noise removed, and finally obtain the vehicle driving noise-reduced speed, vehicle motor noise-reduced speed and vehicle battery noise-reduced power.

[0035] Step S152: constructing a vehicle dynamics model and a battery model corresponding to the new energy vehicle, and performing driving state correlation analysis on the vehicle driving noise reduction speed, the vehicle motor noise reduction speed, and the vehicle battery noise reduction power based on the vehicle dynamics model and the battery model, to obtain a new energy vehicle driving state correlation data set; In an embodiment of the present invention, based on the dynamic model and battery model of the new energy vehicle, it is necessary to analyze and process the previously obtained "noise reduction" data to further establish a vehicle driving state associated data set. Specifically, first, a dynamic model of the new energy vehicle is established, including relevant parameters such as vehicle acceleration, braking, traction, etc. Secondly, a battery model is constructed to describe the battery's charging and discharging characteristics, voltage, current and capacity, etc. Through these two models, the vehicle speed, motor speed and battery power can be combined to establish the relationship between the vehicle's driving state and its dynamics and battery state. For example, at a given driving speed and motor speed, the dynamic model is used to predict the vehicle's energy consumption rate; combined with the battery model, the battery power change and battery discharge efficiency are evaluated. Through these steps, a detailed vehicle driving state associated data set can be obtained, which includes the specific relationship between the various dynamic behaviors of the vehicle and the battery performance, and finally a new energy vehicle driving state associated data set is obtained.

[0036] Step S153: Obtain the corresponding vehicle driving road space location through the road slope and curvature corresponding to the new energy vehicle at different driving road locations, and perform driving space synchronization mapping on the new energy vehicle driving state associated data set based on the vehicle driving road space location to obtain the corresponding vehicle driving state data set at the same driving road location, including the driving speed, motor speed and battery power; In the embodiment of the present invention, the spatial position of the vehicle when driving is obtained by combining the driving position of the new energy vehicle on different road sections with the slope and curvature information of the road. To this end, the specific position of the vehicle can be obtained through a real-time positioning system (such as GPS) and road terrain data (including parameters such as road slope and curvature). At each position, the real-time driving state of the vehicle, especially the vehicle speed, motor speed and battery power of the vehicle, is calculated by combining the slope and curvature of the road. In this process, the slope and curvature of the road need to be determined first, and this information is used to correct the dynamic behavior and energy consumption of the vehicle. For example, when the vehicle is driving on a slope, additional traction is required, and the discharge rate of the battery will be different; and at a turn, the stability and driving strategy of the vehicle will be affected. By comprehensively considering these factors, an accurate driving data set at the same road position is obtained, including vehicle speed, motor speed and battery power, and these data are mapped to the corresponding spatial position, and finally the corresponding vehicle driving state data set at the same driving road position is obtained.

[0037] Step S154: Based on the road slope and curvature corresponding to the new energy vehicle at different driving road locations, vehicle data fusion processing is performed on the vehicle driving state data set corresponding to the same driving road location to obtain a new energy vehicle driving fusion data set.

[0038] In an embodiment of the present invention, the above-mentioned vehicle driving state data set is fused based on the slope and curvature information of the new energy vehicle at different driving road locations. Specifically, it is first necessary to extract information such as vehicle speed, motor speed and battery power from the driving data of different sections, and perform data weighted fusion in combination with the slope and curvature of the road. Through weighted fusion, the influence of the road can be comprehensively considered, the state values ​​in each data set can be adjusted, the deviation caused by the change of road conditions can be eliminated, and more accurate overall driving state data can be obtained. In the process of fusion processing, weighted averaging, Kalman filtering or other data fusion algorithms can be used to adjust the weight of each data point according to the influence of the road slope and curvature. For example, when the vehicle is driving on a steep slope or a section with a large curvature, the weights of the vehicle speed, motor speed and battery power need to be increased to reflect the actual energy consumption and performance status of the vehicle under these conditions. Through this method, the fused data set obtained not only contains accurate vehicle speed, motor speed and battery power data, but also reflects the influence of the road slope and curvature on the vehicle state, and finally obtains a new energy vehicle driving fusion data set.

[0039] Further, step S2 includes the following steps: Step S21: obtaining the current gear position of the new energy vehicle; In an embodiment of the present invention, the current gear position of the new energy vehicle is acquired in real time through the on-board sensor and control system. The vehicle's power transmission system is provided with a gear sensor that can accurately detect the current gear position of the vehicle. After the vehicle's on-board computer system receives the sensor signal, it will automatically identify the current gear position status and record it in the system for subsequent operation and evaluation. In this process, the acquisition of the gear position mainly depends on the vehicle's embedded electronic control unit (ECU) and its coordination with the transmission system. The goal of this step is to ensure real-time and accurate acquisition of the current gear position information as input for subsequent calculations and decisions, and ultimately obtain the current gear position of the new energy vehicle.

[0040] Step S22: determining the state space vector of the new energy vehicle by combining the current gear position of the new energy vehicle with the corresponding driving speed, motor speed, battery power, road slope and curvature in the new energy vehicle driving fusion data set at the current time, so as to obtain the state space vector of the new energy vehicle; In an embodiment of the present invention, by combining the driving fusion data set of the new energy vehicle, the vehicle speed, motor speed, battery power, road slope and curvature and other information at the current time are obtained. The new energy vehicle driving fusion data is obtained by the aforementioned steps, and combined with the communication system (such as GPS) to provide the vehicle driving road condition information, the battery power is calculated by the battery management system (BMS), and the road slope and curvature can be obtained through the high-precision map and GPS module. These real-time acquired values ​​are combined with the current gear to form a state space vector of the new energy vehicle. Specifically, the state space vector is composed of various state values ​​(vehicle speed, motor speed, battery power, road slope, curvature, etc.) arranged in a certain time sequence to form a multidimensional vector describing the current vehicle state, and finally the new energy vehicle state space vector is obtained.

[0041] Step S23: dividing the new energy vehicle state space vector into individual vehicle state discrete space vectors according to corresponding time intervals, and determining the corresponding vehicle speed state value, rotation speed state value, power state value, slope state value and curvature state value in each time range according to each vehicle state discrete space vector; In an embodiment of the present invention, the previously determined state space vector of the new energy vehicle is discretized according to a predetermined time interval. Specifically, the system divides the state space vector into discrete time periods according to the time step, for example, every few seconds or every few minutes is a time interval. The state space vector in each time interval represents the precise state of the vehicle at that moment. According to these discrete state space vectors, the corresponding vehicle speed state value, motor speed state value, battery power state value, road slope state value and curvature state value in each time interval are calculated. For vehicle speed, motor speed and battery power, the state value represents the measured value of these variables in a specific time period; for slope and curvature, the state value represents the environmental information of the vehicle's driving route. This discretization process enables the dynamic behavior of the vehicle to be quantified and further evaluated and decided, and finally determines the corresponding vehicle speed state value, speed state value, power state value, slope state value and curvature state value in each time range.

[0042] Step S24: using the vehicle speed state value, rotation speed state value, power state value, slope state value and curvature state value corresponding to each time range as the state input corresponding to the preset deep reinforcement learning model, and performing gear power and energy consumption evaluation calculation on the current gear of the new energy vehicle based on the vehicle speed state value, rotation speed state value, power state value, slope state value and curvature state value corresponding to each time range, to obtain the gear power performance state value of the new energy vehicle and the gear energy loss of the new energy vehicle; In an embodiment of the present invention, based on the previously obtained vehicle speed state value, rotation speed state value, power state value, slope state value and curvature state value as the input of the deep reinforcement learning model, a preset deep reinforcement learning model is applied to evaluate the power and energy consumption of the current gear of the new energy vehicle. In specific implementation, the deep reinforcement learning model is designed to obtain the power performance state value and energy loss value of the gear according to the real-time state input of the vehicle (i.e., each state value) through a multi-layer neural network. The power performance state value mainly reflects the power output capacity of the vehicle in the current gear, and evaluates whether the vehicle can meet the current driving conditions. The energy loss reflects the energy consumption in the current gear, taking into account factors such as vehicle power demand, driving environment (such as slope and curvature) and battery efficiency. This process requires training the deep reinforcement learning model so that it can output the corresponding gear power performance and energy consumption evaluation results according to the state space vector of the vehicle, and finally obtain the gear power performance state value of the new energy vehicle and the gear energy loss of the new energy vehicle.

[0043] Step S25: Use the deep reinforcement learning model to define the gear action space of the new energy vehicle gear power performance state value and the new energy vehicle gear energy loss to generate the new energy vehicle gear switching action space selection, including the corresponding upshift, downshift and maintain current gear actions of the new energy vehicle.

[0044] In an embodiment of the present invention, the gear switching action space of the new energy vehicle is defined according to the gear power performance state value and energy loss obtained by the above calculation. First, the model uses the power performance state value and energy loss as the basis for decision-making, and sets the action space to upshift, downshift and maintain the current gear. Specifically, when the power performance state value is large and the gear energy loss is within an acceptable range, the deep reinforcement learning model determines that the vehicle is suitable for upshifting to improve driving efficiency; when the power performance state value is small and the energy loss is large, the model determines that the gear should be downshifted to reduce unnecessary energy consumption; when the power performance and energy loss reach a balance, the model determines that the current gear should be maintained to avoid additional energy waste caused by frequent gear switching. At each decision, the model will evaluate the current power demand, energy consumption and driving environment, and select the optimal gear action. Through this decision-making process based on deep reinforcement learning, it is possible to realize intelligent control of gear switching under different driving conditions, optimize the energy efficiency and driving experience of new energy vehicles, and finally generate a gear switching action space selection for new energy vehicles, including upshifting, downshifting and maintaining the current gear actions corresponding to new energy vehicles.

[0045] Further, the gear power and energy consumption evaluation calculation for the current gear of the new energy vehicle based on the corresponding vehicle speed state value, rotation speed state value, power state value, slope state value and curvature state value in each time range in step S24 includes the following steps: Based on the corresponding vehicle speed state value, rotation speed state value and power state value in each time range, the gear power performance evaluation calculation formula is used to perform gear power evaluation calculation on the current gear of the new energy vehicle to obtain the gear power performance state value of the new energy vehicle; In an embodiment of the present invention, a suitable gear power performance evaluation calculation formula is constructed by combining the initial value of the time range, the end value of the time range, the vehicle speed state value, the rotational speed state value, the power state value, the maximum vehicle speed of the new energy vehicle in the current gear, the maximum motor speed, the maximum battery power, the vehicle speed power contribution coefficient, the motor speed power contribution coefficient, the battery power contribution coefficient, the power decay time constant, the time exponential decay factor and related parameters to perform gear power evaluation calculation, so as to quantitatively calculate the power performance state value of the vehicle in the current gear, that is, the energy output capacity of the vehicle in this gear, and finally obtain the gear power performance state value of the new energy vehicle.

[0046] Preferably, the vehicle gear power output of the new energy vehicle in the corresponding time range is simulated and calculated according to the gear power performance state value of the new energy vehicle, and the vehicle gear power peak value corresponding to the new energy vehicle is calculated according to the vehicle gear power output; In an embodiment of the present invention, the gear power performance state value calculated based on the previous step is combined with the current gear of the vehicle to simulate and calculate the vehicle gear power output of the new energy vehicle within the time range. The vehicle gear power output usually represents the power output value that the vehicle can provide in a specific gear. The calculation of the output needs to be combined with the vehicle's power transmission system and real-time control system. The specific calculation method can be performed through dynamic simulation software. According to the change in power output, the peak power of the vehicle in the current gear is further calculated. This process mainly depends on the actual driving state of the vehicle and the power output curve. The simulation calculation needs to comprehensively consider the vehicle speed change, acceleration and load conditions, so as to determine the maximum power output at different time points, and finally obtain the vehicle gear power peak corresponding to the new energy vehicle.

[0047] Preferably, the energy consumption impact analysis of the current gear of the new energy vehicle is performed based on the slope state value and the curvature state value to obtain the energy consumption impact coefficient of the gear of the new energy vehicle; In an embodiment of the present invention, by obtaining the current slope state value and curvature state value, which are usually collected and calculated in real time by the vehicle-mounted sensors and navigation system, the slope state value describes the slope of the road section where the vehicle is located, and the curvature state value reflects the curvature of the vehicle's driving route. Based on these data, the energy consumption impact analysis of the current gear of the new energy vehicle is performed, and the influence of the slope and curvature on the vehicle's energy consumption is analyzed through a mathematical model. When the slope is large, the vehicle needs to overcome greater gravity resistance. When the curvature is large, the vehicle needs more power to maintain steering stability. Therefore, these factors will affect the energy consumption of the vehicle in the current gear. Combined with these analysis results, a gear energy consumption impact coefficient is obtained. The coefficient specifically represents the rate of change of energy consumption of the vehicle in the gear under a specific driving environment, that is, the coefficient is specifically ,in is the slope state value, is the energy consumption of the vehicle in this gear, is the total energy consumption of the vehicle during the entire driving process, is the curvature state value, and finally the energy consumption influence coefficient of the gear position of the new energy vehicle is obtained.

[0048] Preferably, based on the vehicle gear power peak and the new energy vehicle gear energy consumption influence coefficient, the vehicle energy loss evaluation calculation formula is used to evaluate and calculate the energy consumption of the current gear of the new energy vehicle to obtain the gear energy loss of the new energy vehicle; In an embodiment of the present invention, a suitable vehicle energy loss assessment calculation formula is formed by combining the time range initial value, the time range end value, the time range end value, the vehicle speed state value, the rotation speed state value, the power state value, the power energy loss influence weight coefficient, the vehicle gear power peak value, the new energy vehicle gear energy consumption influence coefficient and related parameters to perform energy loss assessment calculation, so as to quantitatively calculate the energy loss value of the new energy vehicle in the current gear, and finally obtain the gear energy loss of the new energy vehicle.

[0049] Among them, the vehicle energy loss assessment calculation formula is as follows: ; In the formula, The energy loss of new energy vehicles during gear shifting. is the initial value of the time range, is the end value of the time range, is the time variable parameter, For new energy vehicles in time The vehicle speed state value at For new energy vehicles in time The speed state value at For new energy vehicles in time The power state value at is the weight coefficient of power energy loss, is the peak power of the vehicle gear, is the energy consumption impact coefficient of the gear position of new energy vehicles, It is the correction coefficient for energy loss during gear shift of new energy vehicles.

[0050] The present invention obtains a vehicle energy loss evaluation calculation formula by using a specific mathematical model and verifying it, which is used to evaluate and calculate the energy consumption of the current gear of the new energy vehicle. The vehicle energy loss evaluation calculation formula can more accurately simulate the energy consumption of the new energy vehicle in actual driving by considering the dynamic states of the vehicle speed, rotation speed, and power within different time ranges. It is not only based on static data, but also evaluates the energy consumption by considering the dynamic performance (such as vehicle speed, rotation speed) and power changes of the vehicle at a specific time point, which can help analyze the energy consumption of the vehicle under actual working conditions and better fit the actual driving scene. The formula takes into account the relationship between multiple factors such as the vehicle's power peak, vehicle speed, rotation speed, and power state. Through the comprehensive calculation of these variables, a more accurate energy consumption evaluation can be obtained, avoiding the one-sidedness of the energy consumption evaluation of a single parameter. For example, vehicle speed and rotation speed directly affect the power output, while the power state affects the actual consumption of the battery. In this formula, the time variable and weighting factor reflect the weighted influence of the state at different moments in the time series. This weighting takes into account the change of energy consumption over time, making the evaluation of vehicle energy consumption more accurate, especially for working conditions where energy consumption is relatively concentrated or concentrated in a specific period of time. In this way, the actual energy consumption of new energy vehicles under different working conditions can be analyzed in a more detailed manner. Through the analysis of the influence of the slope state value and the curvature state value on vehicle energy consumption, the formula can capture the impact of the driving environment on energy consumption. For example, when going uphill, the vehicle needs more power to overcome gravity, resulting in increased energy consumption; and the road section with large curvature will cause frequent acceleration and deceleration, which will affect the power consumption of the vehicle. The energy consumption influence coefficient in the formula reflects the impact of these external factors on energy consumption, thereby evaluating energy consumption more comprehensively. In addition, the introduction of the correction coefficient can make fine adjustments to the calculation process in practical applications to ensure that the calculation results are more in line with the actual situation. Through this correction coefficient, the deviation between the theoretical calculation and the actual test can be corrected to make the evaluation results more accurate. In summary, this formula can fully consider the energy loss of new energy vehicles in gears. , time range initial value , time range end value , time variable parameter , new energy vehicles in time The vehicle speed status value at , new energy vehicles in time Speed ​​state value at , new energy vehicles in time The power status value at , power energy loss influences weight coefficient , vehicle gear power peak , the influence coefficient of energy consumption of new energy vehicle gear , Correction coefficient of energy loss in gear of new energy vehicles , according to the energy loss of new energy vehicles The correlation between the above parameters constitutes a functional relationship This formula can realize the energy consumption evaluation and calculation process of the current gear of new energy vehicles. At the same time, the correction coefficient of energy loss of new energy vehicles is The introduction of can be adjusted according to the errors that occur during the calculation process, thereby improving the accuracy and applicability of the vehicle energy loss assessment calculation formula.

[0051] Furthermore, the gear dynamic performance evaluation calculation formula is specifically as follows: ; In the formula, is the gear power performance status value of new energy vehicles, is the initial value of the time range, is the end value of the time range, is the time variable parameter, For new energy vehicles in time The vehicle speed state value at is the maximum speed of the new energy vehicle in the current gear, is the vehicle speed power contribution coefficient, For new energy vehicles in time The speed state value at is the maximum motor speed of the new energy vehicle in the current gear, is the motor speed power contribution coefficient, For new energy vehicles in time The power state value at It is the maximum battery capacity of the new energy vehicle in the current gear. is the battery power contribution coefficient, is the power attenuation time constant of the new energy vehicle in the current gear, is the time exponential decay factor, It is the correction coefficient of the gear power performance status value of new energy vehicles.

[0052] The present invention obtains a gear power performance evaluation calculation formula by using a specific mathematical model and after verification, which is used to perform gear power evaluation calculation for the current gear of a new energy vehicle. The gear power performance evaluation calculation formula takes into account multiple key parameters, including vehicle speed, rotation speed and power state. Through this multi-dimensional evaluation, the power performance of new energy vehicles in a specific gear can be more accurately reflected, avoiding the error that may be caused by a single indicator. For example, the influence of vehicle speed and rotation speed reflects the actual effect of power output, while the power state represents the available energy of the battery. Time integral is used in the formula, and a time decay factor and a power decay time constant are introduced, which enables the formula to consider the change of time and the decay effect of the battery / motor. For electric vehicles, as the use time increases, the battery power and motor efficiency may decay, so this weighted time decay can better simulate the actual power performance. By proportionalizing the vehicle speed, rotation speed and power state to their maximum values, the formula can accurately reflect the actual contribution of each parameter. This processing method helps to avoid the performance differences of different vehicles under different parameters, making the evaluation process universal and comparable. The energy efficiency of electric vehicles is usually closely related to power output. Through the various coefficients in the formula, the contribution of vehicle speed, rotation speed, and battery status to power output can be quantified. The evaluation results provide data support for subsequent energy consumption analysis and help optimize energy efficiency management. By analyzing the vehicle's power output and influence coefficients in subsequent steps, the energy loss of electric vehicles under different driving conditions can be predicted. This is of great guiding significance for the optimization of energy management systems, especially under long-term high load or low battery conditions. The formula includes multiple adjustment coefficients, which can be personalized according to specific driving conditions. Therefore, this calculation method can not only provide power performance evaluation for different models, but also can be adjusted and optimized in real time according to different driving environments, road conditions and battery status, thereby improving the accuracy of the evaluation results. In summary, this formula fully considers the power performance status value of the gear of new energy vehicles. , time range initial value , time range end value , time variable parameter , new energy vehicles in time The vehicle speed status value at , the maximum speed of new energy vehicles in the current gear , vehicle speed power contribution coefficient , new energy vehicles in time Speed ​​state value at , the maximum motor speed of new energy vehicles in the current gear , motor speed power contribution coefficient , new energy vehicles in time The power status value at , the maximum battery capacity of a new energy vehicle in the current gear , battery power contribution coefficient , the power decay time constant of new energy vehicles in the current gear , time exponential decay factor , correction coefficient of the gear dynamic performance state value of new energy vehicles , according to the gear power performance status value of new energy vehicles The correlation between the above parameters constitutes a functional relationship This formula can realize the gear power evaluation calculation process of the current gear of the new energy vehicle. At the same time, the correction coefficient of the gear power performance state value of the new energy vehicle is The introduction of can be adjusted according to the error conditions that occur during the calculation process, thereby improving the accuracy and applicability of the gear dynamic performance evaluation calculation formula.

[0053] Furthermore, the gear action space definition described in step S25 is specifically that when the power performance state value is large and the gear energy loss is in an acceptable range, the deep reinforcement learning model is used to define the action space corresponding to the new energy vehicle as an upshift action; when the power performance state value is insufficient and the gear energy loss is large, the deep reinforcement learning model is used to define the action space corresponding to the new energy vehicle as a downshift action; when the power performance state value and the gear energy loss are in a balanced state, the deep reinforcement learning model is used to define the action space corresponding to the new energy vehicle as a maintain current gear action.

[0054] Furthermore, the vehicle gear decision reward function described in step S3 is specifically that when the action gear selected by the deep reinforcement learning model results in reduced energy consumption and enhanced power performance, a positive reward is given to the gear decision; when the action gear selected by the deep reinforcement learning model results in excessive energy consumption and weakened power performance to affect the driving smoothness of the new energy vehicle, a negative reward is given to the gear decision.

[0055] Furthermore, the optimal gear decision control strategy corresponding to different driving conditions described in step S4 is specifically that when the vehicle is traveling on congested roads, the model determines whether it is necessary to downshift to maintain the corresponding power output and smoothness of the vehicle based on the vehicle speed and insufficient motor speed; when the vehicle is traveling on high-speed roads, the model determines whether it is necessary to upshift to reduce the motor speed and improve energy utilization efficiency based on sufficient battery power and excessive energy consumption; and when the vehicle is traveling on smooth roads, the model determines whether it is necessary to maintain the current gear to maintain the smooth operation of the vehicle and avoid unnecessary power fluctuations and motor losses based on normal battery power consumption and the fact that gear switching will not bring about obvious power improvement or energy consumption reduction.

[0056] Furthermore, the present invention also provides a new energy vehicle gear decision control system based on deep reinforcement learning, which is used to execute the new energy vehicle gear decision method based on deep reinforcement learning as described above. The new energy vehicle gear decision control system based on deep reinforcement learning includes: The vehicle driving status data fusion module is used to deploy a vehicle speed sensor, a motor speed sensor and a battery status sensor on the new energy vehicle, and use the vehicle speed sensor, the motor speed sensor and the battery status sensor to collect the vehicle status of the new energy vehicle in real time during the real-time driving process, so as to obtain the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle; obtain the road slope and curvature corresponding to the new energy vehicle at different driving road locations, and perform vehicle data fusion processing on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle based on the road slope and curvature corresponding to the new energy vehicle at different driving road locations, so as to obtain the new energy vehicle driving fusion data set; The vehicle gear action model analysis module is used to obtain the current gear of the new energy vehicle and determine the state space vector in combination with the corresponding driving speed, motor speed, battery power, road slope and curvature in the new energy vehicle driving fusion data set at the current time to obtain the new energy vehicle state space vector; the new energy vehicle state space vector is used as the state input corresponding to the preset deep reinforcement learning model, and the deep reinforcement learning model is used to define the gear action space of the new energy vehicle state space vector to generate the new energy vehicle gear switching action space selection, including the corresponding upshift, downshift and keep current gear actions of the new energy vehicle; The model decision reward function design module is used to design the corresponding vehicle gear decision reward function through the gear switching action space selection of the new energy vehicle corresponding to the deep reinforcement learning model; The vehicle gear decision control module is used to perform vehicle gear decision control on the corresponding new energy vehicle state space vector based on the vehicle gear decision reward function using a deep reinforcement learning model, so as to select the optimal gear switching action from the action space corresponding to the model according to the real-time collected vehicle state information, thereby generating the optimal gear decision control strategy corresponding to the new energy vehicle under different driving conditions.

[0057] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0058] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A new energy vehicle gear decision method based on deep reinforcement learning, characterized in that: The following steps are involved: Step S1: by deploying a vehicle speed sensor, a motor speed sensor and a battery status sensor on the new energy vehicle, and using the vehicle speed sensor, the motor speed sensor and the battery status sensor to collect the vehicle status of the real-time driving process of the new energy vehicle in real time, so as to obtain the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle; obtaining the road slope and curvature corresponding to the new energy vehicle at different driving road locations, and performing vehicle data fusion processing on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle based on the road slope and curvature corresponding to the new energy vehicle at different driving road locations, so as to obtain a new energy vehicle driving fusion data set; Step S2: obtaining the current gear of the new energy vehicle and determining the state space vector in combination with the corresponding driving speed, motor speed, battery power, road slope and curvature in the new energy vehicle driving fusion data set at the current time, so as to obtain the state space vector of the new energy vehicle; using the state space vector of the new energy vehicle as the state input corresponding to the preset deep reinforcement learning model, and using the deep reinforcement learning model to define the gear action space of the state space vector of the new energy vehicle, so as to generate the gear switching action space selection of the new energy vehicle, including the corresponding upshift, downshift and keeping current gear actions of the new energy vehicle; Step S3: Designing a corresponding vehicle gear decision reward function by selecting the gear switching action space of the new energy vehicle corresponding to the deep reinforcement learning model; Step S4: Based on the vehicle gear decision reward function, the deep reinforcement learning model is used to perform vehicle gear decision control on the corresponding new energy vehicle state space vector, so as to select the optimal gear switching action from the action space corresponding to the model according to the real-time collected vehicle state information, and generate the optimal gear decision control strategy corresponding to the new energy vehicle under different driving conditions.

2. The gear decision method for new energy vehicles based on deep reinforcement learning according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: by deploying a vehicle speed sensor, a motor speed sensor and a battery status sensor on the new energy vehicle, and using the vehicle speed sensor to collect the driving speed of the new energy vehicle in real time during the real-time driving process, so as to obtain the driving speed of the new energy vehicle; Step S12: using a motor speed sensor to collect the motor speed of the new energy vehicle in real time during the real-time driving process to obtain the motor speed of the new energy vehicle; Step S13: using a battery status sensor to collect the battery power of the new energy vehicle in real time during the real-time driving process to obtain the battery power of the new energy vehicle; Step S14: Obtaining the road slope and curvature corresponding to different driving road locations of the new energy vehicle; Step S15: Based on the road slope and curvature corresponding to different driving road locations of the new energy vehicle, the vehicle data fusion processing is performed on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle to obtain a new energy vehicle driving fusion data set.

3. The gear decision method for new energy vehicles based on deep reinforcement learning according to claim 2 is characterized in that: Step S15 includes the following steps: Step S151: Perform vehicle data Kalman filtering on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle to obtain the driving noise reduction speed of the vehicle, the motor noise reduction speed of the vehicle and the battery noise reduction power of the vehicle; Step S152: constructing a vehicle dynamics model and a battery model corresponding to the new energy vehicle, and performing driving state correlation analysis on the vehicle driving noise reduction speed, the vehicle motor noise reduction speed, and the vehicle battery noise reduction power based on the vehicle dynamics model and the battery model, to obtain a new energy vehicle driving state correlation data set; Step S153: Obtain the corresponding vehicle driving road space location through the road slope and curvature corresponding to the new energy vehicle at different driving road locations, and perform driving space synchronization mapping on the new energy vehicle driving state associated data set based on the vehicle driving road space location to obtain the corresponding vehicle driving state data set at the same driving road location, including the driving speed, motor speed and battery power; Step S154: Based on the road slope and curvature corresponding to the new energy vehicle at different driving road locations, vehicle data fusion processing is performed on the vehicle driving state data set corresponding to the same driving road location to obtain a new energy vehicle driving fusion data set.

4. The gear decision method for new energy vehicles based on deep reinforcement learning according to claim 1 is characterized in that: Step S2 includes the following steps: Step S21: obtaining the current gear position of the new energy vehicle; Step S22: determining the state space vector of the new energy vehicle by combining the current gear position of the new energy vehicle with the corresponding driving speed, motor speed, battery power, road slope and curvature in the new energy vehicle driving fusion data set at the current time, so as to obtain the state space vector of the new energy vehicle; Step S23: dividing the new energy vehicle state space vector into individual vehicle state discrete space vectors according to corresponding time intervals, and determining the corresponding vehicle speed state value, rotation speed state value, power state value, slope state value and curvature state value in each time range according to each vehicle state discrete space vector; Step S24: using the vehicle speed state value, rotation speed state value, power state value, slope state value and curvature state value corresponding to each time range as the state input corresponding to the preset deep reinforcement learning model, and performing gear power and energy consumption evaluation calculation on the current gear of the new energy vehicle based on the vehicle speed state value, rotation speed state value, power state value, slope state value and curvature state value corresponding to each time range, to obtain the gear power performance state value of the new energy vehicle and the gear energy loss of the new energy vehicle; Step S25: Use the deep reinforcement learning model to define the gear action space of the new energy vehicle gear power performance state value and the new energy vehicle gear energy loss to generate the new energy vehicle gear switching action space selection, including the corresponding upshift, downshift and maintain current gear actions of the new energy vehicle.

5. The new energy vehicle gear decision method based on deep reinforcement learning according to claim 4 is characterized in that: The step S24 includes the following steps: performing gear power and energy consumption evaluation calculation for the current gear of the new energy vehicle based on the corresponding vehicle speed state value, rotation speed state value, power state value, slope state value, and curvature state value in each time range: Based on the corresponding vehicle speed state value, rotation speed state value and power state value in each time range, the gear power performance evaluation calculation formula is used to perform gear power evaluation calculation on the current gear of the new energy vehicle to obtain the gear power performance state value of the new energy vehicle; The vehicle gear power output of the new energy vehicle in the corresponding time range is simulated and calculated according to the gear power performance state value of the new energy vehicle, and the vehicle gear power peak value corresponding to the new energy vehicle is calculated according to the vehicle gear power output; Based on the slope state value and the curvature state value, the energy consumption impact of the current gear of the new energy vehicle is analyzed to obtain the energy consumption impact coefficient of the gear of the new energy vehicle; Based on the vehicle gear power peak and the new energy vehicle gear energy consumption impact coefficient, the vehicle energy loss evaluation calculation formula is used to evaluate and calculate the energy consumption of the current gear of the new energy vehicle, and the gear energy loss of the new energy vehicle is obtained; Among them, the vehicle energy loss assessment calculation formula is as follows: ； In the formula, The energy loss of new energy vehicles during gear shifting. is the initial value of the time range, is the end value of the time range, is the time variable parameter, For new energy vehicles in time The vehicle speed state value at For new energy vehicles in time The speed state value at For new energy vehicles in time The power state value at is the weight coefficient of power energy loss, is the peak power of the vehicle gear, is the energy consumption impact coefficient of the gear position of new energy vehicles, It is the correction coefficient for energy loss during gear shift of new energy vehicles.

6. The gear decision method for new energy vehicles based on deep reinforcement learning according to claim 5 is characterized in that: The gear dynamic performance evaluation calculation formula is specifically as follows: ； In the formula, is the gear power performance status value of new energy vehicles, is the initial value of the time range, is the end value of the time range, is the time variable parameter, For new energy vehicles in time The vehicle speed state value at is the maximum speed of the new energy vehicle in the current gear, is the vehicle speed power contribution coefficient, For new energy vehicles in time The speed state value at is the maximum motor speed of the new energy vehicle in the current gear, is the motor speed power contribution coefficient, For new energy vehicles in time The power state value at It is the maximum battery capacity of the new energy vehicle in the current gear. is the battery power contribution coefficient, is the power attenuation time constant of the new energy vehicle in the current gear, is the time exponential decay factor, It is the correction coefficient of the gear power performance status value of new energy vehicles.

7. The new energy vehicle gear decision method based on deep reinforcement learning according to claim 4 is characterized in that: The gear action space definition described in step S25 is specifically as follows: when the power performance state value is large and the gear energy loss is within an acceptable range, the deep reinforcement learning model is used to define the action space corresponding to the new energy vehicle as an upshift action; when the power performance state value is insufficient and the gear energy loss is large, the deep reinforcement learning model is used to define the action space corresponding to the new energy vehicle as a downshift action; when the power performance state value and the gear energy loss are in a balanced state, the deep reinforcement learning model is used to define the action space corresponding to the new energy vehicle as a maintain current gear action.

8. The new energy vehicle gear decision method based on deep reinforcement learning according to claim 1 is characterized in that: The vehicle gear decision reward function described in step S3 is specifically that when the action gear selected by the deep reinforcement learning model results in reduced energy consumption and enhanced power performance, a positive reward is given to the gear decision; when the action gear selected by the deep reinforcement learning model results in excessive energy consumption and weakened power performance to affect the driving smoothness of the new energy vehicle, a negative reward is given to the gear decision.

9. The new energy vehicle gear decision method based on deep reinforcement learning according to claim 1 is characterized in that: The optimal gear decision control strategy corresponding to different driving conditions described in step S4 is specifically that when the vehicle is driving on congested roads, the model determines whether it is necessary to downshift to maintain the corresponding power output and smoothness of the vehicle based on the vehicle speed and insufficient motor speed; when the vehicle is driving on high-speed roads, the model determines whether it is necessary to upshift to reduce the motor speed and improve energy utilization efficiency based on sufficient battery power and excessive energy consumption; and when the vehicle is driving on smooth roads, the model determines whether it is necessary to maintain the current gear to maintain the smooth operation of the vehicle and avoid unnecessary power fluctuations and motor losses based on normal battery power consumption and the fact that gear switching will not bring obvious power improvement or energy consumption reduction.

10. A new energy vehicle gear position decision control system based on deep reinforcement learning, characterized in that: Used to execute the new energy vehicle gear decision method based on deep reinforcement learning as claimed in claim 1, the new energy vehicle gear decision control system based on deep reinforcement learning includes: The vehicle driving status data fusion module is used to deploy a vehicle speed sensor, a motor speed sensor and a battery status sensor on the new energy vehicle, and use the vehicle speed sensor, the motor speed sensor and the battery status sensor to collect the vehicle status of the new energy vehicle in real time during the real-time driving process, so as to obtain the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle; obtain the road slope and curvature corresponding to the new energy vehicle at different driving road locations, and perform vehicle data fusion processing on the driving speed of the new energy vehicle, the motor speed of the new energy vehicle and the battery power of the new energy vehicle based on the road slope and curvature corresponding to the new energy vehicle at different driving road locations, so as to obtain the new energy vehicle driving fusion data set; The vehicle gear action model analysis module is used to obtain the current gear of the new energy vehicle and determine the state space vector in combination with the corresponding driving speed, motor speed, battery power, road slope and curvature in the new energy vehicle driving fusion data set at the current time to obtain the new energy vehicle state space vector; the new energy vehicle state space vector is used as the state input corresponding to the preset deep reinforcement learning model, and the deep reinforcement learning model is used to define the gear action space of the new energy vehicle state space vector to generate the new energy vehicle gear switching action space selection, including the corresponding upshift, downshift and keep current gear actions of the new energy vehicle; The model decision reward function design module is used to design the corresponding vehicle gear decision reward function through the gear switching action space selection of the new energy vehicle corresponding to the deep reinforcement learning model; The vehicle gear decision control module is used to perform vehicle gear decision control on the corresponding new energy vehicle state space vector based on the vehicle gear decision reward function using a deep reinforcement learning model, so as to select the optimal gear switching action from the action space corresponding to the model according to the real-time collected vehicle state information, thereby generating the optimal gear decision control strategy corresponding to the new energy vehicle under different driving conditions.

Images