A double-layer brake energy recovery method based on working condition and braking intention recognition

Abstract

This invention relates to the field of composite braking and energy recovery technology for electric vehicles, and particularly to a two-layer braking energy recovery method based on operating condition and braking intention recognition. The method includes an upper decision-making layer and a lower execution layer. The upper decision-making layer performs operating condition recognition and braking intention recognition. The lower execution layer, based on the upper decision-making results, uses a corresponding optimization algorithm to allocate power between the front and rear electric motors within a preset safe braking area, thereby achieving braking energy recovery. This invention achieves accurate recognition of operating conditions and braking intentions through a two-layer architecture, employing differentiated control strategies: under simple operating conditions, a hybrid optimization algorithm seeks the globally optimal solution; under complex operating conditions, a neural network model maintains stable recovery efficiency; and under coasting conditions, the recovery window is fully explored. Combined with the optimal power allocation model for the front and rear electric motors, the power generation potential of the dual motors is maximized, resulting in a significant improvement in energy recovery efficiency compared to existing technologies, a substantial extension of the electric vehicle's driving range, and a reduction in overall vehicle energy consumption costs.

Description

Technical Field

[0001] This invention relates to the field of composite braking and energy recovery technology for electric vehicles, and in particular to a two-layer braking energy recovery method based on the recognition of operating conditions and braking intentions. Background Technology

[0002] With the rapid development of the electric vehicle industry, regenerative braking technology has become one of the key technologies for improving vehicle range and reducing energy consumption. Most existing regenerative braking solutions are based on fixed rules for braking force distribution according to brake pedal signals, without fully considering the differences in driving conditions and the diversity of driver braking intentions.

[0003] Regarding driving conditions, the system characteristics differ significantly between simple driving conditions such as highways and suburbs and complex driving conditions such as urban congestion. Under simple driving conditions, the system nonlinearity is weak, while under complex driving conditions, the system nonlinearity is strong and there are many disturbances. Existing solutions use a uniform control strategy, resulting in poor adaptability under different driving conditions and difficulty in achieving optimal energy recovery efficiency. In terms of braking intent, existing solutions often only focus on conventional braking (driver pressing the brake pedal) and ignore the important energy recovery window of coasting braking (driver releasing the accelerator pedal). At the same time, conventional braking does not accurately distinguish the degree of braking urgency, resulting in unreasonable braking force distribution.

[0004] Furthermore, in traditional energy recovery systems, the intervention of electric braking is often abrupt, easily producing a "pulling sensation" that affects driving comfort. Moreover, the distribution of braking force does not strictly adhere to safety regulations, posing potential braking safety hazards. Therefore, how to balance braking safety, energy recovery efficiency, and driving comfort has become a problem that existing braking energy recovery technologies need to solve. Summary of the Invention

[0005] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description and other accompanying drawings.

[0006] The purpose of this invention is to overcome the above-mentioned shortcomings and provide a two-layer braking energy recovery method based on the recognition of operating conditions and braking intention.

[0007] To achieve the above objectives, the technical solution of the present invention is: a two-layer braking energy recovery method based on operating condition and braking intention recognition, comprising an upper decision layer and a lower execution layer. The upper decision layer performs operating condition recognition and braking intention recognition, and the lower execution layer, based on the upper decision result, uses a corresponding optimization algorithm to allocate the power of the front and rear motors within a preset safe braking area to realize braking energy recovery.

[0008] In some embodiments, condition identification includes the following steps: S1. Extract the feature vector within the time window of real-time driving data. The feature vector includes, but is not limited to, average vehicle speed. v Maximum speed vmax Average acceleration a Maximum acceleration amaxa Average deceleration adec Maximum deceleration amaxd Parking time percentage Pstop ; S2. A large number of driving cycle segments are pre-divided into simple and complex driving conditions using K-means clustering analysis, and standard feature vectors for each type of driving condition are established. Fsimple and Fcomplex ; S3. Calculate real-time feature vectors Freal Compared with standard feature vectors Fsimple , Fcomplex The Euclidean distances are respectively Dsimple and Dcomplex The current working condition type is determined based on the distance. Dsimple =∥ Freal - Fsimple ∥, Dcomplex =∥ Freal - Fcomplex ∥.

[0009] In some embodiments, braking intent recognition includes conventional braking intent recognition and coasting braking intent recognition, wherein... Conventional braking intent recognition: based on brake pedal opening θb and brake pedal change rate The compensation coefficient is output through the fuzzy controller. k 0, the initial desired braking intensity can be obtained by looking up a table based on the pedal opening. z Compensation is applied to 0 to obtain the final desired braking strength. z = k 0 z 0, and classify braking levels according to the value of Z; Coasting braking intention recognition: based on accelerator pedal opening θa and its rate of change The urgency of coasting braking intention is identified using a fuzzy controller. Icoast urgency Icoast Determine the motor intervention delay time tA .

[0010] In some embodiments, braking levels include light braking, moderate braking, and emergency braking, wherein,z <0.3 indicates light braking; 0.3≤ z ≤0.8 indicates moderate braking. z >0.8 indicates emergency braking.

[0011] In some embodiments, the motor intervention delay time tA The time is 0.5s-3s, during emergency coasting braking. tA =0.5s, intended to be smooth tA =3s.

[0012] In some embodiments, the preset safe braking zone is constructed based on the ECE R13 regulation and the I curve. For any desired braking intensity Z, the intersection points of the I curve and the ECE R13 regulation line are M and N, respectively, and the line segment MN is the range of total braking force distribution between the front and rear axles under this braking intensity.

[0013] In some embodiments, when the upper-level decision is a simple working condition, the lower-level execution layer adopts a hybrid optimization strategy that combines Globalsearch and Fmincon constrained nonlinear programming local solvers. Based on the braking intensity, the braking stage is divided, a dynamic safety boundary for the regenerative braking ratio coefficient is established, an energy recovery efficiency optimization problem is constructed, and the global optimal solution is solved.

[0014] In some embodiments, when the upper-level decision involves a complex operating condition, the lower-level execution layer employs a neural network inverse model for adaptive control. The input layer of the neural network inverse model includes vehicle speed. v Braking strength z Road slope α The inverse neural network model has two hidden layers and uses a sigmoid activation function. The output layer of the inverse neural network model represents the ratio of regenerative braking torque to total braking torque. The Levenberg-Marquardt backpropagation algorithm was used for training, with mean squared error as the loss function.

[0015] In some embodiments, when the upper layer decision is a coasting braking intention, the lower execution layer controls the motor power based on the expected coasting distance, and the specific steps are as follows: A1. The vehicle speed is [missing information] when the driver releases the pedal. v 0, after delay time tA The speed of the car then decreased vA Electric motor power Fe With deceleration ae intervention; A2. Based on vehicle dynamics, considering the deceleration during motorless braking... a Deducing the braking deceleration of the electric motor through geometric relationships ae for: ;in vBThese are empirical parameters.

[0016] A3, according to Fe = m ae Calculate the total power of the electric motor and distribute it to the front and rear motors; among which, m It's about vehicle quality. ae This is used for braking and deceleration of the motor.

[0017] In some embodiments, the lower execution layer solves for the power allocation coefficients of the pre- and post-electric motors through an optimization model. The optimization model aims to maximize the overall system recovery efficiency. Specifically: , The corresponding constraints are: By solving this model, the optimal allocation coefficients can be obtained. This maximizes energy recovery within the safety boundary; among which, For overall system efficiency, T 1. T 2 represents the braking torque of the front and rear motors, respectively. n 1. n 2 represents the speed of the front and rear motors, respectively. η 1. η 2 represents the power generation efficiency of the front and rear motors, respectively. p This is the power distribution coefficient for the front axle electric motor. Fmf - max , Fmr - max These represent the maximum braking force of the front and rear motors, respectively. G For vehicle weight, r For the wheel radius, Fbf , Fbr These are the mechanical braking forces of the front and rear axles, respectively.

[0018] By adopting the above technical solution, the beneficial effects of the present invention are: 1. Ensure braking safety and comply with regulatory requirements: Based on ECE R13 regulations and the I-curve, a closed safety braking zone is constructed, within which all braking force distribution strategies are executed, strictly meeting braking stability and industry regulatory standards. During emergency braking, the regenerative braking ratio is dynamically adjusted to prioritize the safety redundancy of mechanical braking, completely avoiding the risk of sacrificing braking safety in pursuit of energy recovery, thus providing dual protection for the vehicle's braking safety.

[0019] 2. Improve energy recovery efficiency and optimize vehicle economy: A dual-layer architecture enables precise identification of operating conditions and braking intentions, employing differentiated control strategies: under simple operating conditions, a hybrid optimization algorithm seeks the globally optimal solution; under complex operating conditions, a neural network model maintains stable energy recovery efficiency; and under coasting conditions, the energy recovery window is fully exploited. Combined with an optimal power allocation model for the front and rear electric motors, the power generation potential of both motors is maximized. Compared to existing technologies, energy recovery efficiency is significantly improved, extending the electric vehicle's driving range and reducing overall vehicle energy costs.

[0020] 3. Improve driving comfort and eliminate the pain points of traditional technologies: The innovative "expected coasting distance following" algorithm ensures that the coasting distance after the motor brakes intervene matches the driver's psychological expectations, completely eliminating the "pulling sensation" caused by the abrupt intervention of traditional energy recovery systems. Fuzzy control optimizes braking intensity during regular braking, while the motor intervention delay is dynamically adjusted during coasting braking, resulting in a smoother braking process that aligns with driving habits and significantly enhances the driving experience.

[0021] 4. Enhance system adaptability to cover driving needs across all scenarios: By employing K-means clustering and Euclidean distance recognition, the system accurately distinguishes between simple driving conditions such as highways and suburbs, and complex conditions such as urban congestion. Combined with comprehensive recognition of both normal and coasting braking intentions, it achieves full coverage of driving states across all scenarios. Simple conditions prioritize efficiency optimization, while complex conditions emphasize robust control. Differentiated strategies are matched to different braking levels, ensuring the system maintains optimal performance under various road conditions and driving styles, effectively addressing the industry pain point of poor adaptability in existing technologies.

[0022] 5. Strong technological scalability, adaptable to future development trends: The dual-layer architecture of this invention can be deeply integrated with vehicle navigation, vehicle networking, and battery thermal management systems, and can adapt to different driving styles or new operating conditions through OTA algorithm updates. It is applicable to existing dual-motor electric vehicles and can also be extended to multi-motor and intelligent connected vehicle scenarios, providing flexible support for subsequent upgrades and iterations of the vehicle energy management system, and has broad application prospects and technological continuity.

[0023] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure.

[0024] Undoubtedly, such and other objects of the present invention will become more apparent after the following detailed description of the preferred embodiments, which are illustrated in various accompanying drawings and figures.

[0025] To make the above-mentioned beneficial effects and other objects, features and advantages of the present invention more apparent and understandable, one or more preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0026] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0027] In the accompanying drawings, the same parts use the same reference numerals, and the drawings are schematic and not necessarily drawn to actual scale.

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only one or more embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on such drawings without creative effort.

[0029] Explanation of key figure labels: Figure 1 This is a schematic diagram of the safety braking area modeling for the present invention; Figure 2 This is a speed-time diagram under the gliding condition of the present invention; Figure 3 This is a schematic diagram of the neural network braking torque allocation model structure of the present invention; Figure 4 This is a schematic diagram of the overall framework of the regenerative braking system of the present invention. Detailed Implementation

[0030] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings and examples, so that the process of how the present invention uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly. It should be noted that, as long as there is no conflict, the various embodiments and features in the various embodiments of the present invention can be combined with each other, and the resulting technical solutions are all within the protection scope of the present invention.

[0031] Furthermore, numerous specific details are set forth in the following description for illustrative purposes to provide a thorough understanding of the embodiments of the invention. However, it will be apparent to those skilled in the art that the invention may be practiced without the specific details or particular methods described herein.

[0032] Please see Figures 1 - 4 This invention provides a two-layer braking energy recovery method based on operating condition and braking intention recognition, including an upper decision layer and a lower execution layer. The upper decision layer performs operating condition recognition and braking intention recognition, and the lower execution layer allocates the power of the front and rear motors within a preset safe braking area according to the decision results of the upper layer using a corresponding optimization algorithm to realize braking energy recovery.

[0033] In operation, a two-layer core architecture is constructed, consisting of an upper decision-making layer and a lower execution layer. The core functions of the upper layer are condition identification and braking intent recognition, providing a basis for subsequent control decisions. The core function of the lower layer is to allocate braking force using an optimization algorithm based on the upper-layer decisions and safety constraints. Under the constraints of a preset safe braking zone, the corresponding optimization algorithm is matched to rationally allocate the braking force between the front and rear motors, ultimately achieving efficient recovery of braking energy. This architecture breaks away from the single-logic control mode of existing technologies, realizing a shift from passive response to active adaptation. Through layered design, the system can flexibly adjust its control strategy according to different operating conditions and intentions, fundamentally solving the problem of poor adaptability in existing technologies and laying the foundation for the orderly operation of subsequent functional modules.

[0034] According to some embodiments of this application, optionally, the operating condition identification includes the following steps: S1. Extract the feature vector within the time window of real-time driving data. The feature vector includes, but is not limited to, average vehicle speed. v Maximum speed vmax Average acceleration a Maximum acceleration amaxa Average deceleration adec Maximum deceleration amaxd Parking time percentage Pstop ; S2. A large number of driving cycle segments are pre-divided into simple and complex driving conditions using K-means clustering analysis, and standard feature vectors for each type of driving condition are established. Fsimple and Fcomplex ; S3. Calculate real-time feature vectors Freal Compared with standard feature vectors Fsimple , Fcomplex The Euclidean distances are respectively Dsimple and Dcomplex The current working condition type is determined based on the distance. Dsimple =∥ Freal - Fsimple ∥, Dcomplex =∥ Freal - Fcomplex ∥.

[0035] Operating condition identification implementation: Real-time collection of vehicle driving data, extraction of average speed within a time window v =60 km / h、 Top speed vmax =75 km / h、 average acceleration a =0.8 m / s 2. Maximum acceleration amaxa =1.2 m / s 2. Average deceleration adec =−0.5 m / s 2. Maximum deceleration amaxd =−1.0 m / s 2. Parking time percentage Pstop =5%, forming real-time feature vectors Freal ; Standard feature vectors for simple working conditions are obtained in advance through K-means clustering. Fsimple and standard feature vectors for complex working conditions Fcomplex ,,calculate Dsimple =∥ Freal - Fsimple ∥=2.3, Dcomplex =∥ Freal - Fcomplex ∥=5.7, because Dsimple < Dcomplex If so, the current condition is determined to be a simple working condition.

[0036] By employing the above implementation method, seven key features, including average vehicle speed and maximum vehicle speed, are first extracted from real-time driving data to form a feature vector. Then, K-means clustering analysis is used to pre-divide simple and complex operating conditions, establishing standard feature vectors for both types. Finally, the Euclidean distance between the real-time feature vector and the two standard feature vectors is calculated, and the current operating condition type is determined based on the distance. This method enables accurate and real-time identification of operating conditions, providing a reliable basis for upper-level decision-making. By accurately distinguishing between simple and complex operating conditions, the lower-level execution layer can adopt different control algorithms, avoiding the limitations of a unified control strategy under different operating conditions, thus creating a prerequisite for improving energy recovery efficiency and system robustness.

[0037] According to some embodiments of this application, optionally, braking intent recognition includes conventional braking intent recognition and coasting braking intent recognition, wherein, Conventional braking intent recognition: based on brake pedal opening θb and brake pedal change rate The compensation coefficient is output through the fuzzy controller. k 0, the initial desired braking intensity can be obtained by looking up a table based on the pedal opening. z Compensation is applied to 0 to obtain the final desired braking strength. z = k 0 z 0, and classify braking levels according to the value of Z; Coasting braking intention recognition: based on accelerator pedal openingθa and its rate of change The urgency of coasting braking intention is identified using a fuzzy controller. Icoast urgency Icoast Determine the motor intervention delay time tA .

[0038] Braking intent recognition implementation: When the driver presses the brake pedal, the brake pedal opening is detected. θb =30%, brake pedal change rate =0.2% / ms, Fuzzy controller output compensation coefficient k 0=1.1, the initial expected braking intensity is obtained by looking up the table. z If 0 = 0.4, then the final expected braking strength is... z =1.1 0.4 = 0.44, therefore it is determined to be moderate braking; When the driver releases the accelerator pedal, the accelerator pedal opening is detected. θa =−10%, rate of change =−0.1% / ms, Fuzzy controller identifies the urgency level of coasting braking intention Icoast =0.3, determine the motor intervention delay time. tA =2 s。

[0039] By employing the above implementation methods, conventional braking intention recognition: based on the brake pedal opening and rate of change, a fuzzy controller outputs a compensation coefficient to optimize the initial desired braking intensity, resulting in a final braking intensity that matches the driver's needs. Coasting braking intention recognition: based on the accelerator pedal opening (negative values ​​indicate release) and rate of change, a fuzzy controller identifies the urgency of the intention, thereby determining the motor intervention delay time. This fills the gap in existing technologies for recognizing coasting braking intentions while accurately optimizing the judgment accuracy of conventional braking intentions. By identifying the type and urgency of braking intentions, the braking force distribution becomes more aligned with the driver's needs, avoiding energy waste during coasting and preventing insufficient or excessive braking force in conventional braking, while also providing a basis for subsequent comfort control.

[0040] According to some embodiments of this application, optionally, the braking levels include light braking, moderate braking, and emergency braking, wherein, z <0.3 indicates light braking; 0.3≤ z ≤0.8 indicates moderate braking. z >0.8 indicates emergency braking.

[0041] By adopting the above implementation methods, conventional braking is divided into three levels based on the range of the final desired braking intensity:z <0.3 indicates light braking; 0.3≤ z ≤0.8 indicates moderate braking. z A value >0.8 indicates emergency braking, with different braking force distribution strategies corresponding to different levels. Compared to conventional braking, this control is more targeted, with different braking force distribution strategies and regenerative braking ratios for different braking levels. This ensures safety during emergency braking while improving energy recovery efficiency during light and moderate braking, thereby further optimizing braking performance.

[0042] According to some embodiments of this application, optionally, the motor intervention delay time... tA The time is 0.5s-3s, during emergency coasting braking. tA =0.5s, intended to be smooth tA =3s.

[0043] By adopting the above implementation method, the motor intervention delay time is dynamically adjusted within the range of 0.5s-3s according to the urgency of the coasting braking intention. In case of an urgent intention, the motor intervenes immediately within 0.5s; in case of a mild intention, intervention is delayed by 3s to avoid conflict with potential acceleration intentions. This optimizes the intervention timing of coasting braking, improving driving comfort and safety. Through dynamic adjustment of the delay time, the motor braking intervention is made more closely aligned with the driver's operating intentions, preventing accidental motor braking intervention from affecting the driving experience, while also avoiding energy waste caused by late intervention, further balancing the comfort and economy of coasting braking.

[0044] According to some embodiments of this application, optionally, please refer to... Figure 1 The preset safe braking zone is constructed based on the ECE R13 regulation and the I curve. For any desired braking intensity Z, the intersection points of the I curve and the ECE R13 regulation line are M and N, respectively, and the line segment MN is the range of total braking force distribution between the front and rear axles under this braking intensity.

[0045] Application of safe braking zones: For the above-mentioned moderate braking z =0.44, find the intersection points M and N of the I curve and the ECE R13 regulation line, and determine the line segment MN as the braking force distribution range under this braking intensity. The combination of the front and rear motors and mechanical braking forces must fall within the range of this line segment.

[0046] By adopting the above implementation method, a closed safe braking region (HIMM1N1KNO) is constructed based on ECE R13 regulations and the I-curve (ideal front and rear axle braking force distribution curve). For any desired braking intensity, the intersection point of the I-curve and the ECE R13 regulation line is found. The line segment between the two points represents the legal distribution range of the total braking force between the front and rear axles at that braking intensity, ensuring that the braking force distribution always meets braking stability and regulatory requirements. Thus, by establishing a safe boundary for braking energy recovery, all braking force distribution strategies are executed within this region, fundamentally guaranteeing braking safety and effectively avoiding the problem of neglecting braking safety in pursuit of energy recovery efficiency, thereby ensuring the safety of the entire technical solution.

[0047] According to some embodiments of this application, optionally, when the upper-level decision is a simple working condition, the lower-level execution layer adopts a hybrid optimization strategy that combines Globalsearch and Fmincon constrained nonlinear programming local solvers, divides the braking stage based on the braking intensity, establishes a dynamic safety boundary for the regenerative braking ratio coefficient, constructs an energy recovery efficiency optimization problem, and solves the global optimal solution.

[0048] By adopting the above implementation methods, under simple operating conditions, four braking stages are divided according to braking intensity, and a dynamic safety boundary for the regenerative braking proportional coefficient is established. An energy recovery efficiency optimization problem is constructed using a linearized model of the motor torque characteristics. A hybrid optimization strategy combining the Globalsearch global solver and the Fmincon local solver is employed to search for the global optimum within the safety boundary. The control formulas for each stage are as follows: Low-intensity braking: z ≤0.3, ; Medium-intensity braking: 0.3 z ≤0.5, ; High-intensity braking: 0.5 z ≤0.8, ; Emergency braking: 0.8 z , ; Motor load torque: Where Z is the braking intensity, For regenerative braking ratio, For regenerative braking torque, For total braking torque, , , The three opposite potentials, , , It is a three-phase current. , are the moment of inertia and the mechanical angular velocity, is the frictional torque.

[0049] Simple working condition control implementation: Since the current is medium braking (0.3 < z ≤ 0.5), the regenerative braking ratio coefficient =（0.8-0.44） / 0.4=0.9， Using the Globalsearch and Fmincon hybrid optimization algorithm, an energy recovery efficiency optimization problem is constructed. Combining with the optimization model, the front axle motor braking force distribution coefficient is obtained by solving p = 0.6, then the rear axle distribution coefficient is 0.4, and the front and rear motor output torques are calculated T 1 = 180 N ⋅ m , T 2 = 120 N ⋅ m， All constraint conditions are satisfied.

[0050] This algorithm can take into account both the global search ability and the local optimization accuracy, and is suitable for scenarios with weak system nonlinearity under simple working conditions. According to the characteristics of simple working conditions, the energy recovery efficiency is maximized. Through the dynamic safety boundary and the hybrid optimization strategy, the energy recovery efficiency is optimized on the premise of ensuring safety, solving the problem of insufficient recovery efficiency of the existing technology under simple working conditions, and at the same time improving the control response speed.

[0051] According to some embodiments of the present application, optionally, when the upper-layer decision is a complex working condition, the lower-layer execution layer uses a neural network inverse model for adaptive control. The input layer of the neural network inverse model includes the vehicle speed v , the braking intensity z , the road slope α . The hidden layer of the neural network inverse model is a double hidden layer and uses the S-shaped activation function. The output layer of the neural network inverse model is the ratio of the regenerative braking torque to the total braking torque , and it is trained using the Levenberg-Marquardt backpropagation algorithm with the mean square error as the loss function.

[0052] Complex working condition control implementation: When the working condition is identified as urban congestion (complex working condition), the parameters input to the neural network inverse model are the vehicle speed v = 30 km / h， the braking intensity z = 0.5, the road slope α = 2°, and after calculation by the trained neural network model, the regenerative braking torque ratio = 0.75 is output. The front and rear motor braking forces are distributed according to this ratio to achieve adaptive robust control.

[0053] By employing the above implementation methods, a dual-hidden-layer neural network inverse model is constructed. The input layer receives three parameters: vehicle speed, braking intensity, and road gradient. These parameters are processed using a sigmoid activation function, and the output is the proportion of regenerative braking torque to total braking torque. The Levenberg-Marquardt backpropagation algorithm is used to train the model, and the model parameters are optimized using the mean squared error as the loss function to improve prediction accuracy. This approach is capable of handling the challenges of strong nonlinearity and numerous disturbances in complex operating conditions. Adaptive robust control is achieved through a data-driven neural network model, avoiding the limitations of model-based precise control under complex conditions and ensuring that the system maintains stable energy recovery efficiency and braking performance even under complex operating conditions.

[0054] According to some embodiments of this application, optionally, when the upper layer decision is a coasting braking intention, the lower execution layer controls the motor power based on the expected coasting distance, and the specific steps are as follows: A1. The vehicle speed is [missing information] when the driver releases the pedal. v 0, after delay time tA The speed of the car then decreased vA Electric motor power Fe With deceleration ae intervention; A2. Based on vehicle dynamics, considering the deceleration during motorless braking... a Deducing the braking deceleration of the electric motor through geometric relationships ae for: ;in ​ These are empirical parameters. ​ It is an empirical parameter related to driving experience and feel; that is, during unbraked coasting, the speed at which the driver actively brakes to control the distance to the target coasting point. It can be fitted through a large number of experiments. ​ = ​ - ​ ;in ​ Define the starting speed for a certain driving style. ​ Speed ​​at the expected braking point ​ The linear offset between them is a constant value fitted by a large amount of test data.

[0055] A3, according to ​ = m ​ Calculate the total power of the electric motor and distribute it to the front and rear motors; among which, m It's about vehicle quality. ​ This is used for braking and deceleration of the motor.

[0056] Please see ​ Assuming the driver is at a speed vRelease the pedal at 0, and after a delay time ​ The speed of the car then decreased ​ Electric motor power ​ With deceleration ​ Intervention. According to vehicle dynamics, the deceleration during braking without a motor is: a In order to make ​ = ​ (The area of ​​the time-velocity graph equals the distance traveled), thus the required motor braking deceleration can be derived. ​ .

[0057] Coasting brake control implementation: Driver at vehicle speed v 0=50 ​ / h Release the accelerator pedal and the delay time ​ =2 s Then, the vehicle speed decreased. ​ =45 ​ / h， Obtained through experimental fitting ​ = ​ - ​ =10 ​ / h， Then the motor braking deceleration Vehicle quality m =1800 ​ Then electric braking force ​ =1800*0.26=468 N， Distribute the power to the front and rear motors according to the optimal allocation coefficient.

[0058] By adopting the above implementation method, based on the core idea that "the actual coasting distance after the motor brakes intervene is equal to the driver's expected coasting distance," and combining vehicle dynamics and geometric relationships, the motor braking deceleration is derived; the total motor power is calculated based on the vehicle mass, and then distributed to the front and rear motors according to the optimal efficiency model. The parameters were obtained through extensive experimental fitting, closely matching actual driving scenarios. It features comfortable control during coasting, ensuring that the coasting distance after motor braking intervention matches the driver's expectations, completely eliminating the "pulling sensation" of traditional energy recovery. Simultaneously, it maximizes energy recovery during coasting, balancing comfort and economy, effectively solving the problem of balancing comfort and economy in existing technologies.

[0059] According to some embodiments of this application, optionally, the lower execution layer solves for the power distribution coefficients of the electric motor before and after through an optimization model. The optimization model aims to maximize the overall system recovery efficiency, specifically as follows: , The corresponding constraints are: By solving this model, the optimal allocation coefficients can be obtained. This maximizes energy recovery within the safety boundary; among which, For overall system efficiency, T 1. T 2 represents the braking torque of the front and rear motors, respectively. n 1. n 2 represents the speed of the front and rear motors, respectively. ​ 1. ​ 2 represents the power generation efficiency of the front and rear motors, respectively. p This is the power distribution coefficient for the front axle electric motor. ​ - ​ , ​ - ​ These represent the maximum braking force of the front and rear motors, respectively. G For vehicle weight, r For the wheel radius, ​ , ​ These are the mechanical braking forces of the front and rear axles, respectively.

[0060] The above constraints are defined as (safety boundaries): Motor capacity constraints: T 1 and T 2. The regenerative braking force of the front and rear motors must not exceed the maximum regenerative braking force of the front and rear motors at the current speed. ​ - ​ and ​ - ​ .

[0061] Allocation ratio constraints: T 1 and T 2 is also subject to p The defined "theoretical allocation ceiling" constraint, i.e. and .This is p Key aspects that directly influence decision-making space.

[0062] Total braking force constraint: Total braking force of motor ( T 1+ T 2) It must not exceed the theoretical maximum requirement determined by the vehicle deceleration (Z). ).

[0063] Regulatory safety constraints: combined with mechanical braking force ( , ) and electric motor power ( , The total braking force of each axle must fall within the safe range of braking force distribution specified in regulations such as ECE R13 to prevent wheel lock-up.

[0064] By adopting the above implementation method, an optimization model is established with the goal of maximizing the overall system recovery efficiency. Using the output torque, speed, and power generation efficiency of the front and rear motors as variables, and combining constraints such as the maximum braking force of the motors, total braking force demand, and safe operating range, the optimal power allocation coefficient for the front axle motor is solved. The power allocation coefficient for the rear axle motor is then 1 minus the power allocation coefficient for the front axle motor. p。 This allows for optimal coordinated operation of the front and rear motors under various operating conditions, maximizing system energy recovery efficiency. By dynamically distributing braking force, the power generation potential of the front and rear motors is fully utilized, avoiding inefficiency caused by a single motor operating, and further improving the overall vehicle energy economy.

[0065] It should be understood that the embodiments disclosed herein are not limited to the specific processing steps or materials disclosed herein, but should be extended to equivalent substitutions of such features as understood by those skilled in the art. It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting.

[0066] The term "embodiment" in this specification refers to a specific feature or characteristic described in connection with an embodiment that is included in at least one embodiment of the invention. Therefore, phrases or "embodiments" appearing in various places throughout the specification do not necessarily refer to the same embodiment.

[0067] Furthermore, the described features or characteristics can be incorporated into one or more embodiments in any other suitable manner. In the above description, specific details, such as thickness, quantity, etc., are provided to provide a comprehensive understanding of embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented without the aforementioned specific details or may be implemented using other methods, components, materials, etc.

Claims

1. A two-layer braking energy recovery method based on operating condition and braking intention recognition, characterized in that, It includes an upper decision-making layer and a lower execution layer. The upper decision-making layer performs working condition identification and braking intention identification. The lower execution layer, based on the decision-making results of the upper layer, uses a corresponding optimization algorithm to allocate the power of the front and rear electric motors within a preset safe braking area to achieve braking energy recovery. Operating condition identification includes the following steps: S1. Extract the feature vector within the time window of real-time driving data. The feature vector includes, but is not limited to, average vehicle speed. v Maximum speed vmax Average acceleration a Maximum acceleration amaxa Average deceleration adec Maximum deceleration amaxd Parking time percentage Pstop ; S2. A large number of driving cycle segments are pre-divided into simple and complex driving conditions using K-means clustering analysis, and standard feature vectors for each type of driving condition are established. Fsimple and Fcomplex ; S3. Calculate real-time feature vectors Freal Compared with standard feature vectors Fsimple , Fcomplex The Euclidean distances are respectively Dsimple and Dcomplex The current working condition type is determined based on the distance. Dsimple =∥ Freal − Fsimple ∥， Dcomplex =∥ Freal − Fcomplex ∥； Braking intent recognition includes conventional braking intent recognition and coasting braking intent recognition, among which, Conventional braking intent recognition: based on brake pedal opening θb and brake pedal change rate θ ˙ b The compensation coefficient is output through the fuzzy controller. k 0, the initial desired braking intensity can be obtained by looking up a table based on the pedal opening. z Compensation is applied to 0 to obtain the final desired braking strength. z = k 0∗ z 0, and based on the value of Z, identify the conventional braking intention: based on the brake pedal opening. θb and brake pedal change rate θ ˙ b The compensation coefficient is output through the fuzzy controller. k 0, the initial desired braking intensity can be obtained by looking up a table based on the pedal opening. z Compensation is applied to 0 to obtain the final desired braking strength. z = k 0∗ z 0, and classify braking levels according to the value of Z; classify braking levels; Coasting braking intention recognition: based on accelerator pedal opening θa and its rate of change θ ˙ a The urgency of coasting braking intention is identified using a fuzzy controller. Icoast urgency Icoast Determine the motor intervention delay time tA .

2. The dual-layer braking energy recovery method based on operating condition and braking intention recognition according to claim 1, characterized in that, Braking levels include light braking, moderate braking, and emergency braking. z <0.3 indicates light braking; 0.3≤ z ≤0.8 indicates moderate braking. z >0.8 indicates emergency braking.

3. The dual-layer braking energy recovery method based on operating condition and braking intention recognition according to claim 1, characterized in that, Motor intervention delay time tA The time is 0.5s-3s, during emergency coasting braking. tA =0.5s, intended to be smooth tA =3s.

4. A dual-layer braking energy recovery method based on operating condition and braking intention recognition according to claim 1, characterized in that, The preset safe braking zone is constructed based on the ECE R13 regulation and the I curve. For any desired braking intensity Z, the intersection points of the I curve and the ECE R13 regulation line are M and N, respectively, and the line segment MN is the range of total braking force distribution between the front and rear axles under this braking intensity.

5. The dual-layer braking energy recovery method based on operating condition and braking intention recognition according to claim 1, characterized in that, When the upper-level decision is a simple working condition, the lower-level execution layer adopts a hybrid optimization strategy that combines Globalsearch and Fmincon constrained nonlinear programming local solvers. Based on the braking intensity, the braking stage is divided, a dynamic safety boundary for the regenerative braking ratio coefficient is established, an energy recovery efficiency optimization problem is constructed, and the global optimal solution is solved.

6. The dual-layer braking energy recovery method based on operating condition and braking intention recognition according to claim 1, characterized in that, When the upper-level decision involves complex operating conditions, the lower-level execution layer uses a neural network inverse model for adaptive control. The input layer of the neural network inverse model includes vehicle speed. v Braking strength z Road slope α The inverse neural network model has two hidden layers and uses a sigmoid activation function. The output layer of the inverse neural network model represents the ratio of regenerative braking torque to total braking torque. The Levenberg-Marquardt backpropagation algorithm was used for training, with mean squared error as the loss function.

7. The dual-layer braking energy recovery method based on operating condition and braking intention recognition according to claim 3, characterized in that, When the upper layer decides to initiate coasting braking, the lower execution layer controls the motor's power based on the expected coasting distance. The specific steps are as follows: A1. The vehicle speed is [missing information] when the driver releases the pedal. v 0, after delay time tA The speed of the car then decreased vA Electric motor power Fe With deceleration ae intervention; A2. Based on vehicle dynamics, considering the deceleration during motorless braking... a Deducing the braking deceleration of the electric motor through geometric relationships ae for: ;in vB These are empirical parameters; A3, according to Fe = m * ae Calculate the total power of the electric motor and distribute it to the front and rear motors; among which, m It's about vehicle quality. ae This is used for braking and deceleration of the motor.

8. The dual-layer braking energy recovery method based on operating condition and braking intention recognition according to claim 7, characterized in that, The lower execution layer solves for the power allocation coefficients of the electric motors before and after the process by using an optimization model. The optimization model aims to maximize the overall system recovery efficiency, specifically as follows: ， The corresponding constraints are: By solving this model, the optimal allocation coefficients can be obtained. This maximizes energy recovery within the safety boundary; among which, For overall system efficiency, T 1. T 2 represents the braking torque of the front and rear motors, respectively. n 1. n 2 represents the speed of the front and rear motors, respectively. η 1. η 2 represents the power generation efficiency of the front and rear motors, respectively. p This is the power distribution coefficient for the front axle electric motor. Fmf - max , Fmr - max These represent the maximum braking force of the front and rear motors, respectively. G For vehicle weight, r For the wheel radius, Fbf , Fbr These are the mechanical braking forces of the front and rear axles, respectively.

Images