Regenerative braking control method and system based on contribution iteration and multi-agent model prediction

Abstract

The present invention provides a regenerative braking control method and system based on contribution iteration and multi-agent model prediction. According to the driving type of the braking system, all mechanical brakes of the front and rear axles and electric brakes of the front and rear axles are respectively regarded as different agents, and prediction models with contribution are respectively constructed according to the dynamic characteristics of the agents; each agent includes a model prediction control calculation unit, which is used to establish an optimization objective function according to vehicle speed, SOC, safety, comfort and energy recovery efficiency, while considering constraints; according to the contribution iteration method, each agent reaches a consensus to achieve collaborative optimization solution of the braking energy recovery system. The present invention puts the interaction between agents in the stage of establishing the prediction model, eliminating the need for further coordination and optimization of the solution results of all sub-control units, and improving the efficiency of collaborative control.

Description

Technical Field

[0001] The present invention belongs to the technical field of automobile wire control brake and regenerative braking, and specifically relates to a regenerative braking control method and system based on contribution iteration and multi-agent model prediction. Background Art

[0002] As one of the most important systems in a car, the braking system directly affects the driving safety of the car. The braking system of new energy vehicles also needs to have a certain braking energy recovery function. Since the existing braking energy recovery strategy is difficult to take into account both the safety and energy recovery efficiency during braking, a braking energy recovery method based on an optimization algorithm has begun to be proposed. Model predictive control has been increasingly widely used in the field of vehicle braking because of its ability to consider constraints in the optimization algorithm.

[0003] However, MPC control relies on an accurate prediction model to predict the system state within the prediction time step, so as to further solve the rolling optimization. Therefore, accurate modeling of the controlled object is crucial for MPC control. The complex vehicle reference model and its impact on vehicle driving behavior ensure the accuracy of the model predictive controller, but the high dependence on the model limits its scalability and reusability.

[0004] In order to solve the above problems, the application of multi-agent theory in the control field of chassis systems has become more and more in-depth. Intelligent agent models are constructed according to the dynamic characteristics of different subsystems to further carry out collaborative control. However, the existing multi-agent collaborative control methods still need to realize the interaction between subsystems through complex calculations, and the solution efficiency is difficult to really improve. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a regenerative braking control method and system based on contribution iteration and multi-agent model prediction, which can achieve coordinated optimization of the braking energy recovery system.

[0006] The technical solution adopted by the present invention to solve the above technical problems is: a regenerative braking control method based on contribution iteration and multi-agent model prediction, the present invention comprises:

[0007] According to the driving type of the braking system, all mechanical brakes of the front and rear axles and electric brakes of the front and rear axles are regarded as different intelligent agents, and prediction models with contribution degrees are constructed according to the dynamic characteristics of the intelligent agents;

[0008] Each agent includes a model predictive control calculation unit, which is used to establish the optimization objective function based on vehicle speed, SOC, safety, comfort and energy recovery efficiency, while considering the constraints;

[0009] Based on the iterative method of contribution, each intelligent agent can reach a consensus and realize the collaborative optimization solution of the braking energy recovery system.

[0010] According to the above scheme, the prediction model of the agent is obtained specifically in the following way:

[0011] Establish the longitudinal dynamics model of the whole vehicle;

[0012] According to the dynamic characteristics of the intelligent agents, their respective components in the longitudinal dynamic model of the whole vehicle are calculated as the prediction model of each intelligent agent; at the same time, the contribution of all other intelligent agents in the prediction model of each intelligent agent is determined.

[0013] According to the above scheme, the longitudinal dynamic model of the whole vehicle is:

[0014]

[0015] F resist =F f +F w +F i

[0016]

[0017] F i = mgi

[0018] Where m is the vehicle mass, is the longitudinal deceleration, F x1 is the front axle braking force, F x2 is the rear axle braking force, F resist is the driving resistance, F f Represents rolling resistance, F w Represents air resistance; F i represents slope resistance; f1 and f2 represent the rolling resistance coefficients of the front and rear wheels; is the vehicle speed; C D Represents the air resistance coefficient; A * It represents the frontal area of ​​the car during driving; i represents the slope; g is the acceleration due to gravity.

[0019] According to the above scheme, the braking system is a braking system of a front-axle driven pure electric vehicle, including three intelligent agents: front axle electric brake, front axle mechanical brake and rear axle mechanical brake. The prediction model of the intelligent agent is derived from the components in the longitudinal dynamics model of the whole vehicle, which are:

[0020] 1) Front axle electric brake prediction model:

[0021]

[0022] In the formula, is the longitudinal deceleration of the vehicle; X is the vehicle speed; U F1motor =F x1Motor , is the electric braking force of the front axle; C D represents the air resistance coefficient; W F1motor is the front axle electric brake interference term, are the contributions of the front axle mechanical brake and rear axle mechanical brake agents, respectively; a, b, g, i, and f1 are the distance from the front axle to the center of mass, the distance from the rear axle to the center of mass, the gravitational acceleration, the slope, and the front wheel rolling resistance coefficient, respectively; m is the vehicle mass;

[0023] 2) Front axle mechanical brake prediction model:

[0024]

[0025] Where: is the vehicle longitudinal deceleration; U F1mec =F x1mec , is the front axle mechanical braking force; W F1mec is the front axle mechanical brake interference term, is the contribution of the front axle electric brake and rear axle mechanical brake agents,

[0026]

[0027] 3) Rear axle mechanical brake prediction model:

[0028]

[0029] Where: is the vehicle longitudinal deceleration; U F1mec =F x2mec , is the mechanical braking force of the rear axle; W F2mec is the rear axle mechanical brake interference term, The contribution of the front axle electric brake and front axle mechanical brake agents;

[0030] According to the above scheme, the constraints are: the maximum and minimum output constraints of the intelligent agent, and the ground friction ellipse constraint.

[0031] According to the above scheme, the optimization objective function and constraint conditions are specifically as follows:

[0032]

[0033] Q=c·J+(1-c)·W

[0034]

[0035] Where, X (q) The state quantity at the qth iteration of contribution, i.e., deceleration, X des is the expected deceleration, U (q) It represents the control input of the contribution at the i-th iteration, i is the number of iterations, N p is the prediction interval, R X , R U The parameter matrix represents the state and input, J and M are safety and energy recovery efficiency indicators respectively, Q is the overall objective function, c represents the weight of optimal braking safety and optimal energy recovery efficiency, and the input quantity U (q) To meet the actuator maximum With minimum Condition, F x 、F y 、F z represents the vertical force of the ground, μ is the ground adhesion coefficient, and the braking force satisfies the ground friction ellipse condition.

[0036] According to the above scheme, each intelligent agent solves its own optimal control sequence according to the optimization objective function and constraints, and performs collaborative control solution according to the contribution iteration method.

[0037] According to the above scheme, the contribution iteration method specifically includes:

[0038] Each agent updates its contribution by communicating with all other agents, forming a control network together;

[0039] The prediction model of each agent is solved multiple times in an iterative process to reach a consensus among all agents on their contribution to the overall control effort.

[0040] According to the above scheme, in the multiple iterative solution process, in each iteration, each intelligent agent solves the optimal control sequence according to the optimization objective function and constraints, and reports its contribution sequence; in the next iteration, all intelligent agents will share the contribution of the previous iteration to update their respective prediction models until the contribution sequence no longer changes, that is, a consensus is reached.

[0041] A system for implementing the regenerative braking control method based on contribution iteration and multi-agent model prediction, according to the driving type of the braking system, all mechanical brakes of the front and rear axles and the electric brakes of the front and rear axles are regarded as different agents respectively, each agent includes a prediction model with the contribution of the other two agents, and a model predictive control calculation unit; each agent updates its own contribution through communication with all other agents, and together they form a control network; wherein,

[0042] The model predictive control calculation unit of each agent is used to establish the optimization objective function based on vehicle speed, SOC, safety, comfort and energy recovery efficiency, while considering the constraints;

[0043] The prediction model of each intelligent agent is used to perform multiple iterations based on the contribution iteration method, and finally a consensus is reached among each intelligent agent.

[0044] The beneficial effects of the present invention are as follows: a dynamic model is established for different actuators and an MPC control unit is configured, each different MPC control unit is regarded as an intelligent agent, and a consensus is reached between multiple intelligent agents based on the contribution iteration method, thereby realizing the collaborative optimization solution of the braking energy recovery system. This method can greatly reduce the difficulty of dynamic modeling of the system, eliminate the impact of redundant actuators on the amount of solution calculation, improve the real-time performance of the algorithm, facilitate the update and addition of actuators, and have stronger reusability and scalability. At the same time, compared with other collaborative optimization algorithms, the method based on contribution iteration puts the interaction between intelligent agents in the stage of establishing the prediction model, eliminating the need for further coordination and optimization of the solution results of all sub-control units, and improving the efficiency of collaborative control. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of an embodiment of the present invention.

[0046] Figure 2 The figure is a flow chart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The present invention will be further described below in conjunction with specific examples and drawings.

[0048] like Figure 1As shown, the present invention provides a regenerative braking control method based on contribution iteration and multi-agent model prediction, which is applicable to various braking systems. According to the driving type of the braking system, all mechanical brakes of the front and rear axles and electric brakes of the front and rear axles are regarded as different agents respectively. If the braking system is a braking system of a pure electric vehicle driven by the front axle, the front axle electric brake, the front axle mechanical brake and the rear axle mechanical brake are regarded as three different agents. According to the dynamic characteristics of other driving types of electric vehicles, if it is a braking system of a pure electric vehicle driven by the rear axle, the agents are the front and rear axle mechanical brakes and the rear axle electric brake, a total of three; if it is a braking system of a pure electric vehicle driven by the front and rear axles, the agents are the front and rear axle mechanical brakes and the front and rear axle electric brakes, a total of four; if it is a braking system of a pure electric vehicle driven by a distributed drive, the agents are the front and rear axle mechanical brakes and the four-wheel electric brakes, a total of six. This method adopts a hierarchical structure, the upper layer is the driver input layer, the middle layer is the multi-agent collaboration layer, and the lower layer is the actuator response layer. Taking the braking system of a pure electric vehicle driven by the front axle as an example, as Figure 2 As shown, the specific steps include:

[0049] S1. The front axle electric brake, the front axle mechanical brake and the rear axle mechanical brake are regarded as three different intelligent agents, and prediction models with contribution degrees are respectively constructed according to the dynamic characteristics of the intelligent agents.

[0050] First, a prediction model of the vehicle center of mass should be established based on the longitudinal dynamics equation of vehicle braking. Then, according to the dynamic characteristics of the front and rear axle mechanical brakes and the front axle electric brake agent, its component in the longitudinal dynamics model of the vehicle center of mass is calculated, and its contribution model is determined. Figure 1 Various interferences in .

[0051] The prediction model of the agent is specifically obtained in the following way:

[0052] 1.1. Establish the longitudinal dynamics model of the whole vehicle:

[0053]

[0054] F resist =F f +F w +F i

[0055]

[0056] F i = mgi

[0057] Where m is the vehicle mass, is the longitudinal deceleration, F x1 is the front axle braking force, F x2 is the rear axle braking force, Fresist is the driving resistance, F f Represents rolling resistance, F w Represents air resistance; F i represents slope resistance; f1 and f2 represent the rolling resistance coefficients of the front and rear wheels; is the vehicle speed; C D Represents the air resistance coefficient; A * It represents the frontal area of ​​the car during driving; i represents the slope; g is the acceleration due to gravity.

[0058] 1.2. According to the dynamic characteristics of the three intelligent agents, their respective components in the longitudinal dynamic model of the whole vehicle are calculated as the prediction model of each intelligent agent; at the same time, the contribution of the other two intelligent agents in the prediction model of each intelligent agent is determined.

[0059] The prediction models of the three agents are derived from the components in the longitudinal dynamics model of the vehicle, and are:

[0060] 1) Front axle electric brake prediction model:

[0061]

[0062] In the formula, is the longitudinal deceleration of the vehicle; X is the vehicle speed; U F1motof =F x1Motor , is the electric braking force of the front axle; C D Represents the air resistance coefficient; W F1motof is the front axle electric brake interference term, are the contributions of the front axle mechanical brake and rear axle mechanical brake agents, respectively; a, b, g, i, and f1 are the distance from the front axle to the center of mass, the distance from the rear axle to the center of mass, the gravitational acceleration, the slope, and the front wheel rolling resistance coefficient, respectively; m is the vehicle mass;

[0063] 2) Front axle mechanical brake prediction model:

[0064]

[0065] Where: is the vehicle longitudinal deceleration; U F1mec =F x1mec , is the front axle mechanical braking force; W F1mec is the front axle mechanical brake interference term, is the contribution of the front axle electric brake and rear axle mechanical brake agents,

[0066] 3) Rear axle mechanical brake prediction model:

[0067]

[0068] Where: is the vehicle longitudinal deceleration; U F1mec =F x2mec , is the mechanical braking force of the rear axle; W F2mec is the rear axle mechanical brake interference term, The contribution of the front axle electric brake and front axle mechanical brake agents;

[0069] S2. Each intelligent agent includes a model predictive control calculation unit, which is used to establish an optimization objective function based on vehicle speed, SOC, safety, comfort and energy recovery efficiency, while considering constraints. The constraints are: maximum and minimum output constraints of the three intelligent agents, and point friction ellipse constraints.

[0070] Specifically, according to the ideal braking deceleration X des With the predicted state sequence The difference As well as the minimum control input, the efficiency of braking energy recovery is taken into account to construct the objective function of rolling optimization, the actuator constraints and the ground friction ellipse constraints are considered and the optimization problem is transformed into a standard quadratic programming problem for solution.

[0071] In this embodiment, the objective function and constraint conditions are specifically as follows:

[0072]

[0073]

[0074] Q=c·J+(1-c)·W

[0075]

[0076] Where, X (q) The state quantity at the qth iteration of contribution, i.e., deceleration, X des is the expected deceleration, U (q) It represents the control input of the contribution at the i-th iteration, i is the number of iterations, N p is the prediction interval, R X , R U The parameter matrix represents the state and input, J and M are safety and energy recovery efficiency indicators respectively, Q is the overall objective function, c represents the weight of optimal braking safety and optimal energy recovery efficiency, and the input quantity U(q) To meet the actuator maximum With minimum Condition, F x 、F y 、F z It represents the vertical force of the ground, μ is the ground adhesion coefficient, and the braking force should meet the ground friction ellipse condition.

[0077] S3. Based on the iterative method of contribution, each intelligent agent can reach a consensus and realize the collaborative optimization solution of the braking energy recovery system.

[0078] Each intelligent agent solves its own optimal control sequence according to the optimization objective function and constraints, and performs collaborative control solution according to the contribution iteration method.

[0079] The contribution iteration method specifically includes: each intelligent agent updates its own contribution through communication with the other two intelligent agents to jointly form a control network; the prediction model of each intelligent agent is iteratively solved multiple times to reach a consensus among all intelligent agents on their contribution to the overall control work.

[0080] In the multiple iterative solution process, in each iteration, each intelligent agent solves the optimal control sequence according to the optimization objective function and constraints, and reports its contribution sequence; in the next iteration, all intelligent agents will share the contribution of the previous iteration to update their respective prediction models until the contribution sequence no longer changes, that is, a consensus is reached.

[0081] For example, in some embodiments, in the qth iteration, each node k solves its optimal control sequence according to the optimization objective function and the constraint conditions. (Right now ), representing the optimal control sequence of front axle motor braking, front and rear axle mechanical braking, and reporting their contribution sequence (Right now ), respectively representing the contribution sequence of the front axle motor brake, the front and rear axle mechanical brakes). In the next iteration, all agents will share To update their respective interference terms W k (i.e. W F1motor , W F1mec , W F2mec ) until the contribution sequence no longer changes, and a consensus is reached.

[0082] Finally, the mechanical braking forces of the front and rear axles and the electric braking force of the front axle solved after the iteration are input into the controllers of each actuator.

[0083] If it is a braking system of other driving types, a corresponding prediction model and optimization objective function can be formulated according to the ideas and principles of the above embodiments.

[0084] The method of the present invention does not require further optimization, and the model prediction solution of each intelligent agent is the final solution, which improves the collaborative control efficiency of multiple subsystems.

[0085] The present invention also provides a system for implementing the regenerative braking control method based on contribution iteration and multi-agent model prediction, comprising three agents, namely, front axle electric brake, front axle mechanical brake and rear axle mechanical brake, each agent comprising a prediction model with the contribution of the other two agents, and a model prediction control calculation unit; each agent updates its own contribution through communication with the remaining two agents, and together constitutes a control network; wherein the model prediction control calculation unit of each agent is used to establish an optimization objective function based on vehicle speed, SOC, safety, comfort and energy recovery efficiency, while considering constraints; the prediction model of each agent is used to perform multiple iterative solutions based on contribution iteration, so that each agent can finally reach a consensus.

[0086] The method described in the present invention takes into account both safety and energy recovery efficiency: the present invention ensures the optimal balance between vehicle stability and economy by reasonably allocating regenerative braking torque and friction mechanical braking torque, so as to solve the problem that the transmission braking energy recovery control strategy is difficult to take into account both braking safety and energy recovery efficiency.

[0087] The present invention improves reusability and scalability: The present invention establishes dynamic models for different actuators or appropriately decomposes the dynamics of the entire vehicle into actuator models and matches them with MPC control algorithms, regards each different MPC control unit as an intelligent agent, and then performs collaborative control on multiple intelligent agents. This method can greatly reduce the difficulty of dynamic modeling of the system, eliminate the impact of redundant actuators on the amount of solution calculations, and improve the real-time performance of the algorithm. At the same time, this method also facilitates the update and addition of actuators, and has stronger reusability and scalability.

[0088] The present invention improves the efficiency of collaborative control of multiple subsystems: Currently, the optimal strategy based on a multi-agent architecture to achieve collaborative control of multiple subsystems has begun to be widely studied. Most studies focus on minimizing the convex objective function. Through information interaction within a communication cycle, the Pareto optimal strategy can be calculated. However, the Pareto optimal solution is not the only solution, but the Pareto frontier composed of the optimal solution set. To obtain the final solution output to the subsystem, the Pareto frontier still needs to be further optimized and solved, thereby reducing the real-time performance of the system calculation. The model prediction solution of each subsystem of the present invention is the final solution, thereby improving the efficiency of collaborative control of multiple subsystems.

[0089] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all these improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A regenerative braking control method based on contribution iteration and multi-agent model prediction, characterized in that: include: According to the driving type of the braking system, all mechanical brakes of the front and rear axles and electric brakes of the front and rear axles are regarded as different intelligent agents, and prediction models with contribution degrees are constructed according to the dynamic characteristics of the intelligent agents. Each agent includes a model predictive control calculation unit, which is used to establish the optimization objective function based on vehicle speed, SOC, safety, comfort and energy recovery efficiency, while considering the constraints; Based on the iterative method of contribution, each intelligent agent can reach a consensus and realize the collaborative optimization solution of the braking energy recovery system.

2. The method according to claim 1, characterized in that: The prediction model of the agent is specifically obtained in the following way: Establish the longitudinal dynamics model of the whole vehicle; According to the dynamic characteristics of the intelligent agents, their respective components in the longitudinal dynamic model of the whole vehicle are calculated as the prediction model of each intelligent agent; at the same time, the contribution of all other intelligent agents in the prediction model of each intelligent agent is determined.

3. The method according to claim 2, characterized in that: The longitudinal dynamic model of the vehicle during braking is: F resist =F f +F w +F i F i =go Where m is the vehicle mass, is the longitudinal deceleration, F x1 is the front axle braking force, F x2 is the rear axle braking force, F resist is the driving resistance, F f Represents rolling resistance, F w Represents air resistance; F i represents slope resistance; f1 and f2 represent the rolling resistance coefficients of the front and rear wheels; is the vehicle speed; C D Represents the air resistance coefficient; A * It represents the frontal area of ​​the car during driving; i represents the slope; g is the acceleration due to gravity.

4. The method according to claim 2, characterized in that: The braking system is a braking system of a front-axle driven pure electric vehicle, including three intelligent agents: front axle electric brake, front axle mechanical brake and rear axle mechanical brake. The prediction model of the intelligent agent is derived from the components in the longitudinal dynamics model of the whole vehicle, which are: 1) Front axle electric brake prediction model: In the formula, is the longitudinal deceleration of the vehicle; X is the vehicle speed; U F1motor =F x1Motor , is the electric braking force of the front axle; C D represents the air resistance coefficient; W F1motor is the front axle electric brake interference term, are the contributions of the front axle mechanical brake and rear axle mechanical brake agents, respectively; a, b, g, i, and f1 are the distance from the front axle to the center of mass, the distance from the rear axle to the center of mass, the gravitational acceleration, the slope, and the front wheel rolling resistance coefficient, respectively; m is the vehicle mass; 2) Front axle mechanical brake prediction model: Where: is the vehicle longitudinal deceleration; U F1mec =F x1mec , is the front axle mechanical braking force; W F1mec is the front axle mechanical brake interference term, is the contribution of the front axle electric brake and rear axle mechanical brake agents, 3) Rear axle mechanical brake prediction model: Where: is the vehicle longitudinal deceleration; U F1mec =F x2mec , is the mechanical braking force of the rear axle; W F2mec is the rear axle mechanical brake interference term, The contribution of the front axle electric brake and front axle mechanical brake agents; 5. The method according to claim 1, characterized in that: The constraints are: maximum and minimum output constraints of the intelligent agent, and ground friction ellipse constraints.

6. The method according to claim 4, characterized in that: The optimization objective function and constraint conditions are specifically as follows: Q=c·J+(1-c)·W In the formula, X (q) The state quantity at the qth iteration of contribution, i.e., deceleration, X des is the expected deceleration, U (q) It represents the control input of the contribution at the i-th iteration, i is the number of iterations, N p is the prediction interval, R X , R U The parameter matrix represents the state and input, J and M are safety and energy recovery efficiency indicators respectively, Q is the overall objective function, c represents the weight of optimal braking safety and optimal energy recovery efficiency, and the input quantity U (q) To meet the actuator maximum With minimum Condition, F x 、F y 、F z It represents the vertical force of the ground, μ is the ground adhesion coefficient, and the braking force should meet the ground friction ellipse condition.

7. The method according to claim 1, characterized in that: Each intelligent agent solves its own optimal control sequence according to the optimization objective function and constraints, and performs collaborative control solution according to the contribution iteration method.

8. The method according to claim 7, characterized in that: The contribution iteration method specifically includes: Each agent updates its contribution by communicating with all other agents, forming a control network together; The prediction model of each agent is solved multiple times in an iterative process to reach a consensus among all agents on their contribution to the overall control effort.

9. The method according to claim 8, characterized in that: In the multiple iterative solution process, in each iteration, each intelligent agent solves the optimal control sequence according to the optimization objective function and constraints, and reports its contribution sequence; in the next iteration, all intelligent agents will share the contribution of the previous iteration to update their respective prediction models until the contribution sequence no longer changes, that is, a consensus is reached.

10. A system for implementing the regenerative braking control method based on contribution iteration and multi-agent model prediction as described in any one of claims 1 to 9, characterized in that: According to the driving type of the braking system, all mechanical brakes of the front and rear axles and electric brakes of the front and rear axles are regarded as different agents. Each agent includes a prediction model with the contribution of the other two agents and a model predictive control calculation unit. Each agent updates its own contribution through communication with all other agents, and together they form a control network. The model predictive control calculation unit of each agent is used to establish the optimization objective function based on vehicle speed, SOC, safety, comfort and energy recovery efficiency, while considering the constraints; The prediction model of each intelligent agent is used to perform multiple iterations based on the contribution iteration method, and finally a consensus is reached among each intelligent agent.

Images