Energy-saving driving based cloud-supported energy management control method and system for connected bus
By constructing a multi-objective vehicle speed planning optimization model and battery SOC trajectory optimization, combined with adaptive control using reinforcement learning, the energy consumption and stability issues of new energy buses in complex urban traffic environments were solved, achieving global energy saving and robust tracking.
Patent Information
- Application Number
- CN202610348793.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-20
- Publication Date
- 2026-07-03
Smart Images
Figure CN122334565A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent new energy bus technology, and in particular to a cloud-supported energy management and control method and system for connected buses based on energy-saving driving. Background Technology
[0002] Urban traffic environments exhibit significant spatiotemporal coupling and operational uncertainties. The driving status of fixed-route new energy buses is simultaneously affected by multiple factors, including road traffic flow, signal timing, intersection queuing, bus stop stops along the route, and fluctuations in surrounding traffic flow. Frequent vehicle starts and stops, intersection waiting conditions, and mismatches between stop stop rhythms and signal control can easily lead to vehicles entering uneconomical operating states, significantly increasing overall vehicle energy consumption and burdening the power system.
[0003] Existing technologies reduce the number of stops at intersections and improve traffic efficiency by jointly optimizing signal timing parameters and the speed of buses passing through intersections. This method improves bus traffic efficiency to some extent within local road sections or adjacent intersections and is suitable for application scenarios centered on signal coordination and bus priority control. However, this method primarily focuses on local optimization at the signal control and speed guidance layers, failing to comprehensively consider the cumulative impact of fixed-route bus stop behavior on vehicle energy consumption and battery state of charge (SOC) evolution. Furthermore, it does not integrate global speed planning, SOC trajectory planning, and vehicle-side execution control into a unified collaborative optimization framework.
[0004] In addition, existing technologies typically rely on preset control strategies or rule-driven approaches, lacking the ability to adaptively adjust to traffic flow fluctuations, vehicle load changes, and model uncertainties during actual operation. Therefore, in the scenario of energy-saving operation of new energy buses throughout the entire journey in complex urban traffic environments, there are still problems of insufficient overall coordination and limited execution robustness. Summary of the Invention
[0005] To address the aforementioned issues, this invention proposes a cloud-supported energy management and control method and system for connected buses based on energy-saving driving. This invention can improve the overall energy efficiency, operational stability, and control robustness of new energy buses in complex urban traffic environments.
[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a cloud-supported energy management and control method for connected buses based on energy-saving driving, comprising: The system acquires traffic flow data, signal timing, intersection queue length, queue dissipation time, and predicted number of vehicles. Combined with the predicted stop time window of bus stops, it constructs a multi-objective vehicle speed planning optimization model that satisfies green light time window constraints, following safety constraints, and vehicle dynamics constraints. The system solves the multi-objective vehicle speed planning optimization model to obtain the globally optimal smooth vehicle speed trajectory. Based on the globally optimal smooth vehicle speed trajectory and its corresponding acceleration trajectory, the required power of the whole vehicle is calculated and converted into the actual output power of the battery. The SOC state transition equation is obtained through the battery equivalent circuit model. The upper and lower limits of SOC are used as design factors and vehicle load is used as noise factors. The signal-to-noise ratio is calculated and the feasible region of the optimal SOC trajectory is found. The optimal SOC reference trajectory is generated based on the center line of the feasible region of the optimal SOC trajectory. The system acquires real-time vehicle operating status and surrounding traffic information, and calculates the vehicle speed error and SOC error corresponding to the globally optimal smooth vehicle speed trajectory and the optimal SOC reference trajectory. The vehicle speed error, SOC error, real-time vehicle operating status, and surrounding traffic information are integrated into the state space of reinforcement learning, and the vehicle demand acceleration and motor output torque are used as the action space to construct an immediate reward function. The system maps from the state space to the dynamic space by training a policy network, and updates the network parameters by maximizing future cumulative rewards to achieve adaptive tracking and execution of the globally optimal smooth vehicle speed trajectory and the optimal SOC reference trajectory.
[0007] A further technical solution is that the objective function of the multi-objective vehicle speed planning and optimization model is: ;in, For discrete sampling times, The total number of planned steps, For the site serial number, The total number of stations, , , , , , All are weighting coefficients. For instantaneous energy consumption, For time deviation, For the sequence of queue dissipation at intersections, For speed smoothing term, For ride comfort, This refers to the parking prediction item for the station.
[0008] A further technical solution involves constraining the green light time window by limiting the time a vehicle arrives at the intersection, ensuring it arrives and passes through the intersection within the effective green light time window. The effective green light time window is represented as follows: ,in, When the green light turns on. The end of the green light. This refers to the queue dissipation time in the predicted results; The following safety constraint is expressed as follows: ;in, The relative distance between the vehicles. For a safe following distance, The relative speed of the vehicles; The vehicle dynamics constraints are: ;in, for Vehicle speed at all times , These are the upper and lower speed limits, respectively. for Acceleration at all times , These are the upper and lower limits of acceleration, for Accelerate constantly , These are the upper and lower limits of jerk, respectively.
[0009] A further technical solution is to use a Bezier curve trajectory optimization method based on spatiotemporal corridor constraints to solve the multi-objective vehicle speed planning optimization model.
[0010] A further technical solution yields the following formula for calculating the required power of the entire vehicle: in, for The power required by the vehicle at any given time. For powertrain transmission efficiency. , , , These represent the vehicle mass, gravitational acceleration, rolling resistance coefficient, and road slope angle, respectively. , , These represent the air drag coefficient, frontal area, and air density, respectively. This is the rotational mass conversion factor. From the optimal displacement trajectory The planning acceleration obtained by taking the second derivative with respect to time. To assist in system power consumption, The globally optimal smooth vehicle speed trajectory. This is the acceleration trajectory.
[0011] A further technical solution is that the SOC state transition equation is expressed as: ;in, The time rate of change of SOC For battery output power, Open circuit voltage, This refers to the battery's internal resistance. This refers to the battery's rated capacity.
[0012] A further technical solution is that the instantaneous reward function includes a tracking error term, an energy consumption penalty term, a safety penalty term, a comfort penalty term, and a strategy penalty term, specifically expressed as follows: in, , , , , , These are the weighting coefficients for each item. For SOC deviation, For speed deviation, This refers to the instantaneous output power of the battery. For following vehicle safety penalty function, For acceleration command penalty items, Impose additional penalties for parking at intersections.
[0013] Secondly, this invention provides a cloud-supported energy management and control system for connected buses based on energy-saving driving, comprising the following modules: The vehicle speed planning module is configured to: acquire traffic flow data, signal timing, intersection queue length, queue dissipation time and predicted number of vehicles that can pass, combine the predicted stop time window of bus stops, construct a multi-objective vehicle speed planning optimization model that satisfies green light time window constraints, following safety constraints and vehicle dynamics constraints, and solve the multi-objective vehicle speed planning optimization model to obtain the globally optimal smooth vehicle speed trajectory. The SOC optimization module is configured to: calculate the required power of the vehicle based on the globally optimal smooth vehicle speed trajectory and its corresponding acceleration trajectory, convert the required power of the vehicle into the actual output power of the battery, obtain the SOC state transition equation through the battery equivalent circuit model; use the upper and lower limits of SOC as design factors and the vehicle load as noise factors to calculate the signal-to-noise ratio and find the feasible region of the optimal SOC trajectory; and generate the optimal SOC reference trajectory based on the centerline of the feasible region of the optimal SOC trajectory. The tracking control module is configured to: acquire the real-time operating status of the vehicle and surrounding traffic information, and calculate the vehicle speed error and SOC error corresponding to the globally optimal smooth vehicle speed trajectory and the optimal SOC reference trajectory; integrate the vehicle speed error, SOC error, real-time vehicle operating status and surrounding traffic information as the state space of reinforcement learning, and construct an immediate reward function using the vehicle demand acceleration and motor output torque as the action space; realize the mapping from the state space to the dynamic space by training the policy network, and update the network parameters by maximizing the future cumulative reward to achieve adaptive tracking execution of the globally optimal smooth vehicle speed trajectory and the optimal SOC reference trajectory.
[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention acquires traffic flow data, signal timing, and intersection queuing prediction results, and combines them with predicted bus stop stop time windows to construct a multi-objective vehicle speed planning optimization model that satisfies multiple constraints such as green light time windows, following safety, and vehicle dynamics. The globally optimal smooth vehicle speed trajectory obtained by solving the model fully integrates the core traffic characteristics of intersections and stops, taking into account multiple objectives such as green light passage, stop operation rhythm, energy consumption optimization, driving safety, and passenger comfort. This makes the vehicle speed planning results more adaptable to the complex urban traffic environment, reduces the generation of uneconomical operating conditions from the source of driving, and effectively reduces vehicle driving energy consumption.
[0015] This invention accurately calculates the power demand of the entire vehicle throughout its journey based on the globally optimal smooth vehicle speed trajectory and its acceleration trajectory, and further converts it into the actual output power at the battery end, deeply coupling SOC planning with vehicle speed driving conditions; it obtains the SOC state transition equation through the battery equivalent circuit model, and also introduces the Taguchi Robust Design (TRD) method, using the upper and lower limits of SOC as design factors and vehicle load as noise factors to optimize the signal-to-noise ratio and find the feasible region of the optimal SOC trajectory, so that the planned SOC trajectory has strong anti-interference ability against vehicle load fluctuations, effectively solving the problem of energy efficiency instability caused by the switching between full load and no load in actual operation.
[0016] This invention acquires real-time vehicle operating status and surrounding traffic information, calculates vehicle speed error and SOC error, and integrates the above errors, real-time vehicle status, and environmental information as a reinforcement learning state space to construct an immediate reward function. By training the policy network and maximizing future cumulative rewards to update network parameters, it achieves dynamic mapping from state space to action space, enabling vehicle-side control to have self-learning and adaptive adjustment capabilities. It can effectively compensate for deviations caused by traffic flow fluctuations and dynamic model drift, and achieve high-precision and robust tracking of the globally optimal vehicle speed trajectory and SOC reference trajectory, ensuring the effective implementation of the global energy-saving strategy. Attached Figure Description
[0017] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.
[0018] Figure 1 This is a flowchart illustrating the cloud-supported energy management and control method for connected buses based on energy-saving driving, as described in this invention. Figure 2 This is a scenario of intersection-station road segments in the global route planning of this invention; Figure 3 This is a schematic diagram of the green wave band-distance timing of the upstream and downstream intersection signals for buses and other vehicles according to the present invention; Figure 4 This is a schematic diagram of the station arrival time penalty function in the vehicle speed planning objective function of this invention; Figure 5 This is a schematic diagram illustrating the collaborative effect of the vehicle speed trajectory and SOC trajectory throughout the entire process according to the present invention. Detailed implementation method: The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0019] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0020] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.
[0021] Example 1 like Figure 1 As shown, this embodiment provides a cloud-supported energy management and control method for connected buses based on energy-saving driving. This includes a cloud-based global vehicle speed planning process oriented towards intersection queuing and station operation characteristics; a cloud-based collaborative energy management and SOC reference trajectory generation process considering the impact of station stops; and a vehicle-side adaptive tracking and execution control process under cloud-based global planning constraints. The specific steps are as follows: S1: Obtain traffic flow data, signal timing, intersection queue length, queue dissipation time, and predicted number of vehicles that can pass. Combine this with the predicted stop time window of bus stops to construct a multi-objective vehicle speed planning optimization model that satisfies green light time window constraints, following safety constraints, and vehicle dynamics constraints. Solve the multi-objective vehicle speed planning optimization model to obtain the globally optimal smooth vehicle speed trajectory.
[0022] In step S1, traffic flow data, signal timing, intersection queue length, queue dissipation time, and predicted number of vehicles are obtained based on vehicle-road-cloud collaborative perception.
[0023] Meanwhile, based on historical operational data and real-time passenger flow information, the cloud platform utilizes long-term historical arrival and departure timestamp data for this route to predict the baseline stopping time of vehicles at target stations through a time-series analysis model, ultimately obtaining the predicted stopping time window for each bus stop. ,in, To predict the arrival time of vehicles at the station, To predict when a vehicle will leave the station.
[0024] Based on the aforementioned multi-source information, a multi-objective vehicle speed planning and optimization model is constructed that satisfies green light time window constraints, following safety constraints, and vehicle dynamics constraints, such as... Figure 2 As shown, the objective function of the multi-objective vehicle speed planning optimization model is: in, For discrete sampling times, The total number of planned steps, For the site serial number, The total number of stations, , , , , , All are weighting coefficients. to satisfy , For instantaneous energy consumption, For time deviation, For the sequence of queue dissipation at intersections, For speed smoothing term, For ride comfort, This refers to the parking prediction item for the station.
[0025] In this embodiment, instantaneous energy consumption item :
[0026] in, This is the maximum power of the vehicle's drive system. For the vehicle in the The driving power at time t is calculated using the following formula:
[0027] in, The total mass of the vehicle. It is the acceleration due to gravity. The rolling resistance coefficient, The road slope angle, air drag coefficient , which is the vehicle's frontal area. air density, This is the rotational mass conversion factor. This represents the total transmission efficiency.
[0028] Time Deviation Item : ; in, To accumulate travel time, For reference only, the travel time is provided by Calculations show that For departure time, The total running time specified in the running chart.
[0029] Queue dissipation sequence at intersections :
[0030] in, To plan the predicted time for the vehicle to travel to its current position. This refers to the green light activation time for the current signal cycle at the downstream intersection. This refers to the time required for the queue of vehicles to completely dissipate, predicted based on the queue length at the intersection. This is the penalty constant for vehicles arriving at the intersection during a red light.
[0031] Speed smoothing term :
[0032] in, and The first Time and the Vehicle speed at all times This is the maximum speed limit for vehicles.
[0033] Ride comfort :
[0034] in, For the first Vehicle acceleration at all times For the maximum permissible acceleration, For the first The vehicle should accelerate at all times. For the maximum permissible jerk, This is the jerk weighting coefficient.
[0035] Station parking prediction ,like Figure 4As shown:
[0036] in, It is an optimization problem involving the arrival of vehicles at stations. The actual moment, and Planning for vehicle entry and exit The time at each bus stop This refers to the expected length of time the station will be occupied.
[0037] In this embodiment, as Figure 3 As shown, the green light time window constraint is based on signal timing information and the predicted queue dissipation time. It constrains the arrival time of vehicles at the intersection, ensuring they arrive and pass through the intersection within the effective green light time window. The effective green light time window is represented as follows: ,in, When the green light turns on. The end of the green light. This represents the queue dissipation time in the predicted results.
[0038] Following vehicle safety constraints: Predicting surrounding vehicle trajectories using model output Ensure the planned speed and maintain a safe distance from vehicles in front and behind. , is represented as: ;in, The relative distance between the vehicles. For a safe following distance, This represents the relative speed of the vehicles.
[0039] Vehicle dynamics constraints: Vehicle speed, acceleration, and jerk must be within allowable ranges, expressed as: ;in, for Vehicle speed at all times , These are the upper and lower speed limits, respectively. for Acceleration at all times , These are the upper and lower limits of acceleration, for Accelerate constantly , These are the upper and lower limits of jerk, respectively.
[0040] In step S1, the multi-objective vehicle speed planning optimization model is solved using the Bézier curve trajectory optimization method based on spatiotemporal corridor constraints, specifically as follows: First, a continuous drivable spatiotemporal region is constructed through spatial topology analysis. Then, based on signal timing and intersection queue dissipation time obtained from the cloud... The expected length of time for stopping at stations ,exist A series of connected trapezoidal prism spacetime corridors are constructed in the (displacement-time) diagram. .
[0041] The trapezoidal prism spacetime corridor defines the set of hard constraints that the aforementioned vehicles must satisfy in the continuous time domain, including those formed by... Intersection traffic constraints consisting of a defined effective green light window and predicted stopping duration The station operation constraints are formed by the defined mandatory stay area.
[0042] use Piecewise Bézier curves represent the longitudinal displacement trajectory of a vehicle. Provide a parameterized description:
[0043] in, This is a sequence of control points for the curve. By utilizing the convex hull property of Bézier curves, complex nonlinear spatiotemporal constraints (such as traffic light phase boundaries and queue dissipation boundaries) are transformed into control points. Linear boundary constraints.
[0044] By using the parameterized expression of the trajectory, the state variables (such as velocity) in the original objective function are transformed. acceleration (etc.) are represented as control points The derivative function form of . The first The cost of the stage Redefining it as, under the constraints of the spatiotemporal corridor, regarding control points The cost of continuous integration.
[0045] The feasible control set is defined as:
[0046] A quadratic programming (QP) solver is used to quickly retrieve the optimal control point sequence within each rolling cycle:
[0047] in, For vehicles in The planned longitudinal displacement at any given time. for Bernstein basis functions of order 1 The predicted time domain length for the current rolling optimization cycle. This is a set of drivable spatiotemporal corridors constructed based on traffic constraints. , These represent the planned velocity and planned acceleration determined by the control points, respectively. The permitted speed range for vehicles. This represents the absolute value of the maximum permissible acceleration of the vehicle. For the first The weighting coefficients of the target function.
[0048] Its rolling time-domain control framework is as follows: Partial control: Only the first segment of the control command based on the optimal trajectory is applied to guide the vehicle's movement; Status Updates: Real-time acquisition of vehicle location, speed, and the latest traffic forecast information (such as the latest...) wait); Repeated solution: at the new rolling start point, based on the updated corridor boundary. The trajectory was re-optimized.
[0049] The entire process involves multiple rolling optimizations, and the optimal sub-trajectory segments obtained from each iteration are spliced together in chronological order to obtain a globally optimal smooth vehicle speed sequence covering the entire process. .
[0050] By introducing a trapezoidal prism spatiotemporal corridor and a piecewise Bézier curve optimization mechanism, the discrete intersection queuing and traffic light constraints are transformed into a convex optimization problem in the continuous time domain. This method effectively eliminates the discretization error of traditional dynamic programming (DP) and, under the premise of strictly ensuring traffic safety and station stopping physical constraints, achieves the generation of infinitely differentiable smooth energy-saving trajectories, significantly improving driving comfort and energy efficiency.
[0051] S2: Based on the globally optimal smooth vehicle speed trajectory and its corresponding acceleration trajectory, the required power of the whole vehicle is calculated, and the required power of the whole vehicle is converted into the actual output power of the battery. The SOC state transition equation is obtained through the battery equivalent circuit model. The upper and lower limits of SOC are used as design factors and the vehicle load is used as noise factors. The signal-to-noise ratio is calculated and the feasible region of the optimal SOC trajectory is found. The optimal SOC reference trajectory is generated based on the center line of the feasible region of the optimal SOC trajectory.
[0052] In step S2, based on the globally optimal smooth vehicle speed trajectory and its corresponding acceleration trajectory Calculate the required power of the vehicle on a continuous time axis. For the stopping and starting conditions near the station, the formula for calculating the vehicle's required power is as follows: in, for The power required by the vehicle at any given time. For powertrain transmission efficiency. , , , These represent the vehicle mass, gravitational acceleration, rolling resistance coefficient, and road slope angle, respectively. , , These represent the air drag coefficient, frontal area, and air density, respectively. This is the rotational mass conversion factor. From the optimal displacement trajectory The planning acceleration obtained by taking the second derivative with respect to time. To assist in system power consumption, The globally optimal smooth vehicle speed trajectory. This is the acceleration trajectory.
[0053] The power requirement of the whole vehicle Converted into actual output power at the battery terminal :
[0054] in, This represents the total electrical power actually output from the battery. To drive the power demand of machinery under operating conditions, Recoverable mechanical power during braking conditions Battery discharge efficiency encompasses inverter losses, motor drive losses, and transmission system losses. The braking energy recovery charging efficiency encompasses both motor power generation efficiency and battery charging reception efficiency.
[0055] The change in state of charge (SOC) is described using a battery equivalent circuit model. The total journey prediction time is defined as... The SOC state transition equation is expressed as: ;in, The time rate of change of SOC For battery output power, Open circuit voltage, This refers to the battery's internal resistance. This refers to the battery's rated capacity.
[0056] The Taguchi Robust Design Method (TRD) is introduced. In the TRD process, the upper and lower limits of State of Charge (SOC) are treated as design factors, and vehicle load is treated as a noise factor. For each set of experiments (i.e., a specific set of SOC boundaries) in the TRD inner and outer tables, the signal-to-noise ratio is calculated. :
[0057] in, For the number of experiments, For the first The total electrical energy consumed by the vehicle in this experiment to complete the entire driving task is calculated using the following formula: Through TRD optimization, a set of feasible regions for the optimal SOC trajectory that maximizes the signal-to-noise ratio and is least sensitive to noise factors is found. .
[0058] The centerline of the feasible region of the optimal SOC trajectory obtained from the above robust planning is used as the optimal SOC reference trajectory, that is:
[0059] The cloud ultimately outputs the optimal SOC reference trajectory to the downstream vehicle control system. .
[0060] By introducing the Taguchi Robust Design (TRD) method, load fluctuations are treated as noise factors for signal-to-noise ratio optimization, thus finding the optimal SOC reference region that is insensitive to load changes. This mechanism provides a power allocation scheme with anti-interference capabilities for vehicle-side control through cloud-based robustness assessment, effectively reducing the risk of energy efficiency instability caused by switching between full load and no-load conditions during actual operation.
[0061] S3: Acquire the real-time operating status of the vehicle and surrounding traffic information, and calculate the vehicle speed error and SOC error corresponding to the globally optimal smooth vehicle speed trajectory and the optimal SOC reference trajectory; integrate the vehicle speed error, SOC error, real-time vehicle operating status and surrounding traffic information as the state space of reinforcement learning, and construct the immediate reward function with the vehicle demand acceleration and motor output torque as the action space; realize the mapping from the state space to the dynamic space by training the policy network, and update the network parameters by maximizing the future cumulative reward to achieve adaptive tracking execution of the globally optimal smooth vehicle speed trajectory and the optimal SOC reference trajectory.
[0062] In step S3, the vehicle-side control system acquires the current operating status of the vehicle in real time during each sampling period, including vehicle speed, battery SOC, power system operating status, and surrounding traffic information obtained locally, and compares it with the target vehicle speed (globally optimal smooth vehicle speed trajectory) provided by the cloud. and optimal SOC reference trajectory Compare them.
[0063] An adaptive control mechanism based on reinforcement learning is introduced, as follows: state space Integrate planning and tracking errors (vehicle speed errors) in each control cycle SOC error Real-time vehicle status (current speed) Battery SOC acceleration And local environmental information sensed by onboard sensors (such as the relative distance to the vehicle in front). (Relative velocity). Motion space To directly drive the execution commands of the vehicle, including the vehicle's required acceleration. and motor output torque .
[0064] In the reinforcement learning framework, the immediate reward function comprehensively characterizes the tracking accuracy and energy consumption performance of cloud-based planning, including tracking error terms, energy consumption penalty terms, safety penalty terms, comfort penalty terms, and policy penalty terms, specifically expressed as follows: in, , , , , , These are the weighting coefficients for each item. For SOC deviation, For speed deviation, This refers to the instantaneous output power of the battery. For following vehicle safety penalty function, For acceleration command penalty items, Impose additional penalties for parking at intersections.
[0065] Tracking error term: This represents the SOC deviation term. Indicates the speed deviation term; Energy consumption penalty items: This represents the instantaneous output power of the battery, and its calculation formula is: ,in This is a battery power calculation function characterized by motor efficiency MAP, which includes charge and discharge characteristics; Safety penalties: The following vehicle safety penalty function is calculated using the following formula: in, The weighting coefficient for safety penalties; Comfort penalty items: This is an acceleration command penalty term, which is directly calculated by squaring the required acceleration command value output by the autonomous controller. Strategy penalty items: Impose additional penalties for parking at intersections.
[0066] The tracking mechanism of the vehicle-side controller is implemented through deep reinforcement learning, which essentially involves training a policy network. Learning from real-time states To the optimal control action mapping function The specific tracking process is as follows: In each control cycle, the network receives data including tracking error. Vehicle status and environmental information status The network outputs actions through multiple layers of nonlinear computation. That is, the acceleration of demand. and motor torque After the action is executed, the vehicle enters a new state. And based on the reward function Calculate instant rewards.
[0067] Policy Network parameters Continuously optimize to maximize future cumulative rewards Updates are performed. The controller automatically learns patterns by analyzing massive amounts of "state-action-reward" data pairs, ultimately training a mature policy network. This enables adaptive and robust tracking of cloud-planned trajectories, such as... Figure 5 As shown.
[0068] An adaptive tracking execution framework based on deep reinforcement learning (RL) is constructed, utilizing a policy network to achieve optimal mapping from real-time perception state to target trajectory. This method not only achieves high-precision tracking of planning instructions but also compensates for deviations caused by dynamic model drift and traffic fluctuations through the self-learning capability of RL, ensuring the closed-loop robust execution of the global energy-saving strategy in complex environments.
[0069] The "cloud-based long-cycle planning - vehicle-side short-cycle execution" collaborative framework proposed in this embodiment organically integrates continuous spatiotemporal trajectory optimization, TRD robust energy distribution, and RL adaptive control to form a fully closed-loop energy-saving control system. This framework can dynamically adjust the operating rhythm according to traffic environment and passenger flow characteristics, maximizing the system-level energy-saving potential of new energy buses while ensuring punctuality.
[0070] Example 2 This embodiment provides a cloud-supported energy management and control system for connected buses based on energy-saving driving, including the following modules: The vehicle speed planning module is configured to: acquire traffic flow data, signal timing, intersection queue length, queue dissipation time and predicted number of vehicles that can pass, combine the predicted stop time window of bus stops, construct a multi-objective vehicle speed planning optimization model that satisfies green light time window constraints, following safety constraints and vehicle dynamics constraints, and solve the multi-objective vehicle speed planning optimization model to obtain the globally optimal smooth vehicle speed trajectory. The SOC optimization module is configured to: calculate the required power of the vehicle based on the globally optimal smooth vehicle speed trajectory and its corresponding acceleration trajectory, convert the required power of the vehicle into the actual output power of the battery, obtain the SOC state transition equation through the battery equivalent circuit model; use the upper and lower limits of SOC as design factors and the vehicle load as noise factors to calculate the signal-to-noise ratio and find the feasible region of the optimal SOC trajectory; and generate the optimal SOC reference trajectory based on the centerline of the feasible region of the optimal SOC trajectory. The tracking control module is configured to: acquire the real-time operating status of the vehicle and surrounding traffic information, and calculate the vehicle speed error and SOC error corresponding to the globally optimal smooth vehicle speed trajectory and the optimal SOC reference trajectory; integrate the vehicle speed error, SOC error, real-time vehicle operating status and surrounding traffic information as the state space of reinforcement learning, and construct an immediate reward function using the vehicle demand acceleration and motor output torque as the action space; realize the mapping from the state space to the dynamic space by training the policy network, and update the network parameters by maximizing the future cumulative reward to achieve adaptive tracking execution of the globally optimal smooth vehicle speed trajectory and the optimal SOC reference trajectory.
[0071] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
[0072] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.
Claims
1. A cloud-supported energy management and control method for connected buses based on energy-saving driving, characterized in that, include: The system acquires traffic flow data, signal timing, intersection queue length, queue dissipation time, and predicted number of vehicles. Combined with the predicted stop time window of bus stops, it constructs a multi-objective vehicle speed planning optimization model that satisfies green light time window constraints, following safety constraints, and vehicle dynamics constraints. The system solves the multi-objective vehicle speed planning optimization model to obtain the globally optimal smooth vehicle speed trajectory. Based on the globally optimal smooth vehicle speed trajectory and its corresponding acceleration trajectory, the required power of the whole vehicle is calculated and converted into the actual output power of the battery. The SOC state transition equation is obtained through the battery equivalent circuit model. The upper and lower limits of SOC are used as design factors and vehicle load is used as noise factors. The signal-to-noise ratio is calculated and the feasible region of the optimal SOC trajectory is found. The optimal SOC reference trajectory is generated based on the center line of the feasible region of the optimal SOC trajectory. The system acquires real-time vehicle operating status and surrounding traffic information, and calculates the vehicle speed error and SOC error corresponding to the globally optimal smooth vehicle speed trajectory and the optimal SOC reference trajectory. The vehicle speed error, SOC error, real-time vehicle operating status, and surrounding traffic information are integrated into the state space of reinforcement learning, and the vehicle demand acceleration and motor output torque are used as the action space to construct an immediate reward function. The system maps from the state space to the dynamic space by training a policy network, and updates the network parameters by maximizing future cumulative rewards to achieve adaptive tracking and execution of the globally optimal smooth vehicle speed trajectory and the optimal SOC reference trajectory.
2. The cloud-supported energy management and control method for connected buses based on energy-saving driving as described in claim 1, characterized in that, The objective function of the multi-objective vehicle speed planning and optimization model is: ;in, For discrete sampling times, The total number of planned steps, For the site serial number, The total number of stations, , , , , , All are weighting coefficients. For instantaneous energy consumption, For time deviation term, For the sequence of queue dissipation at intersections, For speed smoothing term, For ride comfort, This refers to the parking prediction item for the station.
3. The cloud-supported energy management and control method for connected buses based on energy-saving driving as described in claim 1, characterized in that, The green light time window constraint is a constraint on the time when vehicles arrive at the intersection, ensuring that they arrive and pass through the intersection within the valid green light time window. The valid green light time window is represented as follows: ,in, When the green light turns on. The end of the green light. This refers to the queue dissipation time in the predicted results; The following safety constraint is expressed as follows: ;in, The relative distance between the vehicles. For a safe following distance, The relative speed of the vehicles; The vehicle dynamics constraints are: ;in, for Vehicle speed at all times , These are the upper and lower speed limits, respectively. for Acceleration at all times , These are the upper and lower limits of acceleration, for Accelerate constantly , These are the upper and lower limits of the jerk, respectively.
4. The cloud-supported energy management and control method for connected buses based on energy-saving driving as described in claim 1, characterized in that, A Bézier curve trajectory optimization method based on spatiotemporal corridor constraints is used to solve the multi-objective vehicle speed planning optimization model.
5. The cloud-supported energy management and control method for connected buses based on energy-saving driving as described in claim 1, characterized in that, The formula for calculating the required power of the entire vehicle is as follows: in, for The power required by the vehicle at any given time. For powertrain transmission efficiency. , , , These represent the vehicle mass, gravitational acceleration, rolling resistance coefficient, and road slope angle, respectively. , , These represent the air drag coefficient, frontal area, and air density, respectively. This is the rotational mass conversion factor. From the optimal displacement trajectory The planning acceleration obtained by taking the second derivative with respect to time. To assist in system power consumption, The globally optimal smooth vehicle speed trajectory. This is the acceleration trajectory.
6. The cloud-supported energy management and control method for connected buses based on energy-saving driving as described in claim 1, characterized in that, The SOC state transition equation is expressed as follows: ;in, The time rate of change of SOC For battery output power, Open circuit voltage, This refers to the battery's internal resistance. This refers to the battery's rated capacity.
7. The cloud-supported energy management and control method for connected buses based on energy-saving driving as described in claim 1, characterized in that, The immediate reward function includes a tracking error term, an energy consumption penalty term, a safety penalty term, a comfort penalty term, and a strategy penalty term, specifically expressed as follows: in, , , , , , These are the weighting coefficients for each item. For SOC deviation, For speed deviation, This refers to the instantaneous output power of the battery. For following vehicle safety penalty function, For acceleration command penalty items, Impose additional penalties for parking at intersections.
8. A cloud-supported energy management and control system for connected buses based on energy-saving driving, characterized in that: Includes the following modules: The vehicle speed planning module is configured to: acquire traffic flow data, signal timing, intersection queue length, queue dissipation time and predicted number of vehicles that can pass, combine the predicted stop time window of bus stops, construct a multi-objective vehicle speed planning optimization model that satisfies green light time window constraints, following safety constraints and vehicle dynamics constraints, and solve the multi-objective vehicle speed planning optimization model to obtain the globally optimal smooth vehicle speed trajectory. The SOC optimization module is configured to: calculate the required power of the vehicle based on the globally optimal smooth vehicle speed trajectory and its corresponding acceleration trajectory, convert the required power of the vehicle into the actual output power of the battery, obtain the SOC state transition equation through the battery equivalent circuit model; use the upper and lower limits of SOC as design factors and the vehicle load as noise factors to calculate the signal-to-noise ratio and find the feasible region of the optimal SOC trajectory; and generate the optimal SOC reference trajectory based on the centerline of the feasible region of the optimal SOC trajectory. The tracking control module is configured to: acquire the real-time operating status of the vehicle and surrounding traffic information, and calculate the vehicle speed error and SOC error corresponding to the globally optimal smooth vehicle speed trajectory and the optimal SOC reference trajectory; integrate the vehicle speed error, SOC error, real-time vehicle operating status and surrounding traffic information as the state space of reinforcement learning, and construct an immediate reward function using the vehicle demand acceleration and motor output torque as the action space; realize the mapping from the state space to the dynamic space by training the policy network, and update the network parameters by maximizing the future cumulative reward to achieve adaptive tracking execution of the globally optimal smooth vehicle speed trajectory and the optimal SOC reference trajectory.
9. A computer-readable storage medium having a program stored thereon, characterized in that, When executed by the processor, the program implements the steps in the cloud-supported energy management control method for connected buses based on energy-saving driving as described in any one of claims 1-7.
10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the cloud-supported energy management control method for connected buses based on energy-saving driving as described in any one of claims 1-7.