Hybrid heavy truck energy management method

Abstract

The present application relates to the technical field of energy management method, and particularly relates to a hybrid heavy truck energy management method, which comprises the following steps: a first step of system initialization and state sensing; a second step of distributed prediction and problem construction; a third step of online optimization solution and coordination; a fourth step of control instruction distribution and execution; and a fifth step of real-time iteration. Through the alternating use of offline maps and online maps, the method can significantly optimize energy consumption control, improve system robustness, and reduce comprehensive operating costs in the energy management of the hybrid heavy truck.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy management methods, and particularly to a hybrid heavy truck energy management method. BACKGROUND

[0002] With the continuous progress of society, the transportation industry has developed rapidly. In the transportation industry, heavy trucks are an important component. At present, most heavy trucks use parallel hybrid power systems. In actual use, the coupling of mechanical, electrical and thermal energy flows introduces a large number of state variables and control variables, resulting in a very complex situation for collaborative optimization control. The traditional single optimization algorithm has a huge amount of calculation and is easy to fall into local optimum, which has the following defects: insufficient use of prediction information: existing rule-based or instantaneous optimization strategies (such as equivalent fuel consumption minimization strategy-ECMS) lack foresight for global road condition information (such as slope, speed limit), resulting in suboptimal energy distribution. Poor calculation real-time performance: traditional dynamic programming (DP) and model predictive control (MPC) have too much calculation burden in long-distance scenarios and are difficult to apply in real time on board. Lack of multi-time scale collaboration: a single control layer cannot respond to second-level power distribution and minute-level road conditions at the same time, resulting in deep battery charging and discharging or frequent engine start-stop. Insufficient robustness to uncertain working conditions: when the actual road conditions (traffic flow, slope error) fluctuate, fixed rule strategies are easily ineffective.

[0003] CN116394914A discloses a hybrid electric vehicle predictive energy management method considering SOC trajectory, belonging to the field of plug-in hybrid electric vehicle control. First, based on the vehicle driving path and the target electric quantity consumption, the global target SOC trajectory is linearly solved, and the optimal torque distribution is solved in the prediction time domain using a dynamic programming algorithm to obtain a reference SOC trajectory. Then, taking the SOC following effect and the optimal engine fuel consumption as the target, a linear quadratic output tracking controller is used to solve the optimal engine power. Finally, based on the rules, the demand torque of each power source is calculated, and the control quantity is sent to the corresponding controller to realize torque distribution. The purpose of the present application is to solve the optimal torque distribution in the prediction time domain while controlling the electric quantity consumption in the vehicle driving path, that is, to fully utilize the prediction information of future working conditions to improve the vehicle economy while achieving target electric energy consumption control. The above method has the following defects: it completely relies on global driving path prediction, and dynamic traffic information is difficult to accurately obtain in actual application; the dynamic programming calculation has high complexity and limited real-time performance.

[0004] CN118219935A discloses a hybrid electric vehicle intelligent energy management method based on slope prediction, comprising the following steps: establishing a model of each energy source in the hybrid electric vehicle and a model of the whole vehicle; according to the slope and speed information of the vehicle at the current time, the total demand power is obtained by prediction through a long short-term memory network, the power is processed in layers by using a fuzzy control adaptive low-pass filter and the construction idea of equivalent consumption minimization strategy, a hybrid electric vehicle energy management system based on slope prediction is constructed, the total equivalent hydrogen consumption of each energy source, the service life of the fuel cell and the SOC of the energy storage system under the predicted slope are used as learning goals, and an efficient hybrid experience replay technology is designed to update the network in the intelligent algorithm. The present application can optimize and adjust the energy management of the hybrid electric vehicle on the basis of considering the road slope information, and better update the optimal strategy in complex and random driving environment by real-time prediction of road slope. The defects of the above management method are that the slope prediction depends on high-precision map data, and the applicability in mountainous areas or un-mapped areas is poor; the fuel cell life modeling does not consider the actual attenuation.

[0005] CN119773724A discloses a hybrid truck energy management method and device, and a commercial vehicle, belonging to the technical field of commercial vehicle energy management. During vehicle operation, the vehicle controller obtains commercial vehicle operating state information and transmits it to the hybrid truck control module; the hybrid truck control module inputs the commercial vehicle operating state information into the hybrid truck model; the hybrid truck model calculates the current engine operating point according to the engine speed and torque, and estimates the supercharging parameters of the electronic supercharger based on the exhaust temperature and exhaust flow; based on the vehicle's driving resistance and the driver's operation intention, the torque distribution of the motor and the charging and discharging state of the battery are determined. Reasonable allocation of motor torque assistance realizes the balance optimization of power and energy consumption, improves fuel economy and power performance. The engine power output is improved; the working efficiency and reliability of the electronic supercharger are also improved. The defects of the above device and method are that the traffic information prediction function is not integrated, the predictive control ability is weak; the engine operating point optimization is only based on the current state. SUMMARY

[0006] The technical problem to be solved by the present application is to overcome the shortcomings of the prior art and provide a hybrid electric heavy truck energy management method. In heavy truck energy management, the alternative use of offline maps and online maps can significantly optimize energy consumption control, improve system robustness, and reduce overall operating costs.

[0007] The technical solution adopted by the present application to solve the technical problem is: a hybrid electric heavy truck energy management method, comprising the following steps:

[0008] Step 1, system initialization and state awareness: the vehicle controller collects real-time data from each subsystem controller through the CAN bus network, and the system state includes the current vehicle speed, the state of charge of the power battery, the engine speed and torque, the motor speed and torque, the vehicle demand power, and the power demand of the accessory system;

[0009] Step 2, distributed prediction and problem construction: based on multi-source map information fusion and processing, combined with offline map data and online map information, global reference trajectory planning is carried out, a vehicle speed prediction model and a SOC trajectory prediction model are established, and reference vehicle speed trajectory and SOC target trajectory are generated; the vehicle speed prediction model introduces driver reaction delay and vehicle power lag effect, and the SOC trajectory prediction model dynamically adjusts the SOC target based on the driving distance; including the following steps:

[0010] Step 21, vehicle speed prediction: a vehicle speed prediction model is established, and driver reaction delay and vehicle power lag effect are introduced, and the improved speed prediction equation is:

[0011] (1)

[0012] In formula (1), the meaning of the symbol is:

[0013] represents the reaction time;

[0014] represents the driving distance, which is half of the vehicle headway;

[0015] x represents the vehicle speed distribution prediction error within 30 minutes in the future, which is not greater than 3 km / h;

[0016] q represents the mapping of distance and time;

[0017] t represents the driving time;

[0018] v represents the vehicle speed;

[0019] ρ represents the mapping of driving distance and time;

[0020] Step 3, online optimization solution and coordination: a distributed model predictive control (DMPC) algorithm is used to decompose the global optimization problem into multiple local optimization sub-problems, each intelligent agent solves its own optimal control sequence, and the system overall coordination is realized through information interaction; a rolling optimization mechanism is adopted, that is, in each control period, the optimal control sequence within a limited time domain in the future is solved;

[0021] Step 4, control instruction distribution and execution: the vehicle controller sends the optimal power distribution instructions obtained through optimization calculation to the corresponding subsystem controllers through the CAN bus, and each controller executes the instructions to drive the vehicle to run;

[0022] Step 5, real-time iteration: Steps 1-4 are executed in a very short control cycle, and prediction, optimization and energy distribution are performed again according to the latest state information to achieve adaptive optimization control for changes in driving conditions and driver's intention.

[0023] Further limitation, the distributed prediction and problem construction of the second step includes the following steps:

[0024] Step 22, SOC trajectory prediction: establish a SOC reference trajectory prediction model, dynamically adjust the SOC target based on the driving distance to avoid sudden drop of power, and the SOC reference trajectory is calculated as follows:

[0025] (2)

[0026] In formula (2), the meaning of the symbol is:

[0027] represents the initial SOC value;

[0028] represents the reference SOC value;

[0029] represents the target SOC value;

[0030] L represents the full charge range;

[0031] represents the slope adaptive coefficient;

[0032] d represents the time differential;

[0033] The target function is optimized and calculated, and the power battery thermal management safety is controlled as a constraint condition while ensuring that the reference trajectory meets the lowest energy consumption requirement of the vehicle; constraint conditions: , engine torque limit: ; In the target function, the meaning of the symbol is:

[0034] represents the fuel consumption;

[0035]

[0036] ​​SOC value at K moment;

[0037] engine torque;

[0038] engine speed;

[0039] weighting coefficient, respectively 0.7, 0.2, 0.1;

[0040] △T eng average time interval;

[0041] Power battery safety requirements: ; wherein, motor current;

[0042] Power battery thermal management safety constraints are described by the battery temperature rise model, the formula is as follows:

[0043] (3)

[0044] In formula (3), the meaning represented by the symbol is:

[0045] battery temperature change;

[0046] battery current;

[0047] When , then the discharge power is forced to limit 20%.

[0048] Further limitation, in the above technical solution, the distributed prediction and problem construction of the second step includes the following steps:

[0049] The 23rd step, the reference vehicle speed trajectory Vref(t) and the SOC target trajectory SOCref(t) are generated.

[0050] Further limitation, in the above technical solution, the distributed prediction and problem construction of the second step includes the following steps:

[0051] The 24th step, within 10-30S time, according to the reference trajectory provided by the long-time planning layer and the real-time environmental feedback, the efficient coordination of power system and the stable following of vehicle queue are realized.

[0052] Further limitation, in the above technical solution, each subsystem controller mainly includes battery management system BMS, engine controller ECU and motor controller MCU.

[0053] The beneficial effects of the present application are: the hybrid heavy truck energy management method provided by the present application can significantly optimize energy consumption control, improve system robustness, and reduce comprehensive operating costs in heavy truck energy management, that is, through the deep fusion and dynamic switching of offline maps and online maps, combined with a hierarchical distributed predictive control architecture, the unified improvement of foresight, real-time, robustness and economy is realized.

[0054] I. Prospective energy planning: using the global path information of the offline map and the real-time traffic data of the online map, the optimal vehicle speed and SOC reference trajectory are generated, so that the power system can predict the future road conditions, avoid frequent engine start-stop and deep battery charging and discharging, and optimize energy consumption from the source.

[0055] II. Efficient real-time optimization: the distributed model predictive control (DMPC) algorithm is adopted, the complex global problem is decomposed into local sub-problems for parallel solving, and the decision is made within a millisecond-level control cycle through the rolling optimization mechanism, which significantly reduces the computational burden and meets the real-time requirements of vehicle-mounted.

[0056] III. Enhance system robustness: the dynamic switching mechanism of the hybrid map ensures the continuity of control in areas with poor network signals (such as mountainous areas and tunnels); at the same time, the online optimization layer can quickly respond to working condition fluctuations based on real-time sensor data, overcoming the failure of fixed rule strategies in uncertain environments.

[0057] IV. Reduce comprehensive operating costs: precise energy management directly improves fuel economy; offline calculation reduces dependence on the cloud and continuous data traffic, reducing communication costs; in addition, smooth power distribution and battery thermal management constraints also extend the service life of key components. BRIEF DESCRIPTION OF DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments described in the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0059] Figure 1 is a structure diagram of the hierarchical control framework in the present application;

[0060] Figure 2 is an implementation flow diagram in the present application;

[0061] Figure 3 is a schematic diagram of the core hardware modules inside the controller and their connection relationship with all key external systems. DETAILED DESCRIPTION

[0062] In order to make the technical problems, technical solutions and beneficial effects solved by the present application clearer, the present application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0063] In the present application, multi-source map information fusion and processing: offline map data and online map information are fused to realize global planning; the triggering and switching conditions of the multi-source map are set.

[0064] The hybrid map dynamic switching mechanism is as follows: triggering condition: switch to offline mode when the network signal strength ≤-90dBm; switch to online mode after the signal is recovered and stable for >10s; key sections (such as long steep slopes, tunnels) are forced to add offline data packets. Data fusion strategy: online data covers offline data first, and real-time information is used in case of conflict; offline data supplements online data blind area (such as signal-free areas in mountainous areas)

[0065] The global domain control layer (global planning) in the present application: based on the real-time online map navigation path, slope data and historical working conditions, a spatial domain reference trajectory (vehicle speed, SOC target) is generated; the path is segmented and clustered, and the driving path is divided into several sections according to the slope and congestion state, and the SOC (State of Charge, state of charge) reference trajectory of each section is pre-calculated.

[0066] The real-time domain control layer (real-time optimization) in the present application: based on the reference trajectory output by the global planning layer, combined with real-time sensor data (vehicle speed, acceleration, battery SOC), the power distribution is adjusted at a second level. Model predictive control (MPC) minimizes tracking error and energy consumption targets, and rolls the engine and motor torque distribution.

[0067] See Figure 1 , Figure 2 and Figure 3 , a hybrid energy management method for a hybrid heavy truck, comprising the following steps:

[0068] Step 1, system initialization and state perception: the vehicle controller collects real-time data from each subsystem controller through the CAN bus network, and the system state includes the current vehicle speed, the state of charge of the power battery, the engine speed and torque, the motor speed and torque, the vehicle demand power and the accessory system power demand;

[0069] Step 2, distributed prediction and problem construction: based on multi-source map information fusion and processing, combining offline map data and online map information, global reference trajectory planning is performed, a vehicle speed prediction model and a SOC trajectory prediction model are established, and a reference vehicle speed trajectory and a SOC target trajectory are generated; the vehicle speed prediction model introduces driver reaction delay and vehicle power lag effect, and the SOC trajectory prediction model dynamically adjusts the SOC target based on driving distance;

[0070] Step 3, online optimization solution and coordination: a distributed model predictive control (DMPC) algorithm is used to decompose the global optimization problem into multiple local optimization sub-problems, each agent solves its own optimal control sequence, and the system overall coordination is realized through information interaction; a rolling optimization mechanism is adopted, that is, in each control period, the optimal control sequence in a limited time domain in the future is solved;

[0071] Step 4, control instruction distribution and execution: the optimal power distribution instruction obtained by the vehicle controller through optimization calculation is sent to the corresponding subsystem controller through the CAN bus, and each controller executes the instruction to drive the vehicle to run;

[0072] Step 5, real-time iteration: Steps 1-4 are executed in a very short control period, and prediction, optimization and energy distribution are performed again according to the latest state information to realize adaptive optimization control for changes in driving conditions and driver's intention. This process is repeated in a very short control period. The system will re-predict, optimize and distribute energy based on the latest state information every time the vehicle passes through a control period, so as to realize adaptive optimization control for the changing driving conditions and driver's intention.

[0073] Each subsystem controller mainly includes a battery management system BMS, an engine controller ECU and a motor controller MCU.

[0074] The second step of distributed prediction and problem construction includes the following steps:

[0075] Step 21, vehicle speed prediction model: driver reaction delay and vehicle power lag effect are introduced, and the improved speed prediction equation is:

[0076] (1)

[0077] In formula (1), the meaning of the symbol is:

[0078] represents the reaction time;

[0079] represents the driving distance, which is half of the headway;

[0080] x represents the prediction error of vehicle speed distribution within the next 30 minutes, and this error is no greater than 3 km / h;

[0081] q represents the mapping between distance and time;

[0082] t represents the travel time;

[0083] v indicates vehicle speed;

[0084] ρ represents the mapping between travel distance and time.

[0085] Step 22, SOC Trajectory Prediction: Establish an SOC reference trajectory prediction model, dynamically adjust the SOC target based on driving distance to avoid sudden battery drop, and define the SOC reference trajectory. The calculation is as follows:

[0086] (2)

[0087] The symbols in formula (2) represent the following meanings:

[0088] Indicates the initial SOC value;

[0089] Indicates the reference SOC value;

[0090] Indicates the target SOC value;

[0091] L indicates the range on a full charge;

[0092] This represents the slope adaptive coefficient;

[0093] d represents the time derivative.

[0094] Through the objective function right Perform optimization calculations while ensuring While meeting the minimum energy consumption requirements of the vehicle, the thermal management safety of the power battery is controlled as a constraint; constraint conditions: Engine torque limit: ;

[0095] The symbols in the objective function represent the following meanings:

[0096] Indicates fuel consumption;

[0097] This represents the SOC value at time K;

[0098] Indicates engine torque;

[0099] denotes the engine speed;

[0100] denotes the weighting coefficient, respectively 0.7, 0.2, 0.1;

[0101] △T eng denotes the average time interval.

[0102] Power battery safety requirements: ; wherein, denotes the motor current;

[0103] The power battery thermal management safety constraint is described by a battery temperature rise model, and the formula is as follows:

[0104] (3)

[0105] In formula (3), the meaning represented by the symbol is:

[0106] denotes the battery temperature change;

[0107] denotes the battery current;

[0108] When , then the discharge power is forcibly limited by 20%.

[0109] Step 23, reference vehicle speed trajectory Vref(t), SOC target trajectory SOCref(t);

[0110] Step 24, within 10-30S time, according to the reference trajectory provided by the long-term planning layer and the real-time environmental feedback, realize the efficient coordination of the power system and the stable following of the vehicle queue.

[0111] Wherein, the online optimization solving and coordination of the third step includes the following steps:

[0112] Step 31, distributed solving: DMPC decomposes the complex global optimization problem into multiple more easily handled local optimization problems; each agent independently solves its optimal control sequence, and coordinates with other agents through information interaction to ensure the consistency of the whole system;

[0113] Step 32, rolling optimization: the controller solves the optimal control sequence in the future one limited time domain, and applies the control instruction at the current time to the system; the whole process will be repeated based on the new system state at the next time step, forming the mechanism of "rolling optimization".

[0114] I. System architecture design of the present application

[0115] 1. Information perception module: real-time collection of vehicle state (vehicle speed, battery SOC, motor speed) and environmental data (slope, traffic signal); obtain 3km road topology and traffic flow information in front through V2X.

[0116] 2. Hierarchical control framework, see Figure 1 .

[0117] Upper layer (global planning layer): adopt dynamic programming algorithm, based on 3km road topology information (slope, traffic signal, etc.) obtained by V2X, generate 10-30 minute level reference speed and SOC trajectory with the goal of minimum fuel consumption; lower layer (real-time domain execution layer): adopt model predictive control (MPC), roll optimization of future 5-10 seconds torque distribution and gear selection.

[0118] (1) Environmental perception and data preprocessing

[0119] Multi-source data input: 1) Road topology data: obtain slope, curvature, speed limit information for future 3-5km through offline map, combine with real-time GPS positioning to correct position deviation. 2) Dynamic traffic information: fuse V2X signals (traffic light phase, front vehicle density, congestion prediction) through online map. 3) Vehicle state monitoring: real-time collection of battery SOC, vehicle speed, motor torque, load mass.

[0120] (2) Data filtering and fusion

[0121] (3) Data processing process: 1) adopt Kalman filter to eliminate sensor noise; 2) time and space alignment: match V2X data with vehicle coordinate system, establish SLT (Spatial-Longitudinal-Temporal) three-dimensional decision space.

[0122] (4) Multi-element map information fusion and processing: 1) Offline map data: store static road network topology, historical slope curve, fixed point speed limit, etc. information. Adopt compression technology to reduce storage space, manage by road section; 2) Online map information: real-time obtain traffic congestion index, sudden weather, temporary construction information, dynamic speed limit. Update every 5-10s through V2X communication, delay <500ms; 3) Hybrid map dynamic switching mechanism: trigger condition: switch to offline mode when network signal strength ≤-90dBm; switch to online mode when signal is restored and stable for more than 10s; key road (such as long steep slope, tunnel) forced to add offline data packet. Data fusion strategy: online data covers offline data first, real-time information as the priority in conflict; offline data supplements online data blind area (such as signal-free area in mountainous area).

[0123] (5) Global Reference Trajectory Planning Layer (Time Scale: 10-30 minutes): Responsible for global path planning and energy allocation. Its core logic lies in generating the optimal reference trajectory for vehicle speed and power battery SOC through spatial domain optimization. This trajectory is used to guide the actual operation of the vehicle and provide macro-level decision-making basis for the real-time optimization layer.

[0124] (6) Operating Condition Identification and Classification: Collect historical driving data (vehicle speed, acceleration, gradient) and divide typical operating condition segments (high-speed cruising / city start-stop / mountain climbing) through k-means clustering; 1) High-speed cruising: Maintain the engine's high-efficiency range (1200-2000rpm). 2) City start-stop: Predict the timing of brake energy recovery (dynamically increase the SOC target value by 5%). 3) Long uphill sections: Discharge in advance (reduce the SOC target value by 8%) to avoid the engine's inefficient zone.

[0125] (7) Reference Trajectory Planning: Inputs: Road topology (slope, curvature) in high-precision map, V2X traffic flow data (traffic light phases, congestion prediction), initial battery SOC. Reference Trajectory Planning: 1) Vehicle Speed ​​Prediction Model: Introducing driver reaction delay and vehicle power lag effect to improve the speed prediction equation: ,in, This refers to the reaction time (approximately 1-2 seconds). The distance traveled is half the distance from the front of the vehicle; x represents the speed distribution prediction error within the next 30 minutes, which is no greater than 3 km / h; 2) SOC trajectory prediction model: dynamically adjust the SOC target based on the travel distance to avoid sudden battery drop, and use the SOC reference trajectory. The calculation is as follows: ,in, This is the initial SOC value; For reference SOC value; Target SOC value; L is the range on a full charge; For slope adaptation coefficients (to reserve more power on steep slopes); the following objective function is used to... Perform optimization calculations to ensure The reference trajectory can meet the requirement of minimizing the energy consumption of the entire vehicle, while controlling it with the thermal management safety of the power battery as a constraint. Constraints: Engine torque limit ,in, Fuel consumption; SOC at time K; This refers to engine torque; Engine speed; The weighting coefficients are 0.7, 0.2, and 0.1 respectively (dynamically adjusted to 0.4 for high-temperature conditions). Power battery safety requirements: wherein, is the motor current; power battery thermal management safety constraints, battery temperature rise model; wherein, is the battery temperature change; is the battery current; when , the discharge power is forcibly limited by 20%. Output: reference vehicle speed trajectory Vref(t), SOC target trajectory SOCref(t). The real-time domain control layer (time scale: 5-10 seconds) is the key link connecting the global planning layer and the execution layer, and its main goal is to realize the efficient coordination of the power system and the stable following of the vehicle queue within a limited prediction time domain (usually 10 to 30 seconds) based on the reference trajectory provided by the global planning layer and the real-time environmental feedback.

[0126] (8) Real-time dynamic optimization model: input: 1) Real-time data synchronization in vehicle-mounted sensors, camera: capture the position of the vehicle in front, lane line, traffic sign (sampling rate ≥20Hz), output image sequence. Millimeter wave radar: detect relative speed, distance (accuracy ±0.1m), azimuth angle (accuracy ±0.5°), coverage range 200m. Inertial measurement unit (IMU): real-time acquisition of vehicle yaw rate, longitudinal / lateral acceleration (noise <0.01g), used for road curvature calculation. 2) Real-time dynamic trajectory planning, based on the dynamic optimization model of DMPC: hybrid heavy truck (heavy truck) uses distributed model predictive control (DMPC) for real-time energy management, and the core purpose is to coordinate multiple energy sources such as engine, motor / generator and battery, to realize the optimal fuel economy under the premise of ensuring power performance. Its core calculation process is a closed-loop process of online rolling optimization.

[0127] Step one: system initialization and state perception: at the beginning of each control period, the vehicle control system (VMS) needs to collect real-time data from various subsystem controllers (such as battery management system BMS, engine controller ECU, motor controller MCU, etc.) through the CAN bus network. Key state variables: mainly including current vehicle speed, state of charge (SOC) of power battery, engine speed and torque, motor speed and torque, vehicle demand power (calculated by accelerator pedal and current driving resistance) and accessory system power demand, etc. System self-check: before obtaining data, system self-check is usually performed to determine whether there is a fault in each subsystem, to ensure that the control command can be executed on a healthy system.

[0128] Step two: Distributed prediction and problem formulation: This is the core of DMPC. The controller formulates a local optimization problem for each energy source (treated as an agent) based on the current state. Prediction model: Dynamic mathematical models are established for key components such as the engine, motor, and battery. These models are used to predict the evolution of future states (such as SOC changes and fuel consumption) under a given control input sequence. Objective function: The goal of each local optimization problem is usually to minimize the total "cost" within the prediction horizon. This cost function (J) is usually a weighted sum of fuel consumption, electricity consumption (converted to equivalent fuel consumption), and a penalty term for drastic changes in control commands. Constraints: The optimization must strictly adhere to various physical and safety constraints, such as engine maximum and minimum speed and torque, motor maximum and minimum torque, battery SOC upper and lower limits, battery maximum charge and discharge power, etc.

[0129] Step three: Online optimization solution and coordination: Each energy source agent solves its own local optimization problem in parallel, taking into account the predicted information of neighboring agents (such as vehicle demand and other energy source states). Distributed solution: DMPC breaks down the complex global optimization problem into multiple more manageable local optimization sub-problems. Each agent (such as the engine controller and motor controller) independently solves its optimal control sequence and exchanges necessary prediction information (such as future power demand prediction) with other agents through communication to achieve system-wide coordination and consistency. Rolling optimization: Instead of solving the optimal solution for the entire trip (which is the dynamic programming DP approach), the controller only solves the optimal control sequence within a limited time horizon (prediction horizon) and only applies the current time control command to the system. At the next time step, the entire process is repeated based on the new system state, forming a "rolling optimization" mechanism.

[0130] Step four: Control command distribution and execution: After solving the optimization problem, the vehicle controller (VMS) sends the calculated optimal power distribution commands (such as the power the engine should provide and the power the motor should provide / absorb) to the corresponding subsystem controllers (BMS, ECU, MCU) through the CAN bus. The controllers execute the commands to drive the vehicle.

[0131] Step five: Real-time iteration: The above steps are repeated constantly in a very short control cycle (usually milliseconds). Every time the vehicle passes through a control cycle, the system will re-predict, optimize and allocate energy based on the latest state information, so as to achieve adaptive optimization control of dynamic changes in driving conditions and driver's intention. Condition-responsive mode switching: The short-time layer triggers specific working modes according to the real-time working condition torque demand of the power source; Starting / climbing: dual power source (engine + motor) hybrid power supply, providing peak torque; Downhill: motor switches to generating mode, recovering braking energy; Platform cruise: pure electric / hybrid mode based on SOC threshold decision.

[0132] Hardware architecture: The core of the present application is a powerful central computing unit (vehicle controller VMS), which receives prediction information of different time scales through various communication interfaces and coordinates the management of the entire powertrain.

[0133] See Figure 3 , which describes in detail the core hardware modules inside the controller and their connection with all external key systems, reflecting how they integrate long and short-term prediction information and make energy management decisions.

[0134] Core module function interpretation: main control processor (SoC / MPU), which is the "brain" of the controller. For processing complex prediction models (such as MPC) and machine learning algorithms, high-performance multi-core processors (such as ARM Cortex-A series) are required, usually including: 1) Application Processor (AP): responsible for running the upper layer operating system (such as Linux / AUTOSAR Adaptive) and core energy management algorithms. Microcontroller (MCU): responsible for real-time tasks and communication management with high real-time requirements, in line with the AUTOSAR Classic standard. Communication interface module, Ethernet switch: high-speed backbone, used to connect high-precision map positioning module, remote information processing unit (T-Box), and future expandable connection to intelligent driving domain controller. Multi-channel CAN / CAN FD controller: the absolute main force of heavy truck network. Connects BMS, ECU, MCU, TCU, and all powertrain nodes, and collects vehicle sensor data. V2X and wireless communication interface (4G / 5G): the key to long and short-term prediction. Responsible for receiving global traffic information (long-term prediction) issued by the cloud and real-time local road condition information (short-term prediction) from V2X devices. Digital / Analog I / O: used to process hard-wired signals such as wake-up signals, enable signals, and relay control. Storage unit (Flash & RAM): large capacity Flash: used to store operating systems, applications, high-precision map data (cache), and fault logs. Large capacity RAM is used for data storage and calculation during algorithm execution, especially for rolling update of a large amount of state data within the prediction time domain. Power management and hardware security, multi-path power stabilizing circuit: converts the 24V power of the vehicle storage battery into different voltages required by the internal chips of the controller (such as 5V, 3.3V, 1.2V, etc.), which needs to meet the stringent voltage fluctuation and transient disturbance requirements.

[0135] The present application has the following advantages over traditional hybrid heavy truck energy management methods:

[0136] Dynamic energy recovery efficiency improvement: energy recovery optimization based on slope prediction. Road slope data stored in offline maps (such as digital elevation models) can be used to plan brake points in advance, combined with energy recovery systems, to maximize the recovery of electric energy on downhill sections, extending the endurance mileage by 15%-20%. Online maps dynamically adjust the recovery timing through real-time traffic flow data to avoid invalid braking due to congestion ahead.

[0137] Accurate speed curve matching: generate steady-state speed curve based on historical road conditions offline map, combined with online real-time road conditions (such as accident point location) to correct the prediction, so that the hybrid system runs in the best speed range, reducing invalid engine start-stop, and reducing fuel consumption by more than 7%.

[0138] Network fault tolerance and cost control: Signal-free area continuous management: When the network in the mine area or border road section is interrupted, the offline map maintains the basic navigation and energy strategy execution (such as preloaded fixed slope route), ensures the continuous operation of the SOC (state of charge) balancing algorithm, and automatically switches to the real-time mode after online recovery to avoid abnormal fuel consumption risk.

[0139] Traffic and storage balance: The provincial offline map only needs 800MB storage, avoiding the traffic cost generated by the continuous transmission of real-time data by the 4G / 5G module; the online mode supplements dynamic information through incremental update (only 10K-100K / 4 hours), greatly reducing the communication cost of the vehicle fleet level.

[0140] Security and reliability enhancement: Defense against network attacks: Offline mode reduces the exposure of the vehicle-mounted system to the public network, reducing the risk of hacking and tampering with navigation routes through online vulnerabilities (such as high 27 times of hijacking probability of un-updated system).

[0141] Hardware load optimization: Offline calculation shares the pressure of cloud path planning, avoids the lag of control instructions caused by network delay, and ensures the torque response accuracy of heavy-duty vehicles in curved roads and steep slopes.

[0142] Parallel hybrid heavy-duty trucks encounter highly uncertain climate and driving conditions during driving, and the energy consumption coupling relationship between their power systems and thermal management systems is complex. As nonlinear time-varying systems with multiple state variables and multiple control variables, using global optimization algorithms to reveal energy-saving regulation and control mechanisms will result in large computational load and low solution efficiency. One of the key problems the invention aims to solve is how to improve the global optimization efficiency of energy flow coordination management. Therefore, a reference trajectory optimization algorithm for the state variables of the power system and the thermal management system is proposed to achieve online accurate tracking of the controller.

[0143] For the characteristics of fast response of the power system of the parallel hybrid heavy-duty truck and the large hysteresis of the thermal management system, the same optimization algorithm cannot balance the long-term reference planning of state variables such as temperature and battery SOC, and the short-term accurate control of the system, making it difficult to obtain a relatively ideal global optimal energy consumption. By adopting a hierarchical predictive control strategy and combining the reference trajectory planning of system state variables, online accurate tracking of the reference trajectory by the power system execution components is achieved, thereby achieving the accurate control goal of coordinated optimization.

[0144] The above is only the preferred specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can make equivalent substitutions or changes to the technical solutions and inventive concepts of the present application within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method of energy management for a hybrid heavy duty truck, characterized by, Comprising the following steps: Step 1, system initialization and state awareness: the vehicle controller collects real-time data from each subsystem controller through the CAN bus network, and the system state includes the current vehicle speed, the state of charge of the power battery, the engine speed and torque, the motor speed and torque, the vehicle demand power and the power demand of the accessory system; Step 2, distributed prediction and problem construction: based on multi-source map information fusion and processing, combined with offline map data and online map information, global reference trajectory planning is carried out, a vehicle speed prediction model and a SOC trajectory prediction model are established, and a reference vehicle speed trajectory and a SOC target trajectory are generated; the vehicle speed prediction model introduces driver reaction delay and vehicle power lag effect, and the SOC trajectory prediction model dynamically adjusts the SOC target based on the driving distance; comprising the following steps: Step 21, vehicle speed prediction: a vehicle speed prediction model is established, and driver reaction delay and vehicle power lag effect are introduced, and the improved speed prediction equation is: （1） In formula (1), the meaning of the symbol is: represents the reaction time; represents the distance traveled, taken as half the headway; x represents the vehicle speed distribution prediction error within 30 minutes in the future, and the error is not greater than 3km / h; q represents the mapping of distance and time; t represents the driving time; v represents the vehicle speed; ρ represents the mapping of driving distance and time; Step 22, SOC trajectory prediction: Establish a SOC reference trajectory prediction model, dynamically adjust the SOC target based on the driving distance, to avoid sudden drop of power, and the SOC reference trajectory is calculated as follows: （2） In formula (2), the meaning of the symbol is: represents the initial SOC value; represents a reference SOC value; represents a target SOC value; L represents the full charge cruising range; denotes the slope adaptation coefficient; d represents the time differential; Through the objective function right Perform optimization calculations while ensuring While meeting the minimum energy consumption requirements of the vehicle, the thermal management safety of the power battery is controlled as a constraint; constraint conditions: Engine torque limit: ; In the target function, the meaning of the symbol is: represents the fuel consumption; SOC value at time K; represents the engine torque; NE represents engine speed; represent weighting coefficients, respectively taking values of 0.7, 0.2, 0.1; denotes the average time interval; Power battery safety requirements: ; wherein, denotes the motor current; The power battery thermal management safety constraint is described by a battery temperature rise model, and the formula is as follows: （3） In formula (3), the meaning of the symbol is: represents a change in battery temperature; represents the battery current; When then the discharge power is forced limited by 20%; Step 3, online optimization solution and coordination: a distributed model predictive control algorithm is used to decompose the global optimization problem into multiple local optimization sub-problems, each agent solves its own optimal control sequence, and the system overall coordination is realized through information interaction; a rolling optimization mechanism is adopted, that is, in each control period, the optimal control sequence in a limited time domain in the future is solved; Step 4, control instruction distribution and execution: the vehicle controller sends the optimal power distribution instruction obtained by optimization calculation to the corresponding subsystem controller through the CAN bus, and each controller executes the instruction to drive the vehicle to run; Step 5, real-time iteration: the steps 1-4 are executed in a very short control period, and the prediction, optimization and energy distribution are re-performed according to the latest state information, so as to realize adaptive optimization control for driving conditions and driver intention changes.

2. A method of energy management for a hybrid heavy duty truck as set forth in claim 1, characterized by: The distributed prediction and problem construction of the second step comprises the following steps: Step 23, generating a reference vehicle speed trajectory Vref(t) and a SOC target trajectory SOCref(t).

3. A method of energy management for a hybrid heavy duty truck as set forth in claim 1, characterized by: The distributed prediction and problem construction of the second step comprises the following steps: Step 24, within 10-30S, according to the reference trajectory provided by the long-term planning layer and the real-time environmental feedback, the efficient coordination of the power system and the stable following of the vehicle queue are realized.

4. A method of energy management for a hybrid heavy duty truck as set forth in claim 1, characterized by: Each subsystem controller mainly includes a battery management system BMS, an engine controller ECU and a motor controller MCU.

Images