Energy management system based on fuel cell unmanned vehicle

Abstract

The invention discloses an energy management system based on a fuel cell unmanned vehicle, and relates to the field of unmanned vehicle energy control. The system comprises an environment perception module, a decision planning module, an energy management main module and an execution and state monitoring module, the energy management main module comprises a dynamic energy demand prediction and distribution sub-module and other sub-modules, and the modules exchange data through a hybrid communication network. The core innovation of the invention is a dynamic energy demand prediction and distribution module, and through multi-source data preprocessing, energy demand prediction through an improved instantaneous power method, deviation correction through state feedback and dynamic power distribution, energy demand accurate pre-judgment and self-adaptive distribution are realized. The system does not need a large amount of historical data, is low in calculation complexity, can improve the energy utilization efficiency, prolongs the service life of a fuel cell and an energy storage cell, and is suitable for related vehicles under complex urban road conditions.

Description

Technical Field

[0001] This invention relates to the field of energy control for autonomous vehicles, specifically to an energy management system for fuel cell-based autonomous vehicles. Background Technology

[0002] With the integration of new energy vehicles and autonomous driving technology, fuel cells, with their advantages of zero emissions and long driving range, have become an important power source for autonomous vehicles. Currently, the energy management systems of fuel cell autonomous vehicles mostly adopt a fixed mode of "fuel cell as the primary and energy storage battery as the secondary," adjusting the energy distribution ratio through conventional algorithms such as PID control and fuzzy control. This presents two major problems: First, the energy distribution strategy has a low degree of matching with the vehicle's real-time driving needs and dynamic environmental changes. For example, in scenarios involving sudden hill climbs or rapid acceleration, the fuel cell output response is prone to lag, leading to over-discharge of the energy storage battery. Second, existing systems mostly rely on a single energy consumption model, failing to fully utilize the environmental perception capabilities of autonomous vehicles and unable to predict energy demand fluctuations in advance. This results in frequent start-stop cycles of the fuel cell or operation in an inefficient range, shortening the fuel cell's lifespan and reducing the overall vehicle energy utilization efficiency. Summary of the Invention

[0003] The purpose of this invention is to provide an energy management system for fuel cell-based autonomous vehicles to solve the problems mentioned in the background section.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an energy management system for fuel cell-based autonomous vehicles, comprising an environmental perception module, a decision-making and planning module, an energy management main module, an execution module, and a status monitoring module; wherein the energy management main module includes a dynamic energy demand prediction and allocation module, a fuel cell control submodule, an energy storage battery control submodule, and an energy recovery submodule; the environmental perception module, decision-making and planning module, energy management main module, execution module, and status monitoring module interact with each other via a hybrid communication network composed of CAN bus and Ethernet; the output of the environmental perception module is connected to the input of the decision-making and planning module and the input of the dynamic energy demand prediction and allocation module; the output of the decision-making and planning module is connected to the input of the dynamic energy demand prediction and allocation module; the output of the dynamic energy demand prediction and allocation module is connected to the inputs of the fuel cell control submodule, the energy storage battery control submodule, and the energy recovery submodule, respectively; the outputs of the fuel cell control submodule, the energy storage battery control submodule, and the energy recovery submodule are all connected to the input of the execution module; the input of the status monitoring module is connected to the fuel cell, the energy storage battery, the drive motor, and the output of the execution module, and the output of the status monitoring module is connected to the input of the dynamic energy demand prediction and allocation module; The environmental perception module consists of a lidar, millimeter-wave radar, camera, and GPS / IMU. It is used to collect information on obstacles around the vehicle, traffic light status, lane line information, real-time location, driving speed, and acceleration data to obtain environmental perception data. The environmental perception data is updated at a frequency of 10Hz. The decision-making and planning module generates vehicle driving decisions (such as acceleration, deceleration, steering, and stopping) and a driving trajectory plan for the next 5 seconds based on environmental perception data, and outputs a target speed curve. ( ); The energy management main module includes a fuel cell control submodule for adjusting the output power of the fuel cell, an energy storage battery control submodule for managing the charging and discharging state of the energy storage battery, and an energy recovery submodule for controlling the intensity of braking energy recovery. The dynamic energy demand prediction and allocation module is used to predict the future energy demand of the vehicle and allocate the output ratio of the fuel cell and the energy storage battery. The execution module includes a fuel cell system, an energy storage battery pack (using lithium-ion batteries), a drive motor, a braking system, and a DC / DC converter, and is used to execute control commands from the energy management main module. The condition monitoring module collects the fuel cell output voltage through voltage, current, and temperature sensors. Output current ,temperature Energy storage batteries Value, terminal voltage Current Output power of the drive motor and temperature The data acquisition frequency is 20Hz.

[0005] Preferably, the decision planning module is implemented using the following steps: The first step is data reception and fusion. In real time, 10Hz high-frequency data output by the environmental perception module is received, including obstacle distance information from lidar, relative velocity data from millimeter-wave radar, traffic light / lane line results identified by camera, and positioning and attitude data from GPS / IMU. Redundancy and noise are eliminated through multi-source data fusion algorithm to obtain fused data. The second step is driving decision generation. Based on the fused data, the driving scenario is determined by a rule-based scene matching algorithm. For example, if a stationary obstacle is detected within 50m ahead, the decision is to "slow down and avoid it". If a green light is detected and the lane lines are clear, the decision is to "accelerate and maintain the lane". When the light is red, the decision is to "drive at a constant speed to the stop line and then stop". The third step is trajectory and speed planning. Based on the decision results, a smooth driving trajectory for the next 5 seconds is generated (ensuring compliance with lane constraints). The target speed at each moment is then calculated based on the trajectory, forming a continuous target speed curve. ,in For example, the curve of "accelerating to 40km / h in 0-2s and maintaining a constant speed in 2-5s"; The fourth step is data output. The data is sent in real time to the dynamic energy demand forecasting and allocation module, providing the core basis for energy demand calculation.

[0006] Preferably, the specific implementation steps of the dynamic energy demand prediction and allocation module are as follows: Step 1: Multi-source data preprocessing and feature extraction: The dynamic energy demand forecasting and allocation module receives three types of input data: environmental perception data, decision planning data, and status monitoring data. The environmental perception data includes traffic light status. Road slope Road surface friction coefficient The decision-making and planning data is the target speed curve. Condition monitoring data includes fuel cell condition parameters. , , Energy storage battery , Motor power Then, the three types of input data are preprocessed, and feature parameters are extracted. ,in The filtered curves show the road gradient, friction coefficient, and target speed. Step 2: Short-term energy demand forecasting based on the improved instantaneous power method: Based on feature parameter set Calculate the instantaneous energy demand forecast for the vehicle within the next 5 seconds. , The formula is: ,in The total mass of the vehicle (unit: kg). Based on the target velocity curve Calculated instantaneous acceleration, Unit: m / s² The air drag coefficient is taken as 0.28. The vehicle's frontal area (unit: m²). The density is the air density (unit: kg / m³, taken as 1.225 at room temperature). The acceleration due to gravity (unit: m / s², taken as 9.8) To improve the efficiency of the drive motor (based on the motor temperature collected by the condition monitoring module) correction, hour ; hour ; hour ); At the same time, combined with the traffic light status right Make corrections: when When, i.e., red light, mandatory , , To predict the time to reach the stop line; when When the yellow light is on, calculate the power required to decelerate to a stop. ,make ,in , This is the distance from the current position to the stop line; Step 3: Real-time deviation correction based on state feedback: Introducing an energy demand forecasting bias correction factor Dynamically adjust based on real-time vehicle status data The corrected actual energy demand was obtained. : Among them, the correction factor The formula for calculation is: ,in The data sampling interval is 0.05s. The goal of energy storage batteries Value (taken as 0.6~0.8, dynamically adjusted based on mileage: when remaining mileage > 50km) When the remaining distance is ≤50km, S ), The optimal operating temperature for the fuel cell is 65°C. The correction factors are set to 0.5, 0.3, and 0.2 respectively, satisfying the following conditions: To ensure the stability of the correction factor; when At that time, take ;when At that time, take Avoid over-correction; Step 4: Execution of dynamic power allocation strategy: Based on the revised actual energy demand and energy storage batteries Status, allocate fuel cell output power With the output power of energy storage batteries ,satisfy (when At that time, the vehicle is driven; when At that time, it enters energy recovery mode. , This refers to charging the energy storage battery; the specific allocation rules are as follows: when When the energy storage battery is fully charged: ,in The minimum stable output power of the fuel cell is 15% of its rated power, with the remaining power provided by the energy storage battery. );like ,but , (at this time (energy storage battery charging). when At that time, the energy storage battery has a moderate charge: , This ensures that the fuel cell operates within its high-efficiency range, i.e., 30% to 70% of its rated power. when When the energy storage battery is low on power: ,in This represents the maximum output power of the fuel cell (100% of the rated power), with any remaining power supplemented by the energy storage battery; if ,but At the same time, it triggers an increase in energy recovery priority, that is, the energy recovery efficiency is increased by 10% during braking.

[0007] 1. Preferably, the preprocessing of the three types of input data in step 1 specifically includes: processing the traffic light status... Quantization Processing: Definition A green light means that you are allowed to drive. The yellow light indicates that you are about to slow down. A red light means you must stop; the Kalman filter algorithm is used to analyze the road slope. Road surface friction coefficient and target speed To smooth out noise interference, the filtering formula is as follows: Equations of state: ; Observation equation: ; in, for The state vector at time t, the state vector contains , It is the state transition matrix (3×3 identity matrix). For control matrix (3×1 zero matrix, no external control input). To control the quantity, The process noise is Gaussian noise with a mean of 0 and a variance of 0.01. for The vector of observations at time t, The observation matrix is ​​a 3×3 identity matrix. The observation noise is Gaussian noise with a mean of 0 and a variance of 0.02.

[0008] Compared with the prior art, the beneficial effects of the present invention are: by integrating multi-source environmental and vehicle status data, the present invention realizes the advance prediction and real-time correction of energy demand, and solves the problem of lagging energy distribution in traditional systems; The dynamic power allocation strategy of this invention is based on energy storage batteries. By understanding the state and operating characteristics of the fuel cell, frequent start-ups and shutdowns and operation in inefficient ranges are avoided, thus extending the fuel cell's lifespan; simultaneously, the depth of charge-discharge cycles of the energy storage battery is reduced. The fluctuation range is reduced to 0.6~0.8, improving the cycle life of energy storage batteries.

[0009] The overall system of this invention does not rely on a large amount of historical training data, has low computational complexity (single cycle calculation time ≤10ms), meets the real-time requirements of autonomous vehicles, and is compatible with different types of fuel cells and energy storage batteries, making it highly versatile. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the overall structure of the present invention; Figure 2 This is a schematic diagram of the main energy management module structure of the present invention. Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0013] Please see Figure 1-2This invention provides a technical solution: an energy management system for fuel cell-based autonomous vehicles, comprising an environmental perception module, a decision-making and planning module, an energy management main module, an execution module, and a status monitoring module; wherein the energy management main module includes a dynamic energy demand prediction and allocation module, a fuel cell control submodule, an energy storage battery control submodule, and an energy recovery submodule; the environmental perception module, decision-making and planning module, energy management main module, execution module, and status monitoring module interact with each other via a hybrid communication network composed of CAN bus and Ethernet; the output of the environmental perception module is connected to the input of the decision-making and planning module and the input of the dynamic energy demand prediction and allocation module; the output of the decision-making and planning module is connected to the input of the dynamic energy demand prediction and allocation module; the output of the dynamic energy demand prediction and allocation module is connected to the inputs of the fuel cell control submodule, the energy storage battery control submodule, and the energy recovery submodule; the outputs of the fuel cell control submodule, the energy storage battery control submodule, and the energy recovery submodule are all connected to the input of the execution module; the input of the status monitoring module is connected to the outputs of the fuel cell, the energy storage battery, the drive motor, and the execution module, and the output of the status monitoring module is connected to the input of the dynamic energy demand prediction and allocation module; The environmental perception module consists of a lidar, millimeter-wave radar, camera, and GPS / IMU. It is used to collect information on obstacles around the vehicle, traffic light status, lane line information, real-time location, driving speed, and acceleration data to obtain environmental perception data. The environmental perception data is updated at a frequency of 10Hz. The decision planning module generates vehicle driving decisions (such as acceleration, deceleration, steering, and stopping) and a driving trajectory plan for the next 5 seconds based on environmental perception data, and outputs a target speed curve. ( ); The energy management main module includes a fuel cell control submodule for adjusting the output power of the fuel cell, an energy storage battery control submodule for managing the charging and discharging state of the energy storage battery, and an energy recovery submodule for controlling the intensity of braking energy recovery. The dynamic energy demand prediction and allocation module is used to predict the future energy demand of the vehicle and allocate the output ratio of the fuel cell and the energy storage battery. The execution module includes a fuel cell system, an energy storage battery pack (using lithium-ion batteries), a drive motor, a braking system, and a DC / DC converter, and is used to execute control commands from the main energy management module. The condition monitoring module collects the fuel cell output voltage through voltage, current, and temperature sensors. Output current ,temperature Energy storage batteries Value, terminal voltage Current Output power of the drive motor and temperature The data acquisition frequency is 20Hz.

[0014] Furthermore, the specific implementation steps of the decision-making and planning module are as follows: The first step is data reception and fusion. In real time, 10Hz high-frequency data output by the environmental perception module is received, including obstacle distance information from lidar, relative velocity data from millimeter-wave radar, traffic light / lane line results identified by camera, and positioning and attitude data from GPS / IMU. Redundancy and noise are eliminated through multi-source data fusion algorithm to obtain fused data. The second step is driving decision generation. Based on the fused data, the driving scenario is determined by a rule-based scene matching algorithm. For example, if a stationary obstacle is detected within 50m ahead, the decision is to "slow down and avoid it". If a green light is detected and the lane lines are clear, the decision is to "accelerate and maintain the lane". When the light is red, the decision is to "drive at a constant speed to the stop line and then stop". It should be noted that rule-based scene matching algorithms are existing technology, and will be briefly explained here: The core logic of the rule-based scene matching algorithm is: with "feature quantization - scene matching - instruction output" as a closed loop, it does not rely on a large amount of training data and meets the real-time requirements of the system (matching time ≤ 2ms). Key implementation step: First, transform the fused data into standardized decision parameters (such as traffic light quantification values). (e.g., obstacle distance threshold of 50m), and then match typical scenarios such as following and obstacle avoidance through preset rules, and finally output clear driving instructions; System adaptability: Target velocity curve output by the algorithm It can directly provide core input parameters for the dynamic energy demand prediction and allocation module, ensuring the linkage between energy management and driving decisions.

[0015] The steps are as follows: 1. Scene feature quantification: transform the fused data into decision parameters, such as setting the obstacle safety distance threshold to 50m and the traffic light status according to... =1 (green light) / 0.5 (yellow light) / 0 (red light) quantization, lane line offset is set to ±0.5m as the safe range; 2. Typical scenario matching, combined with parameter matching for four core scenarios: following, obstacle avoidance, intersection passage, and lane keeping. For example, "obstacle within 50m in the same lane + relative speed of 0" matches "stationary obstacle avoidance scenario", "GPS positioning 100m+ from the intersection". =1+Clear lane lines match “intersection acceleration scenario”; 3. Decision command generation, output clear commands for matching scenarios, such as “decelerate to 30km / h + make slight right turn” for obstacle avoidance scenario, and “drive at a constant speed to the stop line + prepare to stop” for red light scenario.

[0016] The third step is trajectory and speed planning. Based on the decision results, a smooth driving trajectory for the next 5 seconds is generated (ensuring compliance with lane constraints). The target speed at each moment is then calculated based on the trajectory, forming a continuous target speed curve. ,in For example, the curve of "accelerating to 40km / h in 0-2s and maintaining a constant speed in 2-5s"; The fourth step is data output. The data is sent in real time to the dynamic energy demand forecasting and allocation module, providing the core basis for energy demand calculation.

[0017] Furthermore, the specific implementation steps of the dynamic energy demand prediction and allocation module are as follows: Step 1: Multi-source data preprocessing and feature extraction: The dynamic energy demand forecasting and allocation module receives three types of input data: environmental perception data, decision planning data, and status monitoring data. The environmental perception data includes traffic light status. Road slope Road surface friction coefficient The decision-making and planning data is the target speed curve. Condition monitoring data includes fuel cell condition parameters. , , Energy storage battery , Motor power Then, the three types of input data are preprocessed, and feature parameters are extracted. ,in The filtered curves show the road gradient, friction coefficient, and target speed. Step 2: Short-term energy demand forecasting based on the improved instantaneous power method: Based on feature parameter set Calculate the instantaneous energy demand forecast for the vehicle within the next 5 seconds. , The formula is: ,in The total mass of the vehicle (unit: kg). Based on the target velocity curve Calculated instantaneous acceleration, Unit: m / s² The air drag coefficient is taken as 0.28. The vehicle's frontal area (unit: m²). The density is the air density (unit: kg / m³, taken as 1.225 at room temperature). The acceleration due to gravity (unit: m / s², taken as 9.8) To improve the efficiency of the drive motor (based on the motor temperature collected by the condition monitoring module) correction, hour ; hour ; hour ); At the same time, combined with the traffic light status right Make corrections: when When, i.e., red light, mandatory , , To predict the time to reach the stop line; when When the yellow light is on, calculate the power required to decelerate to a stop. ,make ,in , This is the distance from the current position to the stop line; Step 3: Real-time deviation correction based on state feedback: Introducing an energy demand forecasting bias correction factor Dynamically adjust based on real-time vehicle status data The corrected actual energy demand was obtained. : Among them, the correction factor The formula for calculation is: ,in The data sampling interval is 0.05s. The goal of energy storage batteries Value (taken as 0.6~0.8, dynamically adjusted based on mileage: when remaining mileage > 50km) When the remaining distance is ≤50km, S ), The optimal operating temperature for the fuel cell is 65°C. The correction factors are set to 0.5, 0.3, and 0.2 respectively, satisfying the following conditions: To ensure the stability of the correction factor; when At that time, take ;when At that time, take Avoid over-correction; Step 4: Execution of dynamic power allocation strategy: Based on the revised actual energy demand and energy storage batteries Status, allocate fuel cell output power With the output power of energy storage batteries ,satisfy (when At that time, the vehicle is driven; when At that time, it enters energy recovery mode. , This refers to charging the energy storage battery; the specific allocation rules are as follows: when When the energy storage battery is fully charged: ,in The minimum stable output power of the fuel cell is 15% of its rated power, with the remaining power provided by the energy storage battery. );like ,but , (at this time (energy storage battery charging). when At that time, the energy storage battery has a moderate charge: , This ensures that the fuel cell operates within its high-efficiency range, i.e., 30% to 70% of its rated power. when When the energy storage battery is low on power: ,in This represents the maximum output power of the fuel cell (100% of the rated power), with any remaining power supplemented by the energy storage battery; if ,but This also triggers an increase in energy recovery priority, meaning that energy recovery efficiency is increased by 10% during braking. Furthermore, in step 1, the three types of input data undergo preprocessing, specifically the following processing steps: processing the traffic light status... Quantization Processing: Definition A green light means that you are allowed to drive. The yellow light indicates that you are about to slow down. A red light means you must stop; the Kalman filter algorithm is used to analyze the road slope. Road surface friction coefficient and target speed To smooth out noise interference, the filtering formula is as follows: Equations of state: ; Observation equation: ; in, for The state vector at time t, the state vector contains , It is the state transition matrix (3×3 identity matrix). For control matrix (3×1 zero matrix, no external control input). To control the quantity, The process noise is Gaussian noise with a mean of 0 and a variance of 0.01. for The vector of observations at time t, The observation matrix is ​​a 3×3 identity matrix. The observation noise is Gaussian noise with a mean of 0 and a variance of 0.02.

[0018] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An energy management system for a fuel cell based unmanned vehicle, characterized by, Comprise: The environmental perception module, decision planning module, energy management main module, execution module and state monitoring module; wherein the energy management main module contains dynamic energy demand prediction and distribution module, fuel cell control sub-module, energy storage battery control sub-module and energy recovery sub-module; The output end of the environmental perception module is connected with the input end of the decision planning module, the input end of the dynamic energy demand prediction and distribution module; The output end of the decision planning module is connected with the input end of the dynamic energy demand prediction and distribution module; The output end of the dynamic energy demand prediction and distribution module is connected with the input end of the fuel cell control sub-module, the energy storage battery control sub-module and the energy recovery sub-module respectively; The output end of the fuel cell control sub-module, the energy storage battery control sub-module and the energy recovery sub-module is connected with the input end of the execution module; The input end of the state monitoring module is connected with the output end of the fuel cell, the energy storage battery, the driving motor and the execution module, and the output end of the state monitoring module is connected with the input end of the dynamic energy demand prediction and distribution module; The environmental perception module is composed of laser radar, millimeter wave radar, camera and GPS / IMU, which is used to collect vehicle surrounding obstacle information, traffic light state, lane line information, real-time position, driving speed and acceleration data, that is, to obtain environmental perception data, and the environmental perception data update frequency is 10Hz; The decision planning module generates vehicle driving decisions and a driving trajectory plan for the next 5s based on the environment perception data, and outputs a target speed curve , ; The fuel cell control sub-module in the energy management main module is used to adjust the output power of the fuel cell, the energy storage battery control sub-module is used to manage the charge and discharge state of the energy storage battery, and the energy recovery sub-module is used to control the intensity of brake energy recovery; The dynamic energy demand prediction and distribution module is used to predict the future energy demand of the vehicle and distribute the output proportion of the fuel cell and the energy storage battery; The execution module comprises fuel cell system, energy storage battery group, driving motor, brake system and DC / DC converter, which is used to execute the control instruction of the energy management main module; The state monitoring module collects the output voltage of the fuel cell, the output current, the temperature, the value of the energy storage battery, the terminal voltage, the current, the output power of the driving motor and the temperature through voltage, current and temperature sensors at a frequency of 20 Hz.​​​​​​​ 2. An energy management system for a fuel cell based unmanned vehicle as claimed in claim 1, wherein: The decision planning module specifically implements the following steps: First step, data receiving and fusion, real-time receiving 10Hz high frequency data output by the environmental perception module, including obstacle distance information of laser radar, relative speed data of millimeter wave radar, traffic light / lane line result recognized by camera and positioning and attitude data of GPS / IMU, eliminating redundancy and noise through multi-source data fusion algorithm to obtain fusion data; Second step, driving decision generation, based on fusion data, the driving scene is judged through rule type scene matching algorithm; Third step, trajectory and speed planning, combined with the decision result to generate a smooth driving trajectory in the future 5s, and calculate the target speed at each time according to the trajectory to form a continuous target speed curve wherein ; Fourth step, data output, will Real-time transmission to dynamic energy demand forecasting and distribution module, to provide the core basis for energy demand calculation.

3. The energy management system for a fuel cell based unmanned vehicle of claim 1, wherein: The dynamic energy demand prediction and distribution module specifically implements the following steps: Step 1, multi-source data preprocessing and feature extraction: The dynamic energy demand prediction and distribution module receives three types of input data: environment perception data, decision planning data, and state monitoring data, wherein the environment perception data includes traffic light status , road slope , road friction coefficient , the decision planning data is a target speed curve , and the state monitoring data includes fuel cell state parameters , , , energy storage battery parameters , , and motor power ; then the three types of input data are preprocessed, and feature parameters are extracted , wherein are filtered road slope, friction coefficient, and target speed curve; Step 2, short-term energy demand prediction based on improved instantaneous power method: Based on a set of characteristic parameters , the instantaneous energy demand prediction value of the vehicle in the next 5s is calculated , , the formula is: , wherein is the total mass of the vehicle, is the instantaneous acceleration calculated based on the target speed curve , , is the air resistance coefficient (0.28), is the vehicle frontal area, is the air density, is the gravitational acceleration, is the efficiency of the drive motor (corrected according to the motor temperature collected by the state monitoring module , , ; , ; , ; At the same time, combined with the traffic light status right Make corrections: when When, i.e., red light, mandatory , , To predict the time to reach the stop line; when When the yellow light is on, calculate the power required to decelerate to a stop. ,make ,in , This is the distance from the current position to the stop line; Step 3, real-time deviation correction based on state feedback: Introducing energy demand prediction deviation correction factor , dynamically adjusted by real-time vehicle state data , to obtain the corrected actual energy demand : , wherein the calculation formula of the correction factor is: , wherein is the data sampling interval, is the target value of the energy storage battery, is the optimal working temperature of the fuel cell; is the correction coefficient, respectively 0.5, 0.3, 0.2, satisfying , to ensure the stability of the correction factor; When , take ; when , take , to avoid overcorrection. Step 4, dynamic power distribution strategy execution: According to the revised actual energy demand and the energy storage battery State, the fuel cell output power is allocated and the energy storage battery output power , meet When (when vehicle driving; when energy recovery mode, , , namely the energy storage battery charging; the specific allocation rules are as follows: When the energy storage battery has sufficient power: where is the minimum stable output power of the fuel cell, i.e. 15% of the rated power, and the remaining power is provided by the energy storage battery, i.e. ); if then , ; When the energy storage battery is moderately charged: , ensure that the fuel cell works in the high efficiency range, i.e. 30%~70% of the rated power. When the energy storage battery is insufficient: where is the maximum output power of the fuel cell, and the remaining power is supplemented by the energy storage battery; if then the energy recovery priority is triggered to be raised, i.e. the energy recovery efficiency is increased by 10% when braking.

4. The energy management system for a fuel cell based unmanned vehicle of claim 1, wherein: The specific processing content of the pre-processing of the three types of data in step 1 is: the traffic light state Quantitative processing: definition , that is, the green light allows driving; , that is, the yellow light is ready to slow down; , that is, the red light needs to stop; through Kalman filtering algorithm, the road slope , the road friction coefficient and the target speed are smoothed to eliminate noise interference, and the filtering formula is as follows: Equation of state: ; Observation equation: ; wherein is the state vector at time k, the state vector containing , is the state transition matrix, is the control matrix, is the control quantity, is the process noise; is the observation vector at time k, is the observation matrix, is the observation noise.

Images