Hierarchical collaborative energy-saving control method for fuel cell heavy-duty trucks in dynamic traffic environments

Abstract

The present invention is applicable to the fields of vehicle speed planning technology and energy management technology, and provides a hierarchical collaborative energy-saving control method for fuel cell heavy trucks in dynamic traffic environments. The method adopts a hierarchical predictive collaborative optimization control framework to achieve collaborative control of external environmental disturbances and power system characteristics; the upper layer adopts an LSTM neural network to predict the speed of the preceding vehicle, establishes a multi-objective optimization function, and constructs a speed planning strategy based on the Pontryagin minimum principle algorithm under the model predictive control framework; the lower layer designs a dual-time domain adaptive equivalent consumption minimum strategy, combines the predicted operating condition data in the rolling time domain, dynamically adjusts the equivalent factor, and generates the optimal control sequence allocation of the fuel cell power; the present invention takes into account external traffic disturbances and power system characteristics, improves the energy utilization efficiency of fuel cell heavy trucks in dynamic traffic environments, and also achieves efficient collaboration between vehicle longitudinal motion control and energy management.

Description

Technical Field

[0001] The present invention relates to the technical fields of vehicle speed planning and energy management, and in particular to a stratified collaborative energy-saving control method for fuel cell heavy trucks in a dynamic traffic environment. Background Art

[0002] Fuel cell heavy-duty trucks, combining a proton exchange membrane fuel cell (PEM) fuel cell system with a power battery pack, are a key solution for meeting the high power and energy demands of heavy-duty trucks thanks to their high energy density, rapid hydrogen refueling, and zero emissions. An efficient energy management strategy (EMS) optimizes power distribution between the fuel cell and battery, improving system efficiency. Speed ​​planning, by generating an optimal speed profile, reduces energy consumption while ensuring safety. The coordinated optimization of these two strategies not only alleviates fuel cell lifespan and cost pressures but also enhances vehicle dynamic performance and overall economic efficiency, becoming a key research direction in promoting the commercialization of fuel cell heavy-duty trucks.

[0003] Disadvantages of existing technology:

[0004] 1. Vehicle speed planning and energy management systems are mutually coupled. The upper-level speed planning generates an energy-saving driving curve, while the lower-level energy management system optimizes energy distribution based on the speed curve. This collaborative optimization of vehicle kinematics and powertrain characteristics improves system energy efficiency. However, existing research on speed planning and energy management has primarily focused on traditional fuel-powered vehicles and pure electric vehicles, while research on multi-objective collaborative optimization for fuel cell heavy-duty trucks remains relatively limited.

[0005] 2. Current research is largely based on modeling idealized, non-interference traffic scenarios, lacking effective responses to the complex operating conditions caused by real-time interactions with vehicles ahead. Existing global optimization methods suffer from high computational complexity and difficulty meeting real-time requirements in real-time control. Furthermore, the equivalent factor is significantly affected by driving conditions, resulting in insufficient dynamic adjustment capabilities. A control strategy that balances external traffic disturbances with the characteristics of the powertrain is urgently needed to improve the energy efficiency of fuel cell heavy-duty trucks in dynamic traffic environments. Summary of the Invention

[0006] The present invention provides a stratified collaborative energy-saving control method for fuel cell heavy trucks in a dynamic traffic environment, which achieves approximately optimal power distribution between fuel cells and power batteries.

[0007] A hierarchical collaborative energy-saving control method for fuel cell heavy-duty trucks in dynamic traffic environments is proposed. The method adopts a hierarchical predictive collaborative optimization control framework to achieve collaborative control of external environmental disturbances and power system characteristics. The hierarchical predictive collaborative optimization control framework includes an upper layer and a lower layer. The upper layer uses an LSTM neural network to predict the speed of the preceding vehicle, establishes a multi-objective optimization function, and constructs a speed planning strategy based on the Pontryagin Minimum Principle (PMP) algorithm within the model predictive control (MPC) framework to achieve safe, energy-saving and comfortable driving of the main vehicle. The lower layer designs a dual-time domain adaptive equivalent consumption minimization strategy (DTD-AECMS) that combines predicted operating condition data in the rolling time domain, dynamically adjusts the equivalent factor, and generates the optimal control sequence allocation for the fuel cell power.

[0008] As a further technical solution of the present invention, the steps of using the LSTM neural network to predict the speed of the preceding vehicle include:

[0009] The long short-term memory network is used to build the preceding vehicle speed prediction framework, and the preceding vehicle historical speed sequence and historical acceleration series As input features, For the current moment, = , Index variable for historical moments, representing the time from the current moment The time step of the backtracking, its value range , is the total length of historical moments, Indicates the speed of the vehicle ahead. Represents the acceleration of the preceding vehicle; through a nonlinear mapping function Generate the preceding vehicle speed sequence in the prediction time domain , , To predict the time domain index variable, representing the The time step of the prediction, its value range , is the prediction time domain step size.

[0010] As a further technical solution of the present invention, the steps of establishing a multi-objective optimization function and constructing a speed planning strategy based on the Pontryagin Minimum Principle (PMP) algorithm under the model predictive control framework to achieve safe, energy-saving and comfortable driving of the main vehicle include:

[0011] Through V2V technology, the speed information and prediction data of the preceding vehicle are obtained in real time, and a multi-objective optimization control model is constructed. The control goal is to minimize the DC bus power demand of the main vehicle. and control smoothness, which refers to the acceleration fluctuation , the state variable is the main vehicle speed , the control variable is the main vehicle acceleration , the optimal control problem of speed planning is expressed as:

[0012] ;

[0013] ;

[0014] ;

[0015] Where, is the weight coefficient, which is used to balance energy consumption and acceleration fluctuation; is the vehicle driving power, The mass of the main vehicle; is the friction coefficient; is the acceleration due to gravity; is the slope angle; is the drag coefficient; is the windward area; is the air density; is the rotation mass conversion factor; 、 、 Represent the efficiency of DC / AC converter, drive motor and main reduction respectively; is the main vehicle speed; is the acceleration of the main vehicle;

[0016] The distance between the preceding vehicle and the main vehicle Can be obtained by Calculated, to avoid collision risk, set a minimum safety distance To ensure driving safety under the following conditions; at the same time, to avoid the reduction of road traffic efficiency due to excessive vehicle distance, the maximum allowable distance is restricted. ;Introduce dynamic upper and lower limit spacing constraints to improve safety and efficiency;

[0017] The minimum safe distance is:

[0018] ;

[0019] Where, A fixed minimum distance threshold is set for car-following control to provide basic safety in the most extreme or unpredictable scenarios, preventing the risk of excessive distance or even collision caused by dynamic distance calculation failure or sudden environmental changes. Driver reaction time;

[0020] The maximum safe distance is:

[0021] ;

[0022] Where, Set a fixed maximum distance threshold for vehicle following control. Set to 2 seconds;

[0023] The state equation and constraints are:

[0024] ;

[0025] Where, Characterize the real-time speed of the vehicle ahead. To suppress drastic speed changes and improve driving comfort, the acceleration feasible region is constrained by upper and lower bounds, i.e. ; The tolerance of the speed difference between the two vehicles at the end of the rolling time domain; is the initial distance to the vehicle in front; is the initial speed of the main vehicle; is the time step between adjacent moments.

[0026] Adopting the rolling optimization strategy, based on the updated speed prediction information of the preceding vehicle in each sampling time domain, the speed planning solution is performed within the prediction time domain through the PMP algorithm until the end of the driving cycle; the optimal control sequence is generated. , but only the first item in the sequence is controlled Applied to the controlled object model to achieve real-time closed-loop control;

[0027] The Hamiltonian function for speed planning of fuel cell heavy trucks within the prediction time domain in the present invention is defined as:

[0028] ;

[0029] Where, is the covariate variable of the Hamiltonian function;

[0030] The necessary conditions for the vehicle system to achieve optimal power distribution and optimal speed at the same time are:

[0031]

[0032] Where, is the state equation, is the co-state equation, is the control equation.

[0033] As a further technical solution of the present invention, a dual-time domain adaptive equivalent consumption minimization strategy (DTD-AECMS) is designed to dynamically adjust the equivalent factor based on the predicted operating condition data in the rolling time domain to generate the optimal control sequence allocation for the fuel cell power. The steps include:

[0034] The reference value of the power battery SOC is set to be consistent with the starting value, and the power battery can be charged by the fuel cell during the entire driving process. Therefore, minimizing the hydrogen consumption of the fuel cell is used as the objective function. At the same time, considering the physical characteristics of the power system, the state variable of the optimization target is the power battery SOC, and the control variable is the fuel cell power. ;

[0035] ;

[0036] ;

[0037] ;

[0038] ;

[0039] ;

[0040] Where, is the fuel cell hydrogen consumption rate, and is related to the fuel cell output power There is a functional relationship; 、 They are The upper and lower limits allowed; 、 are the upper and lower limits of fuel cell power; Output power to the power battery; 、 are the upper and lower limits of the power of the power battery, and The upper and lower limits of the fuel cell power fluctuation value are: = - ; The initial Value; open circuit voltage of power battery and power battery internal resistance With battery related; is the battery current, is the nominal capacity of the battery, is the DC / DC converter efficiency; is the time step between adjacent moments.

[0041] The ECMS algorithm is an engineering implementation of the PMP theory in real-time control scenarios. It achieves the approximation of the global optimal solution through local optimization through a strict mapping relationship between equivalent factors and PMP co-state variables. For the energy management problem of fuel cell heavy trucks, the Hamiltonian function is defined as:

[0042] ;

[0043] Where: is a co-state variable, and the state variable is the power battery , the control variable is the fuel cell output power ;

[0044] As a real-time optimization method based on the PMP framework, ECMS maps electricity consumption to equivalent hydrogen consumption by introducing an equivalent factor to build a unified energy optimal control framework. The calculation of the total hydrogen consumption rate is expressed as:

[0045]

[0046] Where: is the equivalent hydrogen consumption rate, is the power battery energy consumption rate, is the equivalence factor;

[0047] The power consumption rate of the power battery The relationship is expressed as:

[0048] ;

[0049] Where, The low calorific value of hydrogen;

[0050] Derived and The equivalence between the two;

[0051] ;

[0052] The PMP method has thus far yielded a global optimal solution for various driving conditions. This optimal solution can serve as a training dataset for the equivalent factor adjustment model. Given the significant time-varying characteristics of the equivalent factor due to the randomness and unpredictability of driving conditions, it is necessary to establish a dynamic, nonlinear relationship between the equivalent factor and the vehicle's operating state, SOC, and future speed trends. Neural network algorithms, with their strong nonlinear mapping and model-free optimization properties, can avoid the reliance of traditional regression methods on explicit models. By dynamically tuning parameters to approximate the optimal solution space for complex systems, they are suitable for equivalent factor prediction.

[0053] A neural network training architecture based on the offline optimization database of PMP principle is constructed to achieve online dynamic prediction of equivalent factors. The model input layer adopts multi-dimensional feature vector design: , required power, main vehicle speed, acceleration and main vehicle planned speed in the rolling time domain; the output layer is the optimal co-state variable at the current moment , and then through and The equivalence relationship between the two is converted into the corresponding equivalent factors ;

[0054] Optimal Equivalence Factor It is sensitive to working conditions and needs to make corresponding adaptive adjustments according to unknown working conditions. Feedback, where the speed ratio in the future section is used as a feedforward adjustment of the equivalent factor, Feedback acts as a post-compensation mechanism, combining the effects of the two on the equivalent factor Perform adaptive correction, the formula is shown as follows:

[0055] ;

[0056] ;

[0057] ;

[0058] ;

[0059] Where, is the corrected equivalence factor, is the adjustment coefficient, is the proportionality coefficient; For the prediction time domain The standard deviation of vehicle speed within is the average vehicle speed in the prediction time domain.

[0060] Explanation of terms:

[0061] Speed ​​planning: During driving, the vehicle performs adaptive following based on the current position, speed, and acceleration, as well as the expected speed and acceleration of the target state.

[0062] Energy management: Through reasonable strategy settings, the vehicle's energy sources and energy consumption units are managed to achieve a balance between the safety, economy and power of the entire vehicle.

[0063] Beneficial effects achieved by the present invention:

[0064] 1. This paper focuses on the energy-saving driving and energy management issues of fuel cell heavy-duty trucks in intelligent connected traffic environments, and proposes a hierarchical predictive collaborative optimization control framework that takes into account external traffic disturbances and power system characteristics, thereby improving the energy utilization efficiency of fuel cell heavy-duty trucks in dynamic traffic environments.

[0065] 2. Through upper-layer speed planning based on LSTM-based preceding vehicle speed prediction and the Pontryagin Minimum Principle (PMP) algorithm within the Model Predictive Control (MPC) framework, dynamic traffic influences are integrated into speed planning, providing forward-looking information for speed planning and ensuring safe, energy-efficient, and comfortable driving for the main vehicle.

[0066] 3. The lower layer adopts the adaptive equivalent consumption minimization strategy (DTD-AECMS) that integrates dual-time domain information, integrating information from the current and future time domains, and using a neural network algorithm to quickly and accurately obtain the equivalent factor value. It then dynamically adjusts the equivalent factor based on SOC feedback in combination with future operating condition data to obtain the optimal control sequence allocation, achieving efficient coordination between vehicle longitudinal motion control and energy management. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is the structural diagram of the fuel cell heavy-duty truck powertrain.

[0068] Figure 2 Schematic diagram of the structure of the hierarchical prediction collaborative optimization control framework in an embodiment of the present invention.

[0069] Figure 3 Graph showing fuel cell system efficiency and hydrogen consumption rate in an embodiment of the present invention.

[0070] Figure 4 Graph showing the relationship between the open circuit voltage, internal resistance, and SOC of a power battery in an embodiment of the present invention. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0072] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0073] This paper proposes a hierarchical collaborative energy-saving control method for fuel cell heavy-duty trucks in dynamic traffic environments. Through dynamic interaction between the upper and lower layers, it achieves real-time collaborative optimization of speed planning and energy management, thereby reducing energy consumption. First, the upper layer dynamically generates an energy-saving speed trajectory for the main vehicle that meets safety, economy, and comfort requirements by predicting the speed of the preceding vehicle based on an LSTM network and combining it with the Pontryagin Minimum Principle (PMP) algorithm under the Model Predictive Control (MPC) framework. Second, the lower layer proposes a dual-time domain adaptive equivalent consumption minimization strategy (DTD-AECMS), which uses a neural network algorithm to quickly and accurately estimate the equivalent factor value. Then, based on the future planned vehicle speed and real-time SOC feedback, it dynamically adjusts the equivalent factor to obtain the optimal control sequence allocation for power, achieving near-optimal power distribution between the fuel cell and the power battery.

[0074] Figure 1 The diagram shows the powertrain structure of a fuel cell heavy-duty truck, primarily consisting of a fuel cell, power battery, drive motor, differential, and DC / DC converter. The fuel cell system is connected to the DC bus via a DC / DC converter, while the drive motor achieves bidirectional conversion of mechanical energy into electrical energy via a bidirectional DC / AC converter.

[0075] See also Figure 2 An embodiment of the present invention provides a hierarchical collaborative energy-saving control method for fuel cell heavy trucks in dynamic traffic environments. The method adopts a hierarchical predictive collaborative optimization control framework to achieve collaborative control of external environmental disturbances and power system characteristics. The hierarchical predictive collaborative optimization control framework includes an upper layer and a lower layer. The upper layer adopts an LSTM neural network to predict the speed of the preceding vehicle, establishes a multi-objective optimization function, and constructs a speed planning strategy based on the Pontryagin Minimum Principle (PMP) algorithm under the model predictive control (MPC) framework to achieve safe, energy-saving and comfortable driving of the main vehicle. The lower layer designs a dual-time domain adaptive equivalent consumption minimum strategy (DTD-AECMS), combines the predicted operating condition data in the rolling time domain, dynamically adjusts the equivalent factor, and generates the optimal control sequence allocation of the fuel cell power.

[0076] In this embodiment, the steps of using the LSTM neural network to predict the speed of the preceding vehicle include:

[0077] In a speed planning system with interference from a preceding vehicle, the preceding vehicle's driving behavior is dynamically coupled with the host vehicle's driving state. Effectively predicting its future speed information is a prerequisite for optimizing the host vehicle's economical driving strategy. Proactively acquiring the preceding vehicle's speed time series characteristics to provide closed-loop guidance for the ego vehicle's driving decisions can significantly reduce vehicle energy consumption and enhance the optimization potential of the energy management system.

[0078] In view of this, the present invention adopts long short-term memory network to construct the preceding vehicle speed prediction framework, and the preceding vehicle historical speed sequence and historical acceleration series As input features, For the current moment, = , Index variable for historical moments, representing the time from the current moment The time step of the backtracking, its value range , is the total length of historical moments, Indicates the speed of the vehicle ahead. Represents the acceleration of the preceding vehicle; through a nonlinear mapping function Generate the preceding vehicle speed sequence in the prediction time domain , , To predict the time domain index variable, representing the The time step of the prediction, its value range , is the prediction time domain step size.

[0079] In this embodiment, the steps of establishing a multi-objective optimization function and constructing a speed planning strategy based on the Pontryagin Minimum Principle (PMP) algorithm under the model predictive control (MPC) framework to achieve safe, energy-saving and comfortable driving of the main vehicle include:

[0080] The present invention uses V2V technology to obtain the speed information and prediction data of the preceding vehicle in real time, and constructs a multi-objective optimization control model. The control goal is to minimize the DC bus power demand of the main vehicle. and control smoothness, which refers to the acceleration fluctuation , the state variable is the main vehicle speed , the control variable is the main vehicle acceleration , the optimal control problem of speed planning is expressed as:

[0081] ;

[0082] ;

[0083] ;

[0084] Where, is the weight coefficient, which is used to balance energy consumption and acceleration fluctuation; is the vehicle driving power, The mass of the main vehicle; is the friction coefficient; is the acceleration due to gravity; is the slope angle; is the drag coefficient; is the windward area; is the air density; is the rotation mass conversion factor; 、 、 Represent the efficiency of DC / AC converter, drive motor and main reduction respectively; is the main vehicle speed; is the acceleration of the main vehicle;

[0085] The distance between the preceding vehicle and the main vehicle Can be obtained by Calculated, to avoid collision risk, set a minimum safety distance To ensure driving safety under the following conditions; at the same time, to avoid the reduction of road traffic efficiency due to excessive vehicle distance, the maximum allowable distance is restricted. However, a fixed minimum safe distance may lead to collision risks, and a fixed maximum distance may lead to overly conservative following, reducing road traffic efficiency. Therefore, the present invention introduces dynamic upper and lower limit distance constraints to improve safety and efficiency.

[0086] The minimum safe distance is:

[0087] ;

[0088] Where, A fixed minimum distance threshold is set for car-following control to provide basic safety in the most extreme or unpredictable scenarios, preventing the risk of excessive distance or even collision caused by dynamic distance calculation failure or sudden environmental changes. Driver reaction time;

[0089] The maximum safe distance is:

[0090] ;

[0091] Where, Set a fixed maximum distance threshold for vehicle following control. Set to 2 seconds;

[0092] The state equation and constraints are:

[0093] ;

[0094] Where, Characterize the real-time speed of the vehicle ahead. To suppress drastic speed changes and improve driving comfort, the acceleration feasible region is constrained by upper and lower bounds, i.e. ; The tolerance of the speed difference between the two vehicles at the end of the rolling time domain; is the initial distance to the vehicle ahead; is the initial speed of the main vehicle; is the time step between adjacent moments.

[0095] Obviously, the speed planning problem has nonlinear characteristics. PMP, as a global optimization method, can transform the global optimization problem into an instantaneous problem. Its core idea is to find the optimal control variable to minimize the Hamiltonian function. The present invention adopts a rolling optimization strategy. Based on the updated speed prediction information of the preceding vehicle in each sampling time domain, the speed planning solution is performed within the prediction time domain by the PMP algorithm until the end of the driving cycle. Specifically, the algorithm will generate the optimal control sequence , but only the first item in the sequence is controlled Applied to the controlled object model to achieve real-time closed-loop control;

[0096] The Hamiltonian function for vehicle speed planning within the prediction time domain of a fuel cell heavy truck in the present invention is defined as:

[0097] ;

[0098] Where, is the covariate variable of the Hamiltonian function;

[0099] The necessary conditions for the vehicle system to achieve optimal power distribution and optimal speed at the same time are:

[0100]

[0101] Where, is the state equation, is the co-state equation, is the control equation.

[0102] In this embodiment, a dual-time domain adaptive equivalent consumption minimization strategy (DTD-AECMS) is designed to dynamically adjust the equivalent factor based on the predicted operating condition data in the rolling time domain to generate the optimal control sequence allocation for the fuel cell power. The steps include:

[0103] After solving the upper-level optimization problem to obtain the speed trajectory, the system power demand is calculated using the driver model. To address the underlying energy allocation problem, an adaptive equivalent energy consumption minimization strategy (DTD-AECMS) is proposed that integrates current operating conditions and future dual-time domain information to determine the equivalent factor. This method uses PMP to solve the optimal equivalent factor under different operating conditions and utilizes a neural network (NN) algorithm to quickly and accurately obtain the equivalent factor value. A dynamic mapping relationship is constructed between the equivalent factor and the state of charge (SOC) state and future operating conditions to achieve adaptive correction of the equivalent factor. Finally, with minimizing fuel cell hydrogen consumption as the objective function, the DTD-AECMS algorithm is used to obtain the optimal control sequence allocation for the hybrid system power, thus achieving energy management for fuel cell vehicles.

[0104] Energy management control problem description.

[0105] The research on energy management control strategy of fuel cell heavy truck belongs to the problem of multi-energy source optimization. The core goal of this system is to minimize energy consumption while meeting the driving requirements of the vehicle by dynamically coordinating the power output ratio of the fuel cell and the power battery. The present invention sets the reference value of the power battery SOC to be consistent with the starting value, and the power battery can be charged by the fuel cell during the entire driving process. Therefore, the minimization of the fuel cell hydrogen consumption is taken as the objective function. At the same time, considering the physical characteristics of the power system, the state variable of the optimization target is the power battery SOC, and the control variable is the fuel cell power. ;

[0106] ;

[0107] ;

[0108] ;

[0109] ;

[0110] ;

[0111] Where, is the fuel cell hydrogen consumption rate, and is related to the fuel cell output power There is a functional relationship. Based on the actual vehicle data collected in reality, the present invention uses a fuel cell with a rated power of 112KW. The relationship between the output power, hydrogen consumption and working efficiency of the battery system is as follows: Figure 3 As shown; 、 They are The upper and lower limits allowed; 、 are the upper and lower limits of fuel cell power; Output power to the power battery; 、 are the upper and lower limits of the power of the power battery, and The upper and lower limits of the fuel cell power fluctuation value are: = - ; The initial Value; open circuit voltage of power battery and power battery internal resistance With battery According to the actual vehicle battery data collection of fuel cell heavy trucks, the relationship curve between the open circuit voltage and internal resistance of the power battery and SOC is as follows: Figure 4As shown; is the battery current, is the nominal capacity of the battery, is the DC / DC converter efficiency; is the time step between adjacent moments.

[0112] The optimal equivalent factor database was established.

[0113] The ECMS algorithm is an engineering implementation of the PMP theory in real-time control scenarios. It achieves the approximation of the global optimal solution through local optimization through a strict mapping relationship between equivalent factors and PMP co-state variables. For the energy management problem of fuel cell heavy trucks in this invention, the Hamiltonian function is defined as:

[0114] ;

[0115] Where: is a co-state variable, and the state variable is the power battery , the control variable is the fuel cell output power ;

[0116] As a real-time optimization method based on the PMP framework, ECMS maps electricity consumption to equivalent hydrogen consumption by introducing an equivalent factor to build a unified energy optimal control framework. The calculation of the total hydrogen consumption rate is expressed as:

[0117]

[0118] Where: is the equivalent hydrogen consumption rate, is the power battery energy consumption rate, is the equivalence factor;

[0119] The power consumption rate of the power battery The relationship is expressed as:

[0120] ;

[0121] Where, The low calorific value of hydrogen;

[0122] Combining the above formulas, we can deduce and The equivalence between the two;

[0123] ;

[0124] The PMP method has thus far yielded a global optimal solution for various driving conditions. This optimal solution can serve as a training dataset for the equivalent factor adjustment model. Given the significant time-varying characteristics of the equivalent factor due to the randomness and unpredictability of driving conditions, it is necessary to establish a dynamic, nonlinear relationship between the equivalent factor and the vehicle's operating state, SOC, and future speed trends. Neural network algorithms, with their strong nonlinear mapping and model-free optimization properties, can avoid the reliance of traditional regression methods on explicit models. By dynamically tuning parameters to approximate the optimal solution space for complex systems, they are suitable for equivalent factor prediction.

[0125] Based on the above characteristics, the present invention constructs a neural network training architecture based on the PMP principle offline optimization database, aiming to achieve online dynamic prediction of equivalent factors. The model input layer adopts multi-dimensional feature vector design: the current , required power, main vehicle speed, acceleration and main vehicle planned speed in the rolling time domain; the output layer is the optimal co-state variable at the current moment , and then through and The equivalence relationship between the two is converted into the corresponding equivalent factors ;

[0126] Equivalence Factor Adaptive correction.

[0127] Optimal Equivalence Factor It is sensitive to working conditions and needs to make corresponding adaptive adjustments according to unknown working conditions. This paper is based on the future working speed and Feedback, where the speed ratio in the future section is used as a feedforward adjustment of the equivalent factor, Feedback acts as a post-compensation mechanism, combining the effects of the two on the equivalent factor Perform adaptive correction, the formula is shown as follows:

[0128] ;

[0129] ;

[0130] ;

[0131] ;

[0132] Where, is the corrected equivalence factor, is the adjustment coefficient, which reflects the changing trend of future working conditions to a certain extent. is the proportionality coefficient; For the prediction time domain The standard deviation of vehicle speed within is the average vehicle speed in the prediction time domain.

[0133] The dynamic adjustment mechanism of the equivalent factor of the present invention is based on the prediction time domain Real-time adjustment to adapt to changing working conditions. When it is at a high level, it indicates that there is a significant acceleration condition in the prediction time domain. By increasing the equivalent factor to improve the electric energy weight distribution, the charge state balance of the power battery is maintained; when When the value is low, it indicates that the vehicle will experience a deceleration process during the forecast period. By reducing the equivalent factor, the coordinated control of braking energy recovery and driving power distribution is optimized to optimize the overall energy utilization efficiency. It should be noted that the " At a higher level" or A "lower value" does not refer to a universal, fixed absolute value. It is a value that requires specialized engineering calibration based on specific target operating conditions and energy management control objectives.

[0134] It should be noted that, in this document, the term "comprises" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or apparatus that includes a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0135] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A hierarchical collaborative energy-saving control method for fuel cell heavy trucks in dynamic traffic environments, characterized by: The method uses a hierarchical predictive collaborative optimization control framework to achieve coordinated control of external environmental disturbances and power system characteristics. The hierarchical predictive collaborative optimization control framework includes an upper layer and a lower layer. The upper layer uses an LSTM neural network to predict the speed of the preceding vehicle, establishes a multi-objective optimization function, and constructs a speed planning strategy based on the Pontryagin minimum principle algorithm within the model predictive control framework to achieve safe, energy-saving and comfortable driving of the main vehicle. The lower layer designs a dual-time domain adaptive equivalent consumption minimization strategy that combines predicted operating condition data in the rolling time domain to dynamically adjust the equivalent factor and generate an optimal control sequence allocation for the fuel cell power. The steps of establishing a multi-objective optimization function, constructing a speed planning strategy based on the Pontryagin minimum principle algorithm under the model predictive control framework, and achieving safe, energy-saving and comfortable driving of the main vehicle include: Through V2V technology, the speed information and prediction data of the preceding vehicle are obtained in real time, and a multi-objective optimization control model is constructed. The control goal is to minimize the DC bus power demand of the main vehicle. and control smoothness, which refers to the acceleration fluctuation , the state variable is the main vehicle speed , the control variable is the main vehicle acceleration , the optimal control problem of speed planning is expressed as: ； ； ； Where, is the weight coefficient, which is used to balance energy consumption and acceleration fluctuation; is the vehicle driving power, The mass of the main vehicle; is the friction coefficient; is the acceleration due to gravity; is the slope angle; is the drag coefficient; is the windward area; is the air density; is the rotation mass conversion factor; 、 、 Represent the efficiency of DC / AC converter, drive motor and main reduction respectively; is the main vehicle speed; is the main vehicle acceleration, is the prediction time domain step length; The distance between the preceding vehicle and the main vehicle Can be obtained by Calculated, to avoid collision risk, set a minimum safety distance To ensure driving safety under the following conditions; at the same time, to avoid the reduction of road traffic efficiency due to excessive vehicle distance, the maximum allowable distance is restricted. ;Introduce dynamic upper and lower limit spacing constraints to improve safety and efficiency; The minimum safe distance is: ； Where, A fixed minimum distance threshold is set for car-following control to provide basic safety in the most extreme or unpredictable scenarios, preventing the risk of excessive distance or even collision caused by dynamic distance calculation failure or sudden environmental changes. Driver reaction time; The maximum safe distance is: ； Where, Set a fixed maximum distance threshold for vehicle following control. Set to 2 seconds; The state equation and constraints are: ； Where, Characterize the real-time speed of the vehicle ahead. To suppress drastic speed changes and improve driving comfort, the acceleration feasible region is constrained by upper and lower bounds, i.e. ; The tolerance of the speed difference between the two vehicles at the end of the rolling time domain; is the initial distance to the vehicle ahead; is the initial speed of the main vehicle; is the time step between adjacent moments; Adopting the rolling optimization strategy, based on the updated speed prediction information of the preceding vehicle in each sampling time domain, the speed planning solution is performed within the prediction time domain through the PMP algorithm until the end of the driving cycle; the optimal control sequence is generated. , but only the first item in the sequence is controlled Applied to the controlled object model to achieve real-time closed-loop control; The Hamiltonian function of vehicle speed planning within the prediction time domain of fuel cell heavy trucks is defined as: ； Where, is the covariate variable of the Hamiltonian function; The necessary conditions for the vehicle system to achieve optimal power distribution and optimal speed at the same time are: ； Where, is the state equation, is the co-state equation, is the control equation.

2. The stratified collaborative energy-saving control method for fuel cell heavy trucks in a dynamic traffic environment according to claim 1 is characterized in that: The steps for using LSTM neural network to predict the speed of the preceding vehicle include: The long short-term memory network is used to build the preceding vehicle speed prediction framework, and the preceding vehicle historical speed sequence and historical acceleration series As input features, For the current moment, = , Index variable for historical moments, representing the time from the current moment The time step of the backtracking, its value range , is the total length of historical moments, Indicates the speed of the vehicle ahead. Represents the acceleration of the preceding vehicle; through a nonlinear mapping function Generate the preceding vehicle speed sequence in the prediction time domain , , To predict the time domain index variable, representing the The time step of the prediction, its value range , is the prediction time domain step size.

3. The stratified collaborative energy-saving control method for fuel cell heavy trucks in a dynamic traffic environment according to claim 1 is characterized in that: The steps of designing a dual-time domain adaptive equivalent consumption minimization strategy, combining the predicted operating condition data in the rolling time domain, dynamically adjusting the equivalent factor, and generating the optimal control sequence allocation for the fuel cell power include: The reference value of the power battery SOC is set to be consistent with the starting value, and the power battery can be charged by the fuel cell during the entire driving process. Therefore, minimizing the hydrogen consumption of the fuel cell is used as the objective function. At the same time, considering the physical characteristics of the power system, the state variable of the optimization target is the power battery SOC, and the control variable is the fuel cell power. ; ； ； ； ； ； Where, is the fuel cell hydrogen consumption rate, and is related to the fuel cell output power There is a functional relationship; 、 They are The upper and lower limits allowed; 、 are the upper and lower limits of fuel cell power; Output power to the power battery; 、 are the upper and lower limits of the power of the power battery, and The upper and lower limits of the fuel cell power fluctuation value are: = - ; The initial Value; open circuit voltage of power battery and power battery internal resistance With battery There is a functional relationship; is the battery current, is the nominal capacity of the battery, is the DC / DC converter efficiency; is the time step between adjacent moments; The ECMS algorithm is an engineering implementation of the PMP theory in real-time control scenarios. It achieves the approximation of the global optimal solution through local optimization through a strict mapping relationship between equivalent factors and PMP co-state variables. For the energy management problem of fuel cell heavy trucks, the Hamiltonian function is defined as: ； Where: is a co-state variable, and the state variable is the power battery , the control variable is the fuel cell output power ; As a real-time optimization method based on the PMP framework, ECMS maps electricity consumption to equivalent hydrogen consumption by introducing an equivalent factor to build a unified energy optimal control framework. The calculation of the total hydrogen consumption rate is expressed as: ； Where: is the equivalent hydrogen consumption rate, is the power battery energy consumption rate, is the equivalence factor; The power consumption rate of the power battery The relationship is expressed as: ； Where, The low calorific value of hydrogen; Derived and The equivalence between the two; ； The PMP method has thus far yielded a global optimal solution for various driving conditions. This optimal solution can serve as a training dataset for the equivalent factor adjustment model. Given the significant time-varying characteristics of the equivalent factor due to the randomness and unpredictability of driving conditions, it is necessary to establish a dynamic, nonlinear relationship between the equivalent factor and the vehicle's operating state, SOC, and future speed trends. Neural network algorithms, with their strong nonlinear mapping and model-free optimization properties, can avoid the reliance of traditional regression methods on explicit models. By dynamically tuning parameters to approximate the optimal solution space for complex systems, they are suitable for equivalent factor prediction. A neural network training architecture based on the offline optimization database of PMP principle is constructed to achieve online dynamic prediction of equivalent factors. The model input layer adopts multi-dimensional feature vector design: , required power, main vehicle speed, acceleration and main vehicle planned speed in the rolling time domain; the output layer is the optimal co-state variable at the current moment , and then through and The equivalence relationship between the two is converted into the corresponding equivalent factors ; Optimal Equivalence Factor It is sensitive to working conditions and needs to make corresponding adaptive adjustments according to unknown working conditions. Feedback, where the speed ratio in the future section is used as a feedforward adjustment of the equivalent factor, Feedback acts as a post-compensation mechanism, combining the effects of the two on the equivalent factor Perform adaptive correction, the formula is shown as follows: ； ； ； ； Where, is the corrected equivalence factor, is the adjustment coefficient, is the proportionality coefficient; For the prediction time domain The standard deviation of vehicle speed within is the average vehicle speed in the prediction time domain.

Images