Fuel cell anticipatory start-stop method and system for coupled control of automotive powertrain

Abstract

The application is suitable for the technical field of fuel cell automobile control, and provides a fuel cell predictive start-stop method and system for automobile power system coupling control, which comprises the following steps: acquiring multi-dimensional data in real time; inputting the multi-dimensional data into a pre-trained driving trend prediction model to output predicted driving trend information; calculating the whole vehicle demand power at the current time according to the average speed prediction result; based on the sea squirt swarm algorithm, taking the maximum demand power prediction result as the target power reference of the fuel cell, taking the whole vehicle demand power as the total power benchmark, constructing a multi-objective optimization function, and solving the multi-objective optimization function in real time to obtain optimal start-stop control parameters; and generating and executing the coordinated start-stop control instruction according to the optimal start-stop control parameters. The application can realize the dynamic coordination of fuel cell start-stop control and the power system, and achieve the dual goals of efficient energy utilization and millisecond-level real-time response.

Description

Technical Field

[0001] This invention belongs to the field of fuel cell vehicle control technology, and particularly relates to a predictive start-stop method and system for fuel cell powertrain coupling control. Background Technology

[0002] With the current trend of accelerating the transformation of the energy structure towards clean energy, the green transformation of the transportation sector has attracted much attention as a key direction. Compared with lithium battery new energy vehicles, hydrogen fuel cell vehicles have irreplaceable potential in heavy-duty long-distance transportation scenarios due to their advantages such as high energy density and fast refueling time, making them a key development direction for the future automotive industry. However, the characteristics of the power system of hydrogen fuel cell heavy trucks determine that they face core challenges such as "slow dynamic response, sensitivity to battery aging, and difficulty in energy consumption optimization." As a key link connecting the power system and operational needs, the fuel cell start-stop control strategy directly determines the economy, durability, and operational reliability of the entire vehicle. Its technological optimization is crucial for the large-scale deployment of hydrogen fuel cell heavy trucks.

[0003] Current fuel cell start-stop control technologies still have room for improvement and are difficult to fully adapt to the demands of dynamic traffic environments. On the one hand, the start-stop of a fuel cell system involves the dynamic coupling of multiple subsystems, especially the air system, which exhibits significant time-varying characteristics. During startup, the cathode air pressure needs to be gradually built up using an air compressor, and the airflow needs to be adjusted to match the proton exchange membrane's requirements. During shutdown, targeted purging is required to maintain a safe threshold for membrane water content. These start-stop subsystems, in turn, need to be deeply coupled with multiple aspects of the power system. However, current research has limitations: one type of research focuses on optimizing the dynamic response of the fuel cell system itself, without considering the global demands of the power system, such as the drive motor's power output and the state of the power battery. Another type of research focuses on the control of the power system components, neglecting the dynamic constraints of fuel cell start-stop, thus failing to realize the coupling value between the two and not considering the impact of time scale differences among multiple subsystems, which can easily lead to battery shock and power overshoot during startup.

[0004] On the other hand, in traditional optimization algorithms, dynamic programming requires knowledge of the global operating conditions to optimize start-stop parameters, and the computational load of model predictive control increases exponentially with the prediction time domain. Both of these are difficult to meet the millisecond-level cycle requirements of hydrogen fuel cell start-stop control, making it difficult to optimize start-stop parameters in real time to respond to external traffic disturbances, which further affects operational stability.

[0005] Specifically, in the existing technical solutions: (1) Patent CN120503662A discloses a multimodal energy management control method and system for new energy vehicles based on event triggering, including: first calculating the total demand power, using a Bayesian optimized RBFNN model combined with historical speed data, and predicting short-term vehicle speed with a 5-second sliding window; classifying four types of working conditions according to features such as average speed, and matching different power output modes; constructing a cost function with SOC constraints based on model predictive control, and converting it into quadratic programming to find the optimal power; introducing an event triggering mechanism to reduce the amount of computation; but not coupling energy distribution and start-stop control, without aging adaptation, and the network data only contains historical vehicle speed. (2) Patent CN120621167A discloses a fuel cell energy management method and device based on fuzzy control, including: determining the required power of the whole vehicle and the SOC of the power battery as influencing factors, collecting voltage and current to calculate SOC; fuzzifying the data and determining the membership degree; combining the preset rule table to infer the fuzzy set to which the power belongs, and using the maximum membership degree method to defuzzify to obtain the predicted value; monitoring the actual value to optimize the fuzzy process; but it does not associate start-stop control and does not coordinate the time scale of the subsystem.

[0006] (3) Patent CN118770006A discloses an energy management strategy based on fuzzy rules, including: constructing a fuzzy rule base based on real-time vehicle speed and power battery SOC, dynamically allocating the output power of fuel cells and power batteries; avoiding the fuel cell from operating in a low-load state through rule constraints, thereby achieving power supply and demand balance and extending the service life of fuel cells. However, the rule base relies on expert experience and has poor adaptability to untrained operating conditions; it does not combine driving trend prediction and network data, resulting in limited global optimization effect.

[0007] (4) Patent CN120229111A discloses a power distribution and drive control system for a fuel cell heavy truck, including: using multiple sensors to collect vehicle status, equipment parameters and environmental data, capturing working condition characteristics and predicting power demand through neural networks; combining efficiency and lifespan mapping models to find the optimal operating point of the fuel cell, and realizing dynamic power distribution and drive control; however, the start-stop parameters are not optimized, and the network data only contains maps and slopes, lacking service information such as traffic flow and hydrogen refueling stations, resulting in insufficient prediction accuracy.

[0008] In summary, the existing technology currently has the following drawbacks:

[0009] (1) Insufficient coupling between fuel cell start-stop control and vehicle power system, and the coupling mechanism needs improvement: Most existing solutions focus on the dynamic optimization of the fuel cell start-stop system, without considering the overall power system requirements such as the power output of the drive motor and the charging and discharging status of the power battery; vehicle-level solutions focus more on the power control of the power system, ignoring the dynamic characteristic constraints of the start-stop phase, and failing to establish an effective linkage mechanism between the two. This decoupling leads to the failure to reserve instantaneous energy consumption during the start-up phase, causing a shock to the power battery, and the dynamic effect of the fuel cell leads to high power overshoot, accelerating battery aging. During the shutdown phase, parameter mismatch causes hydrogen energy waste or subsystem damage, which is more prominent under complex conditions such as congestion and heavy-load hill climbing.

[0010] (2) The time scale differences of multiple subsystems are not fully coordinated, and the response rhythm is prone to mismatch: Existing solutions often do not take into account the difference in response speed between the start and stop of the drive motor, power battery and fuel cell, and do not design time coordination logic in a targeted manner. Under sudden working conditions, the responses of each subsystem may be asynchronous, and it is difficult to avoid the occurrence of a short-term power gap or power overshoot, which will have a certain impact on the stability of the vehicle operation. Summary of the Invention

[0011] The purpose of this invention is to provide a predictive start-stop method for fuel cells in automotive powertrain coupling control, aiming to solve the above-mentioned technical problems.

[0012] This invention is implemented as follows: a predictive start-stop method for fuel cells in a vehicle powertrain coupling control system, comprising the following steps:

[0013] Real-time acquisition of multi-dimensional data; the multi-dimensional data includes vehicle status data as well as environmental and traffic data;

[0014] Multi-dimensional data is input into a pre-trained driving trend prediction model, which outputs predicted driving trend information for a preset future time period. The predicted driving trend information includes average vehicle speed prediction results, maximum power demand prediction results, and operating condition type prediction results.

[0015] Based on the preset vehicle dynamics model, the vehicle power demand at the current moment is calculated according to the average vehicle speed prediction results.

[0016] Based on the tunic algorithm, combined with the preset fuel cell start-stop dynamic model and power battery SOC model, the maximum demand power prediction result is used as the target power reference of the fuel cell, and the vehicle demand power is used as the total power benchmark. A multi-objective optimization function is constructed, and the multi-objective optimization function is solved in real time to obtain the optimal start-stop control parameters.

[0017] Based on the optimal start-stop control parameters, generate and execute coordinated start-stop control commands.

[0018] Furthermore, the vehicle's own status data includes historical vehicle speed and historical average acceleration; the environmental and traffic data includes gradient, speed of the vehicle in front, distance to the vehicle in front, remaining time of the traffic light ahead, distance between the vehicle and the refueling station, and waiting time at the refueling station.

[0019] Furthermore, the driving trend prediction model is constructed using Bi-LSTM, specifically including:

[0020] Input layer: consists of 8 neurons, with an input vector of... , where v ave For historical vehicle speeds, a avg Let i be the historical average acceleration, i be the slope, and T be the acceleration. light v represents the remaining time of the traffic light ahead. lead For the speed of the vehicle in front, d lead To determine the distance to the vehicle in front, d h2 t is the distance between the vehicle and the refueling station. queue Queue times at refueling stations;

[0021] Hidden layers: consist of 2 layers, each with 64 neurons, using ReLU or Softmax activation functions, with a Dropout layer between the two layers;

[0022] Output layer: consists of 3 neurons, with an output vector of... , where v pred For the prediction of future average vehicle speed, P req_max The forecast result for the maximum future demand power is denoted by type, which represents the forecast result for the operating condition type.

[0023] The driving trend prediction model learns by iteratively optimizing parameters and mastering the correspondence between input features and patterns. During training, training set data is processed in batches and input into the driving trend prediction model. After each batch of data undergoes forward propagation, the prediction results are compared with the actual labels, and a hybrid loss function L is used. total The calculation method is as follows:

[0024] ;

[0025] in, The mean squared error is used to optimize regression tasks and is calculated as follows:

[0026] ;

[0027] In the formula, This represents the number of samples in the training set. These are the actual vehicle speed and power, respectively.

[0028] Cross-entropy loss is used to optimize operating conditions, and is calculated as follows:

[0029]

[0030] In the formula, Representing the The first sample A real label, This refers to the first The corresponding sample of the th sample The predicted probability of the item; These are the weighting coefficients;

[0031] The backpropagation algorithm is used to determine the gradient of the hybrid loss function with respect to the parameters of the driving trend prediction model. Then, the optimizer is used to adjust the parameters of the driving trend prediction model according to the gradient. This process is repeated until the predetermined training cycle is completed. After each training cycle, the training effectiveness of the driving trend prediction model is evaluated by calculating the loss value.

[0032] After training the driving trend prediction model, test set data is input into the trained driving trend prediction model to obtain prediction results, which are then compared with actual data to evaluate the accuracy detection and recognition performance of the driving trend prediction model.

[0033] Furthermore, the formula for the vehicle dynamics model is as follows:

[0034] ;

[0035] In the formula, For the power required by the whole vehicle, For the efficiency of the transmission system, The angle between the plane where the vehicle is located and the horizontal plane. For vehicle quality, It is the acceleration due to gravity. The rolling resistance coefficient, For vehicle speed, The air drag coefficient, For windward area, This is the rotational mass conversion factor.

[0036] Furthermore, the fuel cell start-stop dynamic model synchronously simulates voltage output characteristics, gas path pressure dynamics, and auxiliary system energy consumption, wherein the auxiliary system includes an air compressor; wherein, the fuel cell voltage is modeled as a function of current density, cathode oxygen partial pressure, and membrane water content, as shown in the following equation:

[0037] ;

[0038] In the formula, It is activation loss. It is an ohm loss. It is a concentration loss. It is the number of stacked units; It is the Nernst potential, which depends on the temperature of the reactants and products. And partial pressure:

[0039] ;

[0040] In the formula, This indicates the partial pressure of hydrogen at the anode of the fuel cell. This indicates the partial pressure of oxygen at the cathode of the fuel cell;

[0041] The dynamic model of the airflow pressure consists of three parts: the intake manifold, the cathode, and the exhaust manifold. The intake manifold connects the air compressor outlet to the cathode inlet. Based on the linear nozzle principle, the outlet flow rate of the intake manifold is proportional to the difference between the pressure inside the manifold and the cathode pressure, as shown in the following formula:

[0042] ;

[0043] In the formula, This refers to the mass flow rate at the intake manifold outlet. This is the intake manifold flow coefficient. This refers to the pressure inside the intake manifold. This refers to the pressure inside the cathode cavity;

[0044] Based on the ideal gas law and the law of conservation of mass, the pressure inside the intake manifold dynamically changes with the mass difference between the intake and exhaust gases, as shown in the following formula:

[0045] ;

[0046] In the formula, Let be the ideal gas constant of air. For the intake manifold volume, This refers to the mass flow rate at the air compressor outlet. The outlet gas temperature of the air compressor. This refers to the temperature of the gas inside the intake manifold.

[0047] The exhaust manifold connects the cathode outlet to the outside. After the exhaust gas flows from the cathode into the exhaust manifold, the outflow is regulated by the back pressure valve, resulting in a pressure difference change within the manifold, as shown in the following formula:

[0048] ;

[0049] In the formula, This refers to the pressure inside the exhaust manifold. The temperature of the gas inside the exhaust manifold. For exhaust manifold volume, This represents the cathode outlet mass flow rate. This refers to the mass flow rate at the exhaust manifold outlet.

[0050] Based on the principle of adiabatic compression, the formula for the energy consumption of the auxiliary system is as follows:

[0051] ;

[0052] In the formula, For air compressor energy consumption, This refers to the intake molar flow rate of the air compressor. For the molar mass of air, The specific heat capacity of air at constant pressure. This refers to the inlet air temperature of the air compressor. For air compressor efficiency. Where c is the inlet pressure of the air compressor, and c is the air adiabatic index. This refers to the intake manifold pressure.

[0053] Furthermore, the power battery SOC model is constructed based on the equivalent circuit model method; wherein, the output power of the power battery... for:

[0054] ;

[0055] In the formula, Represents open-circuit voltage. Indicates battery current. The internal resistance of the power battery; battery current. The calculation formula is as follows:

[0056] ;

[0057] The SOC value of the power battery at the current moment is calculated using the ampere-hour integral method, as shown in the following formula:

[0058] ;

[0059] In the formula, Q represents the current SOC value of the power battery; bat This refers to the battery capacity.

[0060] Furthermore, the multi-objective optimization function takes the vehicle energy consumption as an example. Battery power fluctuation and power control overshoot To optimize the target, start-stop control parameters include energy distribution ratio. Air compressor speed Back pressure valve opening ;in, Power required for the entire vehicle; This refers to the output power of the fuel cell.

[0061] Furthermore, based on the slug-sea swarm algorithm, combined with the pre-set fuel cell start-stop dynamic model and power battery SOC model, the maximum demand power prediction result is used as the target power reference for the fuel cell, and the vehicle demand power is used as the total power benchmark. A multi-objective optimization function is constructed, and the multi-objective optimization function is solved in real time to obtain the optimal start-stop control parameters. The specific steps include:

[0062] (1) Chaotic initialization of the population:

[0063] Population size N=30; dimension D=3, corresponding to the optimal energy allocation ratio Optimal air compressor speed Optimal back pressure valve opening ;

[0064] Generate a uniformly distributed initial population using an improved Tent chaotic map:

[0065]

[0066] In the formula, , A random number in the range [0,1]. It is a chaotic sequence;

[0067] Mapping chaotic sequences to the solution space :

[0068]

[0069] In the formula, , Define the upper and lower bounds of the variable;

[0070] (2) Randomly generate the initial population and construct a multi-objective optimization function J:

[0071] ;

[0072] In the formula, Let be the weighting coefficient, satisfying ;

[0073] The constraints st are as follows:

[0074] ;

[0075] In the formula, Let be the output power of the fuel cell at time t. The minimum permissible fuel cell output power, The maximum permissible output power of the fuel cell; Let be the output power of the power battery at time t. The minimum permissible power battery output power, The maximum permissible power battery output power; Let t be the SOC value of the power battery at time t. The minimum allowable SOC value, The maximum allowed SOC value;

[0076] Among them, the energy consumption of the whole vehicle The calculation formula is as follows:

[0077] ;

[0078] In the formula, For the hydrogen consumption of fuel cells, This refers to the equivalent hydrogen consumption of the power battery.

[0079] Battery power fluctuation The calculation formula is as follows:

[0080] ;

[0081] In the formula, This is the SOC fluctuation penalty coefficient. This is the penalty coefficient for fuel cell power fluctuations;

[0082] Power control overshoot The calculation formula is as follows:

[0083] ;

[0084] In the formula, Real-time power of the fuel cell; Determined from the maximum demand power prediction results;

[0085] (3) Leader-follower iterative optimization:

[0086] Leaders renew:

[0087] ;

[0088] In the formula, For the upper realm, The lower bound is given by , rand is a random number in the range [0,1], and l is the number of iterations.

[0089] calculate and If the former is smaller, then update the leader position;

[0090] followers renew:

[0091]

[0092] Introducing networked data feedback: predicting the gradient of the incline. hour, The search weight increased from 1.0 to 1.5;

[0093] (4) Convergence judgment:

[0094] The fitness error is determined when the number of iterations reaches T=50, or after three consecutive iterations. Stop when the time is right and output the optimal start / stop control parameters.

[0095] Furthermore, the steps of generating and executing coordinated start-stop control commands based on optimal start-stop control parameters specifically include:

[0096] Power allocation is calculated based on optimal start-stop control parameters to ensure dynamic balance under multiple constraints;

[0097] The power calculation formula is as follows:

[0098] ;

[0099] Multi-constraint control logic:

[0100] (1) Discharge constraint, The discharge power change rate must be ≤5kW / s and the SOC must be ≥30%.

[0101] (2) Charging constraints, The following conditions must be met: SOC ≤ 80%, and charging power change rate ≤ 5kW / s;

[0102] (3) Constraints of fuel cells: Need to be Within the range; Rated power of the fuel cell;

[0103] Parameter fine-tuning mechanism: If the constraint is not met, fine-tune by ±0.05. value:

[0104] If SOC < 30% and P bat >0: Increasing k by 0.05 increases P fc Percentage;

[0105] If SOC > 80% and P bat <0: k decreases by 0.05, reducing P fc Percentage;

[0106] If P fc >0.9×P fc_rated Reduce k by 0.05 to avoid overload.

[0107] Another object of the present invention is to provide a fuel cell predictive start-stop system for coupled control of an automotive powertrain, for implementing the above-mentioned fuel cell predictive start-stop method, comprising:

[0108] The data acquisition module is used to acquire multi-dimensional data in real time; the multi-dimensional data includes vehicle status data as well as environmental and traffic data.

[0109] The driving trend prediction module is used to input multi-dimensional data into a pre-trained driving trend prediction model and output predicted driving trend information for a preset future time period. The predicted driving trend information includes average vehicle speed prediction results, maximum power demand prediction results, and operating condition type prediction results.

[0110] The demand power calculation module is used to calculate the vehicle's demand power at the current moment based on the preset vehicle dynamics model and the average vehicle speed prediction results.

[0111] The multi-objective optimization module is used to construct a multi-objective optimization function based on the Zunhaishao swarm algorithm, combined with the preset fuel cell start-stop dynamic model and power battery SOC model. It uses the maximum demand power prediction result as the target power reference of the fuel cell and the total vehicle demand power as the total power benchmark. The module then solves the multi-objective optimization function in real time to obtain the optimal start-stop control parameters.

[0112] The start-stop control module is used to generate and execute coordinated start-stop control commands based on the optimal start-stop control parameters.

[0113] The predictive start-stop method for fuel cells in automotive powertrain coupling control provided by this invention has the following advantages compared to existing technologies:

[0114] 1. Deep integration of fuel cell start-stop and vehicle power system, resulting in stronger dynamic characteristic adaptability: This invention establishes a coupled optimization mechanism between fuel cell air system start-stop parameters and energy distribution. By using the Zunhaishao swarm algorithm to synchronously output the energy distribution ratio and start-stop parameters, it ensures sufficient power supply when the air compressor is loaded, matches the air flow with the fuel cell power, reduces membrane damage rate, and reduces energy consumption during start-stop.

[0115] 2. Coordinate the time scale differences of multiple subsystems to improve system stability: To address the differences in the start-stop response speeds of the drive motor, power battery, and fuel cell, a time coordination logic and priority scheduling mechanism are designed to synchronize the response rhythm of each subsystem and avoid exacerbating coordination problems due to inconsistent time scales. Compared with existing technologies, the system stability during the start-up phase is better.

[0116] 3. Superior real-time performance and higher control precision: The optimization time of the Zunhaichao group algorithm meets the millisecond-level control cycle and the power change rate is constrained to ≤5kW / s. Compared with existing technologies, it reduces the battery control overshoot during start-stop and improves the overall vehicle operation stability.

[0117] 4. High utilization rate of connected data and better overall energy consumption: The integration of environmental, traffic and service-related connected data in all dimensions can coordinate the planning of remaining electricity and hydrogen refueling time. Compared with existing technologies, it reduces hydrogen consumption per 100 kilometers and stabilizes SOC fluctuation within a reasonable range of 30%-80%, thereby reducing energy consumption. Attached Figure Description

[0118] Figure 1 This is a structural block diagram of the coupling control system for the power system of a hydrogen fuel cell heavy-duty truck provided in an embodiment of the present invention.

[0119] Figure 2 This is a schematic flowchart of the fuel cell predictive start-stop method provided in an embodiment of the present invention. Detailed Implementation

[0120] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0121] This invention focuses on the problem of coordinated control between the fuel cell start-stop and the vehicle power system in a connected vehicle environment for hydrogen fuel cell heavy-duty trucks. Specifically, the power system of a hydrogen fuel cell heavy-duty truck consists of multiple subsystems, including a fuel cell, a power battery, a drive motor, an air compressor, and a back pressure valve. The start-stop of the fuel cell is not a simple on / off operation, but a complex process requiring dynamic coordination among multiple subsystems through actions such as pressurizing the air compressor and adjusting the opening of the back pressure valve. The control accuracy of this process directly affects the vehicle's economy, durability, and stability. However, existing vehicle control strategies suffer from insufficient coordination during the fuel cell start-stop phase, failing to establish an effective linkage mechanism with multiple core components of the vehicle power system (such as the drive motor and power battery), resulting in a disconnect between start-stop control and power demand. This problem is more pronounced in the typical complex operating conditions of heavy-duty long-distance transportation. For example, when there are frequent starts and stops in congested sections, the power battery has to passively bear the instantaneous overload discharge because there is no reserve for the energy consumption of the air compressor. When starting and accelerating or climbing hills under heavy load, the power response of the fuel cell lags behind the demand of the motor, which can easily lead to a power gap that affects driving performance. These problems not only lead to a significant increase in the energy consumption of the whole vehicle, but also accelerate the aging of the power battery and the core components of the fuel cell, which seriously restricts the efficiency release and large-scale application of hydrogen fuel cell heavy trucks, becoming a key bottleneck for the development of the industry.

[0122] In existing control strategies, fuel cell start-stop and vehicle powertrain systems exhibit decoupling: at the vehicle level, start-stop is treated as a static process without reserving dynamic energy consumption, while at the system level, the focus is on its own dynamic optimization without matching the vehicle's power requirements. This decoupling problem is further highlighted in specific technical solutions. Existing fuel cell start-stop control technologies often fall into two categories. One category triggers start-stop based solely on a single operating condition parameter, treating fuel cell start-stop as a mechanical operation with fixed parameters, ignoring its own dynamic process and future operating condition trends. This results in the lack of reserved loading energy consumption during the start-up phase, causing impact on the power battery, and frequent starts accelerating battery life degradation. During the shutdown phase, mismatch between purging parameters and vehicle load leads to hydrogen waste or subsystem damage. The other category focuses more on the dynamics of the fuel cell subsystem but fails to deeply integrate the response rhythm of multiple components in the vehicle powertrain system, ultimately causing a disconnect between start-stop control and the actual needs of the vehicle and the equipment status. Fuel cell start-stop and multiple components of the powertrain system have a two-way mutual influence. Start-stop parameter settings determine the subsystem's response rhythm and energy demand, while the overall vehicle powertrain system status and equipment status, in turn, affect the rationality of start-stop control. Without an effective coupling mechanism, problems such as power deficit and increased energy consumption can easily arise. It is this systemic decoupling that leads to a serious lack of coordination between fuel cell start-stop control and multiple components of the vehicle powertrain system, ultimately becoming a key bottleneck restricting the optimization of energy consumption and the extension of lifespan in hydrogen fuel cell heavy-duty trucks. The key path to solving this core problem and unleashing the efficiency potential of heavy-duty trucks lies in the coordinated coupling mechanism of powertrain system components. Supported by multi-dimensional operating condition prediction data provided by the connected environment, dynamic coordination between start-stop control and multiple components of the powertrain system can be achieved, ultimately reaching multi-objective real-time optimization with optimal energy consumption, minimum degradation, and minimal overshoot.

[0123] The main technical problems addressed by the embodiments of the present invention are as follows:

[0124] 1. In existing fuel cell start-stop control strategies for hybrid electric vehicles, there is a certain degree of decoupling between vehicle-level and system-level applications. At the vehicle level, fuel cell start-stop control is simplified to a static triggering process, focusing on matching global demands such as motor power output and vehicle load changes, neglecting the dynamic characteristics of the start-stop phase. At the system level, fuel cell start-stop control primarily considers the dynamic processes of multiple subsystems working together, such as air compressor pressurization and back pressure valve opening adjustment, but fails to match the overall vehicle operating conditions, ultimately resulting in a disconnect between start-stop control and actual vehicle needs. The insufficient integration of these differences in existing technology leads to vehicle-level fuel cell start-stop decisions failing to account for dynamic system energy consumption, easily causing power battery impact; and system-level fuel cell start-stop control failing to match the overall vehicle operating rhythm, potentially causing power overshoot and component damage.

[0125] 2. Existing fuel cell start-stop control strategies for hybrid electric vehicles have shortcomings in their coordination with the time-scale characteristics of multiple subsystems. The drive motor, as the power actuator, needs to adapt quickly to sudden changes in road conditions; fuel cell start-stop is a dynamic process involving multiple steps, and parameter adjustments inherently have lag; although the power battery acts as a power buffer, its charging and discharging response still has a delay, and frequent rapid charging and discharging accelerates degradation. However, existing solutions do not design specific collaborative logic to address this issue, only employing uniform static control rules, leading to a mismatch in the response rhythms of various subsystems. Especially in scenarios where the vehicle requires high power output, when the fuel cell is triggered to supplement power, due to the inherent lag in fuel cell start-stop, existing static rules, in order to quickly meet high power demands, often set the initial power target after fuel cell start-up too high. After actual fuel cell start-up, due to factors such as a sudden increase in gas pressure and lag in membrane water content adjustment, the power output is prone to significant overshoot. Furthermore, the lack of coordinated buffering by the power battery not only wastes energy but also subjectes the fuel cell and drive motor to sudden power changes, accelerating wear and ultimately leading to decreased power system reliability and shortened lifespan of core components.

[0126] To address the above technical issues, such as Figure 1 As shown, this invention focuses on the coupling mechanism between fuel cell start-stop and the vehicle power system, providing a coupling control system for the power system of a hydrogen fuel cell heavy-duty truck. This system integrates multiple subsystems, including the fuel cell, power battery, air compressor, and back pressure valve. It fully considers the significant time-scale inconsistencies arising from the physical characteristics of each subsystem, as well as the complex coupling relationship between start-stop control and the vehicle power system. To overcome the limitations of traditional hierarchical control in coordinating multiple time-scale subsystems, this invention also proposes a predictive start-stop method for fuel cells in a connected environment, which can be implemented through the following technical solutions:

[0127] To address the actual operational requirements of fuel cell start-stop control in a vehicle environment, this paper proposes an integrated fuel cell start-stop control method that considers the time coordination of the entire vehicle power system, leveraging the power allocation flexibility of energy distribution and the timing control capabilities of start-stop control. This method fully explores the energy-saving potential of heavy-duty trucks under different operating conditions by combining the time-scale characteristics of the drive motor, power battery, and air compressor. By establishing a dynamic coordination logic: during the start-up phase, the energy consumption required for air compressor loading is reserved in advance based on the charging and discharging response rhythm of the power battery. Simultaneously, power coordination is achieved by dynamically allocating the output ratio of the fuel cell and the power battery, allowing them to share the power demand of the entire vehicle. This avoids the power gap caused by the high power demand of the drive motor and the lag in battery response, and also reduces the instantaneous power overshoot that is prone to occur when the fuel cell outputs alone through the buffering effect of the power battery, ensuring stable power output during the start-up process. During the shutdown phase, the energy allocation ratio is adjusted to maintain low power output and reduce hydrogen redundancy by synchronously adapting to the second-level purging sequence of the fuel cell.

[0128] Specifically, such as Figure 2 As shown, in one embodiment of the present invention, a fuel cell predictive start-stop method for coupled control of an automotive powertrain system is provided, comprising the following steps:

[0129] S1. Real-time acquisition of multi-dimensional data;

[0130] S2. Input multi-dimensional data into a pre-trained driving trend prediction model and output predicted driving trend information for a future preset time period; the predicted driving trend information includes average vehicle speed prediction results, maximum power demand prediction results, and operating condition type prediction results.

[0131] S3. Based on the preset vehicle dynamics model, calculate the vehicle power demand at the current moment according to the average vehicle speed prediction results.

[0132] S4. Based on the Slug Group Algorithm (SSA), combined with the preset fuel cell start-stop dynamic model and power battery SOC model, the maximum demand power prediction result is used as the target power reference of the fuel cell, and the vehicle demand power is used as the total power benchmark. A multi-objective optimization function is constructed, and the multi-objective optimization function is solved in real time to obtain the optimal start-stop control parameters.

[0133] S5. Generate and execute coordinated start-stop control commands based on the optimal start-stop control parameters.

[0134] In this embodiment of the invention, a dynamic coupling mechanism between fuel cell start-stop control and multiple components of the power system is established to achieve coordinated optimization of start-stop and the vehicle power system. Instantaneous energy consumption is reserved in advance during the start-up phase, and power synergy is achieved by dynamically allocating the output ratio of the fuel cell and the power battery, allowing both to share the power demand of the vehicle, ensuring stable power output during start-up. During the shutdown phase, the power demand for purging is adapted, promoting dynamic coordination between the two according to real-time operating conditions, and minimizing battery impact and hydrogen waste. Furthermore, this embodiment coordinates the time scale differences of multiple subsystems, clarifies the response priority of each subsystem, synchronizes the response rhythm of each system, and reduces the probability of power gaps and overshoot. Simultaneously, SSA is used to reduce the computational complexity of optimization, aiming to meet millisecond-level control cycles as much as possible; a multi-objective optimization function including energy consumption, decay, and overshoot is constructed, and the optimization results are dynamically corrected to ensure good matching of demand during sudden changes in operating conditions, balancing economy and reliability.

[0135] In practical applications, multi-dimensional data includes vehicle status data as well as environmental and traffic data. Vehicle status data includes, but is not limited to, historical vehicle speed and historical average acceleration. Environmental and traffic data comes from connected vehicle data and specifically includes, but is not limited to, gradient, speed of the vehicle in front, distance to the vehicle in front, remaining time at the traffic light ahead, distance between the vehicle and the refueling station, and refueling station queue time. It should be noted that multi-dimensional data can be preprocessed, such as by extended Kalman filtering, before being input into the subsequent model.

[0136] In a preferred embodiment of the present invention, the driving trend prediction model is a core prerequisite for fuel cell start-stop prediction and energy demand planning of heavy-duty trucks in a connected environment. It needs to accurately capture the temporal characteristics of complex operating conditions under heavy-load scenarios based on multi-dimensional data, and output driving trend information within the next short period, providing key input for subsequent power demand calculation of the vehicle dynamics model and response rhythm matching of the drive motor model. Considering the strong temporal correlation of driving states in heavy-load scenarios, this driving trend prediction model is constructed using a Bi-LSTM (Bidirectional Long Short-Term Memory) network, fully utilizing its ability to capture dependencies in long-term time-series data and avoiding prediction biases in traditional time-series models under complex operating conditions; specifically including:

[0137] (1) Constructing the Bi-LSTM neural network model:

[0138] Input layer: consists of 8 neurons, with an input vector of... , where v ave For the historical 5-second speed, a avg The average acceleration over the past 5 seconds is given by i, where i is the slope and T is the acceleration over the past 5 seconds. light v represents the remaining time of the traffic light ahead. lead For the speed of the vehicle in front, d lead To determine the distance to the vehicle in front, d h2t is the distance between the vehicle and the refueling station. queue Queue times at refueling stations;

[0139] Hidden layers: consist of 2 layers, each with 64 neurons, using ReLU (to avoid gradient vanishing) or Softmax activation functions, and a Dropout layer (with a dropout probability of 0.2) between the two layers to prevent overfitting.

[0140] Output layer (classification layer): consists of 3 neurons, with an output vector of... , where v pred P is the predicted average vehicle speed for the next 10 seconds. req_max This is the prediction result of the maximum power demand in the next 10 seconds, with type indicating the prediction result of the operating condition (e.g., 0 = stable, 1 = climbing, 2 = congestion).

[0141] (2) Model training and testing:

[0142] The driving trend prediction model learns by iteratively optimizing parameters and mastering the correspondence between input features and patterns. During training, training set data is processed in batches and input into the driving trend prediction model. After each batch of data undergoes forward propagation, the prediction results are compared with the actual labels, and a hybrid loss function L is used. total The calculation method is as follows:

[0143] ;

[0144] in, Mean squared error, used to optimize regression tasks ( The calculation method is as follows:

[0145] ;

[0146] In the formula, This represents the number of samples in the training set. These are the actual vehicle speed and power, respectively.

[0147] Cross-entropy loss is used to optimize the operating condition type (type), and is calculated as follows:

[0148]

[0149] In the formula, Representing the The first sample One real label (0 or 1). This refers to the first The corresponding sample of the th sample The predicted probability of the item; This is the weighting coefficient, typically set to 0.6;

[0150] The backpropagation algorithm is used to determine the gradient of the hybrid loss function with respect to the parameters of the driving trend prediction model. Then, the Adam optimizer is used to adjust the parameters of the driving trend prediction model according to the gradient. This process is repeated until the predetermined training cycle is completed. After each training cycle, the training effect of the driving trend prediction model is evaluated by calculating the loss value. If the loss value continues to decrease, it indicates that the model is gradually improving. If the loss value stagnates or fluctuates, it may be necessary to adjust the model parameters or expand the training dataset.

[0151] After training the driving trend prediction model, test set data is input into the trained driving trend prediction model to obtain prediction results, which are then compared with actual data to evaluate the accuracy detection and recognition performance of the driving trend prediction model.

[0152] In a preferred embodiment of the present invention, based on the vehicle dynamics model and the average vehicle speed prediction results, the real-time vehicle power demand can be calculated, providing a target benchmark for energy allocation; wherein, the formula of the vehicle dynamics model is as follows:

[0153] ;

[0154] In the formula, For the power required by the whole vehicle, For the efficiency of the transmission system, The angle between the plane where the vehicle is located and the horizontal plane. For vehicle quality, It is the acceleration due to gravity. The rolling resistance coefficient, For vehicle speed (when calculating the power demand of the entire vehicle, the predicted average vehicle speed can be substituted into the equation). ), The air drag coefficient, For windward area, This is the rotational mass conversion factor.

[0155] Furthermore, this embodiment of the invention also constructs a drive motor model; the drive motor is the core power actuator of the hydrogen fuel cell heavy-duty truck, and its core function is to convert the electrical energy supplied by the power battery and fuel cell into mechanical energy to drive the vehicle. The calculation formula for the output power of the drive motor is as follows:

[0156] ;

[0157] In the formula, For the output power of the drive motor, For the motor output torque, This refers to the motor speed. This refers to the efficiency of the motor.

[0158] ;

[0159] In the formula, The power allocated from the battery to the motor. The power allocated to the motor for the fuel cell.

[0160] ;

[0161] ;

[0162] Where k is the power distribution ratio (i.e., energy distribution ratio) of the fuel cell.

[0163] In a preferred embodiment of the present invention, the core of the fuel cell start-up and shutdown process is the multi-physics coupling of water, gas, heat and electricity. The fuel cell start-up and shutdown dynamic model needs to simultaneously simulate voltage output characteristics, gas path pressure dynamics and auxiliary system energy consumption. The auxiliary system includes an air compressor to ensure parameter matching (voltage and power, pressure and flow) during the start-up and shutdown phases, and avoid membrane damage or hydrogen waste caused by parameter mismatch.

[0164] Voltage is a direct representation of the energy output of a fuel cell, and its value dynamically changes with current density, reactant partial pressure, and membrane water content. Voltage stability requires close monitoring during start-up and shutdown. The fuel cell voltage is modeled as a function of current density, cathode oxygen partial pressure, and membrane water content, as shown in the following equation:

[0165] ;

[0166] In the formula, It is activation loss. It is an ohm loss. It is a concentration loss. It is the number of stacked units; It is the Nernst potential, which depends on the temperature of the reactants and products. And partial pressure:

[0167] ;

[0168] In the formula, This indicates the partial pressure of hydrogen at the anode of the fuel cell. This indicates the partial pressure of oxygen at the cathode of the fuel cell;

[0169] The pressure dynamics of the gas path system directly affect the reactant supply efficiency, and a stable pressure field needs to be established quickly during start-up and shutdown. The model of the gas path pressure dynamics consists of three parts: the intake manifold, the cathode, and the exhaust manifold. The intake manifold connects the air compressor outlet and the cathode inlet, and its outlet flow rate and internal pressure dynamics need to be simulated. The pressure difference between the intake manifold and the cathode is relatively small, so a simplified linear model is adopted. Based on the linear nozzle principle, the outlet flow rate of the intake manifold (i.e., the air flow rate entering the cathode) is proportional to the difference between the pressure inside the manifold and the cathode pressure, as shown in the following formula:

[0170] ;

[0171] In the formula, This refers to the mass flow rate at the intake manifold outlet. This is the intake manifold flow coefficient. This refers to the pressure inside the intake manifold. This refers to the pressure inside the cathode cavity;

[0172] Based on the ideal gas law and the law of conservation of mass, the pressure inside the intake manifold dynamically changes with the mass difference between the intake and exhaust gases, as shown in the following formula:

[0173] ;

[0174] In the formula, Let be the ideal gas constant of air. For the intake manifold volume, This refers to the mass flow rate at the air compressor outlet. The outlet gas temperature of the air compressor. This refers to the temperature of the gas inside the intake manifold.

[0175] The exhaust manifold connects the cathode outlet to the outside environment. It's necessary to simulate its internal pressure dynamics. After the exhaust gas flows from the cathode into the exhaust manifold, the outflow is regulated by a back pressure valve, causing a pressure difference change within the manifold. The formula is as follows:

[0176] ;

[0177] In the formula, This refers to the pressure inside the exhaust manifold. The temperature of the gas inside the exhaust manifold. For exhaust manifold volume, This represents the cathode outlet mass flow rate. This refers to the mass flow rate at the exhaust manifold outlet.

[0178] The air compressor is the most energy-intensive device in the PEMFC auxiliary system, and its parasitic power consumption directly affects the system's net output power. During startup, the power consumption must be provided by the battery; precise quantification is crucial to avoiding battery overload. Based on the adiabatic compression principle and considering the air compressor efficiency, the formula for the auxiliary system's energy consumption is as follows:

[0179] ;

[0180] In the formula, For air compressor energy consumption, This refers to the intake molar flow rate of the air compressor. For the molar mass of air, The specific heat capacity of air at constant pressure. This refers to the inlet air temperature of the air compressor. For air compressor efficiency. Where c is the inlet pressure of the air compressor, and c is the air adiabatic index. This refers to the intake manifold pressure.

[0181] In a preferred embodiment of the present invention, the power battery can overcome the shortcomings of fuel cells, such as slow response speed and inability to recover braking energy; the power battery SOC model is constructed based on the equivalent circuit model method; wherein, the output power of the power battery... for:

[0182] ;

[0183] In the formula, Represents open-circuit voltage. Indicates battery current. The internal resistance of the power battery; battery current. The calculation formula is as follows:

[0184] ;

[0185] The state of charge (SOC) of a power battery is a key parameter, reflecting its remaining usable capacity. This embodiment of the invention uses the ampere-hour integration method to calculate the SOC value of the power battery at the current moment, as shown in the following formula:

[0186] ;

[0187] In the formula, Q represents the current SOC value of the power battery; bat This refers to the battery capacity.

[0188] In a preferred embodiment of the present invention, the multi-objective optimization function is based on the vehicle energy consumption. ( Minimum), battery power fluctuation (degradation) ( Minimum) and power control overshoot ( The optimization objective is to minimize the energy consumption; the optimization variables are three dimensions, namely three start-stop control parameters, namely, the energy distribution ratio. Air compressor speed Back pressure valve opening ;in, Power required for the entire vehicle; This refers to the output power of the fuel cell.

[0189] In a preferred embodiment of the present invention, based on the Slug Group Algorithm (SSA), combined with a preset fuel cell start-stop dynamic model and power battery SOC model, the maximum demand power prediction result is used as the target power reference for the fuel cell, and the total vehicle demand power is used as the total power benchmark. A multi-objective optimization function is constructed, and the multi-objective optimization function is solved in real time to obtain the optimal start-stop control parameters. The specific steps include:

[0190] (1) Chaotic initialization of the population:

[0191] Population size N=30; dimension D=3, corresponding to the optimal energy allocation ratio Optimal air compressor speed Optimal back pressure valve opening ;

[0192] Generate a uniformly distributed initial population using an improved Tent chaotic map:

[0193]

[0194] In the formula, , A random number in the range [0,1]. It is a chaotic sequence;

[0195] Mapping chaotic sequences to the solution space :

[0196]

[0197] In the formula, , The upper and lower bounds of the variables are shown in Table 1;

[0198] Table 1

[0199] Optimize variables Upper Realm The Lower World 0.9 0.1 90000 r / min 30000 r / min 90% 10%

[0200] (2) Randomly generate the initial population and construct a multi-objective optimization function (fitness) J:

[0201] ;

[0202] In the formula, Let be the weighting coefficient, satisfying ;

[0203] The constraints st are as follows:

[0204] ;

[0205] In the formula, Let be the output power of the fuel cell at time t. The minimum permissible fuel cell output power, The maximum permissible output power of the fuel cell; Let be the output power of the power battery at time t. The minimum permissible power battery output power, The maximum permissible power battery output power; Let t be the SOC value of the power battery at time t. The minimum allowable SOC value, The maximum allowed SOC value;

[0206] Among them, the overall vehicle energy consumption (hydrogen consumption) The calculation formula is as follows:

[0207] ;

[0208] In the formula, For the hydrogen consumption of fuel cells, This refers to the equivalent hydrogen consumption of the power battery.

[0209] State of charge (SOC) fluctuations and fuel cell power fluctuations are the core factors accelerating the aging of power batteries and fuel cells. For power batteries, rapid SOC fluctuations directly cause membrane damage, leading to capacity decay. Meanwhile, the slow response and lagging power adjustment of fuel cells transfer the load of sudden power changes in the vehicle to the power battery, forcing it to passively and rapidly charge and discharge, indirectly increasing SOC fluctuations and damage risk. For the fuel cell itself, power fluctuations directly impact critical components; the proton exchange membrane tears due to repeated wet-dry cycles and sudden pressure changes, significantly shortening its lifespan.

[0210] Meanwhile, real-time monitoring of aging conditions presents numerous challenges. Essentially, internal microscopic damage cannot be directly measured, and estimations based solely on macroscopic parameters result in significant errors. Furthermore, vibrations and temperature differences under the complex operating conditions of heavy-duty trucks interfere with monitoring equipment. The strong coupling between aging and operating conditions also greatly reduces data reliability. Therefore, incorporating SOC volatility and fuel cell power fluctuations into the objective function as aging penalty terms is crucial. This approach eliminates the need for additional equipment, allows for real-time calculation, and avoids monitoring difficulties. It also suppresses aging-inducing factors at their source, enabling targeted optimization of high-frequency fluctuating operating conditions such as start-stop and congestion. Simultaneously, it synergizes with energy consumption and control stability objectives, ensuring that multi-dimensional optimizations do not conflict with each other.

[0211] Battery power fluctuation The calculation formula is as follows:

[0212] ;

[0213] In the formula, This is the SOC fluctuation penalty coefficient. This is the penalty coefficient for fuel cell power fluctuations;

[0214] Power control overshoot The calculation formula is as follows:

[0215] ;

[0216] In the formula, Real-time power of the fuel cell; The maximum power demand prediction results are determined by the output of the above driving trend prediction model;

[0217] In addition, the default values ​​for the above weighting coefficients are as follows: ;

[0218] When battery power fluctuates significantly ( When >0.5), The maximum value is increased to 0.4, prioritizing the reduction of battery power fluctuation weight; for brand new batteries. To balance multiple objectives;

[0219] Based on the driving trend prediction model's output of the driving condition type prediction result (0=smooth, 1=climbing, 2=congestion), the adjustment rules are as follows:

[0220] Stable (0), prioritize reducing the weight of overall vehicle energy consumption, i.e. ;

[0221] Climbing (1), balancing the weight of vehicle energy consumption and power control overshoot, i.e. ;

[0222] Congestion (2), prioritize increasing the weight of power control overshoot, i.e. ;

[0223] Weight normalization: ensure The normalization formula is: ;

[0224] (3) Leader-follower iterative optimization:

[0225] Leaders Update (Global Search):

[0226] ;

[0227] In the formula, For the upper realm, The lower bound is given by , rand is a random number in the range [0,1], and l is the number of iterations.

[0228] calculate and If the former is smaller, then update the leader position;

[0229] followers Update (partial refinement):

[0230]

[0231] Introducing networked data feedback: predicting the gradient of the incline. hour, The search weight increased from 1.0 to 1.5;

[0232] (4) Convergence judgment:

[0233] The fitness error is determined when the number of iterations reaches T=50, or after three consecutive iterations. Stop when the time is right, and output the optimal start / stop control parameters ( , , The time taken for a single optimization is ≤0.1s, which meets the 100ms level control cycle.

[0234] In a preferred embodiment of the present invention, the step of generating and executing a coordinated start-stop control command based on optimal start-stop control parameters specifically includes:

[0235] (1) Energy allocation decision:

[0236] Power allocation is calculated based on optimal start-stop control parameters to ensure dynamic balance under multiple constraints;

[0237] The power calculation formula is as follows:

[0238] ;

[0239] Multi-constraint control logic:

[0240] (1) Discharge constraint, The discharge power change rate must be ≤5kW / s and the SOC must be ≥30%.

[0241] (2) Charging constraints, The following conditions must be met: SOC ≤ 80%, and charging power change rate ≤ 5kW / s;

[0242] (3) Constraints of fuel cells: Need to be Within the range; Rated power of the fuel cell;

[0243] Parameter fine-tuning mechanism: If the constraint is not met, fine-tune by ±0.05. value:

[0244] If SOC < 30% and P bat >0: Increasing k by 0.05 increases P fc Percentage;

[0245] If SOC > 80% and P bat <0: k decreases by 0.05, reducing P fc Percentage;

[0246] If P fc >0.9×P fc_rated Reduce k by 0.05 to avoid overload.

[0247] (2) Fuel cell start-stop control and execution:

[0248] 1. Start-stop control logic:

[0249] Start-up / shutdown target parameters: Air compressor speed target ( ), back pressure valve opening target ( (From SSA optimization results), ensuring start / stop parameters and energy allocation ratios ( Collaboration;

[0250] Real-time feedback signals: cathode pressure, membrane water content, fuel cell output power The state of charge (SOC) of the power battery is used to dynamically correct control commands and avoid overshoot or undershoot. The anode hydrogen pressure needs to be dynamically matched with the cathode air pressure (pressure difference ≤ 10 kPa).

[0251] 2. Execution process:

[0252] 2.1 Start-up execution: Calculate the target power of the air compressor. ( ), Power battery pre-discharge, pre-discharge power =0.8* This ensures uninterrupted power supply when the air compressor is under load; the FCU controls the air compressor speed to increase from 0 to the target value, the back pressure valve opening to increase from 0 to the target value, and the boost controller gradually increases the output voltage. The entire startup process takes ≤2s, and the power fluctuation is ≤5kW.

[0253] 2.2 Shutdown Procedure: The FCU gradually reduces the air compressor speed, and the back pressure valve opening decreases synchronously. When the load drops to 20KW, a prolonged shutdown purging process is performed until the membrane water content decreases to the expected value. After the temperature drops to 0, the shut-off valve is closed, and the shutdown process takes ≤120 seconds.

[0254] 2.3 Power Adjustment: MCU Receiver After receiving the command, the charging and discharging current is adjusted via vector control, with a current change rate ≤10A / s (corresponding to a power change rate ≤5kW / s).

[0255] Performance metrics: Execution response latency ≤100ms (from command issuance to power stabilization within ±0.5kW of the target value).

[0256] In another embodiment of the present invention, a fuel cell predictive start-stop system for coupled control of an automotive powertrain is also provided, for implementing the above-described fuel cell predictive start-stop method, comprising:

[0257] The data acquisition module is used to acquire multi-dimensional data in real time; the multi-dimensional data includes vehicle status data as well as environmental and traffic data.

[0258] The driving trend prediction module is used to input multi-dimensional data into a pre-trained driving trend prediction model and output predicted driving trend information for a preset future time period. The predicted driving trend information includes average vehicle speed prediction results, maximum power demand prediction results, and operating condition type prediction results.

[0259] The demand power calculation module is used to calculate the vehicle's demand power at the current moment based on the preset vehicle dynamics model and the average vehicle speed prediction results.

[0260] The multi-objective optimization module is used to construct a multi-objective optimization function based on the Zunhaishao swarm algorithm, combined with the preset fuel cell start-stop dynamic model and power battery SOC model. It uses the maximum demand power prediction result as the target power reference of the fuel cell and the total vehicle demand power as the total power benchmark. The module then solves the multi-objective optimization function in real time to obtain the optimal start-stop control parameters.

[0261] The start-stop control module is used to generate and execute coordinated start-stop control commands based on the optimal start-stop control parameters.

[0262] It should be noted that each of the above modules can be implemented as a computer program, which can run on a computer device. The computer device's memory can store the computer program that makes up each module, enabling the processor to execute each step of the above method.

[0263] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0264] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A predictive start-stop method for fuel cells in a coupled control system of an automotive powertrain, characterized in that, Includes the following steps: Real-time acquisition of multi-dimensional data; the multi-dimensional data includes vehicle status data as well as environmental and traffic data; Multi-dimensional data is input into a pre-trained driving trend prediction model, which outputs predicted driving trend information for a future preset time period. The predicted driving trend information includes the average vehicle speed prediction result, the maximum power demand prediction result, and the operating condition type prediction result. Based on the preset vehicle dynamics model, the vehicle power demand at the current moment is calculated according to the average vehicle speed prediction results. Based on the tunic algorithm, combined with the preset fuel cell start-stop dynamic model and power battery SOC model, the maximum demand power prediction result is used as the target power reference of the fuel cell, and the vehicle demand power is used as the total power benchmark. A multi-objective optimization function is constructed, and the multi-objective optimization function is solved in real time to obtain the optimal start-stop control parameters. Based on the optimal start-stop control parameters, generate and execute coordinated start-stop control commands.

2. The fuel cell predictive start-stop method for coupled control of an automotive powertrain system according to claim 1, characterized in that, The vehicle's own status data includes historical vehicle speed and historical average acceleration; the environmental and traffic data includes gradient, speed of the vehicle in front, distance to the vehicle in front, remaining time at the traffic light ahead, distance between the vehicle and the refueling station, and waiting time at the refueling station.

3. The fuel cell predictive start-stop method for coupled control of an automotive powertrain system according to claim 2, characterized in that, The driving trend prediction model is constructed using Bi-LSTM, specifically including: Input layer: consists of 8 neurons, with an input vector of... , where v ave For historical vehicle speeds, a avg Let i be the historical average acceleration, i be the slope, and T be the acceleration. light v represents the remaining time of the traffic light ahead. lead For the speed of the vehicle in front, d lead To determine the distance to the vehicle in front, d h2 t is the distance between the vehicle and the refueling station. queue Queue times at refueling stations; Hidden layers: consist of 2 layers, each with 64 neurons, using ReLU or Softmax activation functions, with a Dropout layer between the two layers; Output layer: consists of 3 neurons, with an output vector of... , where v pred For the prediction of future average vehicle speed, P req_max The forecast result for the maximum future demand power is denoted by type, which represents the forecast result for the operating condition type. The driving trend prediction model learns by iteratively optimizing parameters and mastering the correspondence between input features and patterns. During training, training set data is processed in batches and input into the driving trend prediction model. After each batch of data undergoes forward propagation, the prediction results are compared with the actual labels, and a hybrid loss function L is used. total The calculation method is as follows: ； in, The mean squared error is used to optimize regression tasks and is calculated as follows: ； In the formula, This represents the number of samples in the training set. These are the actual vehicle speed and power, respectively. Cross-entropy loss is used to optimize operating conditions, and is calculated as follows: In the formula, Representing the The first sample A real label, This refers to the first The corresponding sample of the th sample The predicted probability of the item; These are the weighting coefficients; The backpropagation algorithm is used to determine the gradient of the hybrid loss function with respect to the parameters of the driving trend prediction model. Then, the optimizer is used to adjust the parameters of the driving trend prediction model according to the gradient. This process is repeated until the predetermined training cycle is completed. After each training cycle, the training effectiveness of the driving trend prediction model is evaluated by calculating the loss value. After training the driving trend prediction model, test set data is input into the trained driving trend prediction model to obtain prediction results, which are then compared with actual data to evaluate the accuracy detection and recognition performance of the driving trend prediction model.

4. The fuel cell predictive start-stop method for coupled control of an automotive powertrain system according to claim 1, characterized in that, The formula for the vehicle dynamics model is as follows: ； In the formula, For the power required by the whole vehicle, For the efficiency of the transmission system, The angle between the plane where the vehicle is located and the horizontal plane. For vehicle quality, It is the acceleration due to gravity. The rolling resistance coefficient, For vehicle speed, The air drag coefficient, For windward area, This is the rotational mass conversion factor.

5. The fuel cell predictive start-stop method for coupled control of an automotive powertrain system according to claim 1, characterized in that, The fuel cell start-stop dynamic model synchronously simulates voltage output characteristics, gas path pressure dynamics, and auxiliary system energy consumption. The auxiliary system includes an air compressor. The fuel cell voltage is modeled as a function of current density, cathode oxygen partial pressure, and membrane water content, as shown in the following equation: ； In the formula, Refers to the voltage of the fuel cell; It is activation loss. It is an ohm loss. It is a concentration loss. It is the number of stacked units; It is the Nernst potential, which depends on the temperature of the reactants and products. And partial pressure: ； In the formula, This indicates the partial pressure of hydrogen at the anode of the fuel cell. This indicates the partial pressure of oxygen at the cathode of the fuel cell; The dynamic model of the airflow pressure consists of three parts: the intake manifold, the cathode, and the exhaust manifold. The intake manifold connects the air compressor outlet to the cathode inlet. Based on the linear nozzle principle, the outlet flow rate of the intake manifold is proportional to the difference between the pressure inside the manifold and the cathode pressure, as shown in the following formula: ； In the formula, This refers to the mass flow rate at the intake manifold outlet. This is the intake manifold flow coefficient. This refers to the pressure inside the intake manifold. This refers to the pressure inside the cathode cavity; Based on the ideal gas law and the law of conservation of mass, the pressure inside the intake manifold dynamically changes with the mass difference between the intake and exhaust gases, as shown in the following formula: ； In the formula, Let be the ideal gas constant of air. For the intake manifold volume, This refers to the mass flow rate at the air compressor outlet. The outlet gas temperature of the air compressor. This refers to the temperature of the gas inside the intake manifold. The exhaust manifold connects the cathode outlet to the outside. After the exhaust gas flows from the cathode into the exhaust manifold, the outflow is regulated by the back pressure valve, resulting in a pressure difference change within the manifold, as shown in the following formula: ； In the formula, This refers to the pressure inside the exhaust manifold. The temperature of the gas inside the exhaust manifold. For exhaust manifold volume, This represents the cathode outlet mass flow rate. This refers to the mass flow rate at the exhaust manifold outlet. Based on the principle of adiabatic compression, the formula for the energy consumption of the auxiliary system is as follows: ； In the formula, For air compressor energy consumption, This refers to the intake molar flow rate of the air compressor. For the molar mass of air, The specific heat capacity of air at constant pressure. This refers to the inlet air temperature of the air compressor. For air compressor efficiency. Where c is the inlet pressure of the air compressor, and c is the air adiabatic index. This refers to the intake manifold pressure.

6. The fuel cell predictive start-stop method for coupled control of an automotive powertrain system according to claim 1, characterized in that, The power battery SOC model is constructed based on the equivalent circuit model method; wherein, the power battery output power for: ； In the formula, Represents open-circuit voltage. Indicates battery current. The internal resistance of the power battery; battery current. The calculation formula is as follows: ； The SOC value of the power battery at the current moment is calculated using the ampere-hour integral method, as shown in the following formula: ； In the formula, Q represents the current SOC value of the power battery; bat This refers to the battery capacity.

7. The fuel cell predictive start-stop method for coupled control of an automotive powertrain system according to claim 6, characterized in that, The multi-objective optimization function takes the vehicle energy consumption as an example. Battery power fluctuation and power control overshoot To optimize the target, start-stop control parameters include energy distribution ratio. Air compressor speed Back pressure valve opening ;in, Power required for the entire vehicle; This refers to the output power of the fuel cell.

8. The fuel cell predictive start-stop method for coupled control of an automotive powertrain system according to claim 7, characterized in that, Based on the tunic swarm algorithm, combined with a pre-defined fuel cell start-stop dynamic model and a power battery SOC model, the steps include: using the maximum demand power prediction result as the target power reference for the fuel cell and the vehicle demand power as the total power benchmark; constructing a multi-objective optimization function; and solving the multi-objective optimization function in real time to obtain the optimal start-stop control parameters. (1) Chaotic initialization of the population: Population size N=30; dimension D=3, corresponding to the optimal energy allocation ratio Optimal air compressor speed Optimal back pressure valve opening ; Generate a uniformly distributed initial population using an improved Tent chaotic map: In the formula, , A random number in the range [0,1]. It is a chaotic sequence; Mapping chaotic sequences to the solution space : In the formula, , Define the upper and lower bounds of the variable; (2) Randomly generate the initial population and construct a multi-objective optimization function J: ； In the formula, Let be the weighting coefficient, satisfying ; The constraints st are as follows: ； In the formula, Let be the output power of the fuel cell at time t. The minimum permissible fuel cell output power, The maximum permissible output power of the fuel cell; Let be the output power of the power battery at time t. The minimum permissible power battery output power, The maximum permissible power battery output power; Let t be the SOC value of the power battery at time t. The minimum allowable SOC value, The maximum allowed SOC value; Among them, the energy consumption of the whole vehicle The calculation formula is as follows: ； In the formula, For the hydrogen consumption of fuel cells, This refers to the equivalent hydrogen consumption of the power battery. Battery power fluctuation The calculation formula is as follows: ； In the formula, This is the SOC fluctuation penalty coefficient. This is the penalty coefficient for fuel cell power fluctuations; Power control overshoot The calculation formula is as follows: ； In the formula, Real-time power of the fuel cell; Determined from the maximum demand power prediction results; (3) Leader-follower iterative optimization: Leaders renew: ； In the formula, For the upper realm, The lower bound is given by , rand is a random number in the range [0,1], and l is the number of iterations. calculate and If the former is smaller, then update the leader position; followers renew: Introducing networked data feedback: predicting the gradient of the incline. hour, The search weight increased from 1.0 to 1.5; (4) Convergence judgment: The fitness error is determined when the number of iterations reaches T=50, or after three consecutive iterations. Stop when the time is right and output the optimal start / stop control parameters.

9. The fuel cell predictive start-stop method for coupled control of an automotive powertrain system according to claim 8, characterized in that, The steps for generating and executing coordinated start-stop control commands based on optimal start-stop control parameters specifically include: Power allocation is calculated based on optimal start-stop control parameters to ensure dynamic balance under multiple constraints; The power calculation formula is as follows: ； Multi-constraint control logic: (1) Discharge constraint, The discharge power change rate must be ≤5kW / s and the SOC must be ≥30%. (2) Charging constraints, The following conditions must be met: SOC ≤ 80%, and charging power change rate ≤ 5kW / s; (3) Constraints of fuel cells: Need to be Within the range; Rated power of the fuel cell; Parameter fine-tuning mechanism: If the constraint is not met, fine-tune by ±0.

05. value: If SOC < 30% and P bat >0: Increasing k by 0.05 increases P fc Percentage; If SOC > 80% and P bat <0: k decreases by 0.05, reducing P fc Percentage; If P fc >0.9×P fc_rated Reduce k by 0.05 to avoid overload.

10. A fuel cell predictive start-stop system for coupled control of an automotive powertrain, used to implement the fuel cell predictive start-stop method according to any one of claims 1-8, characterized in that, include: The data acquisition module is used to acquire multi-dimensional data in real time. The multi-dimensional data includes vehicle status data as well as environmental and traffic data; The driving trend prediction module is used to input multi-dimensional data into a pre-trained driving trend prediction model and output predicted driving trend information for a future preset time period. The predicted driving trend information includes the average vehicle speed prediction result, the maximum power demand prediction result, and the operating condition type prediction result. The demand power calculation module is used to calculate the vehicle's demand power at the current moment based on the preset vehicle dynamics model and the average vehicle speed prediction results. The multi-objective optimization module is used to construct a multi-objective optimization function based on the Zunhaishao swarm algorithm, combined with the preset fuel cell start-stop dynamic model and power battery SOC model. It uses the maximum demand power prediction result as the target power reference of the fuel cell and the total vehicle demand power as the total power benchmark. The module then solves the multi-objective optimization function in real time to obtain the optimal start-stop control parameters. The start-stop control module is used to generate and execute coordinated start-stop control commands based on the optimal start-stop control parameters.

Images